0% found this document useful (0 votes)

14 views52 pages

Yang 2015

This article provides a review of the past 50 years of research on MIMO detection methods, from 1965 to 2015. It discusses the fundamentals and challenges of MIMO detection, including different channel and system models. It then reviews representative MIMO detection methods over time and draws lessons for designing scalable algorithms for large-scale MIMO systems. Finally, it discusses detection strategies for two types of large-scale MIMO and reviews recent advances in the field.

Uploaded by

Sofia Bouchenak

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

14 views52 pages

Yang 2015

Uploaded by

Sofia Bouchenak

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 52

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/COMST.2015.2475242, IEEE Communications Surveys & Tutorials
SUBMITTED TO IEEE COMMUNICATIONS SURVEYS & TUTORIALS, 2015 1

Fifty Years of MIMO Detection: The Road to

Large-Scale MIMOs
Shaoshi Yang, Member, IEEE and Lajos Hanzo, Fellow, IEEE

Abstract—The emerging massive/large-scale multiple-input ACO ant colony optimization.

multiple-output (LS-MIMO) systems that rely on very large an- AME asymptotic-multiuser-efficiency.
tenna arrays have become a hot topic of wireless communications. APP a posteriori probability.
Compared to multi-antenna aided systems being built at the time ASIC application-specific integrated circuit.
of writing, such as the long-term evolution (LTE) based fourth AWGN additive white Gaussian noise.
BALM block alternating likelihood maximization.
generation (4G) mobile communication system which allows for BC-SDPR bound-constrained semidefinite programming relaxation.
up to eight antenna elements at the base station (BS), the LS- BER bit-error rate.
MIMO system entails an unprecedented number of antennas, BI-GDFE block-iterative generalized decision feedback equalizer.
say 100 or more, at the BS. The huge leap in the number of BS BLER block-error rate.
antennas opens the door to a new research field in communication BP belief propagation.
theory, propagation and electronics, where random matrix theory BPSK binary phase-shift keying.
begins to play a dominant role. Interestingly, LS-MIMOs also BS base station.
constitute a perfect example of one of the key philosophical CAGR compound annual growth rate.
principles of the Hegelian Dialectics, namely that “quantitative CCI co-channel interference.
CDM code-division multiplexing.
change leads to qualitative change”. CDMA code-division multiple-access.
In this treatise, we provide a recital on the historic her- CLPS closest lattice-point search.
itages and novel challenges facing LS-MIMOs from a detection CMOS complementary metal-oxide semiconductor.
perspective. Firstly, we highlight the fundamentals of MIMO CPDA complex-valued probabilistic data association.
detection, including the nature of co-channel interference (CCI), CR cognitive radio.
the generality of the MIMO detection problem, the received signal DFD decision-feedback detector.
models of both linear memoryless MIMO channels and dispersive DMT diversity-multiplexing tradeoff.
MIMO channels exhibiting memory, as well as the complex- DS-CDMA direct-sequence code-division multiple-access.
DSNR decreasing signal-to-noise ratio.
valued versus real-valued MIMO system models. Then, an
EB exabytes.
extensive review of the representative MIMO detection methods EM expectation-maximization.
conceived during the past fifty years (1965-2015) is presented, and EXIT extrinsic information transfer.
relevant insights as well as lessons are inferred for the sake of FCSD fixed-complexity sphere decoding/decoder.
designing complexity-scalable MIMO detection algorithms that FDM frequency-division multiplexing.
are potentially applicable to LS-MIMO systems. Furthermore, FDMA frequency-division multiple-access.
we divide the LS-MIMO systems into two types, and elaborate FEC forward-error-correction.
on the distinct detection strategies suitable for each of them. The FER frame-error rate.
type-I LS-MIMO corresponds to the case where the number of FH-CDMA frequency-hopped code-division multiple-access.
FIR finite impulse response.
active users is much smaller than the number of BS antennas,
GA genetic algorithm.
which is currently the mainstream definition of LS-MIMO. The GSNR greatest signal-to-noise ratio.
type-II LS-MIMO corresponds to the case where the number HNN Hopfield neural network.
of active users is comparable to the number of BS antennas. IAI interantenna interference.
Finally, we discuss the applicability of existing MIMO detection IC integrated circuit.
algorithms in LS-MIMO systems, and review some of the recent ICI interchannel interference.
advances in LS-MIMO detection. IDD iterative detection and decoding.
IMSE increasing mean-square error.
Index Terms—Co-channel interference (CCI), equalization, ISI intersymbol interference.
large-scale/massive MIMO, multiuser detection, MIMO detec- JPDA joint probabilistic data association.
tion. LAS likelihood ascent search.
LDPC low-density parity-check.
LLL Lenstra-Lenstra-Lovász.
G LOSSARY LMSE least mean-square error.
LR lattice-reduction.
3G third generation.
LS least-squares.
4G fourth generation.
LS-MIMO large-scale multiple-input multiple-output.
5G fifth generation.
LSD list sphere decoding.
A-CPDA approximate complex-valued probabilistic data association.
LTE long-term evolution.
LTE-A Long Term Evolution-Advanced.
M2M machine-to-machine.
The financial support of the Research Councils UK (RCUK) under the MAI multiple-access interference.
India-UK Advanced Technology Center (IU-ATC), of the EU under the MAME maximum asymptotic-multiuser-efficiency.
auspices of the Concerto project, and of the European Research Councils MAP maximum a posteriori.
Advanced Fellow Grant is gratefully acknowledged. MBER minimum bit error rate.
The authors are with the School of Electronics and Computer Science, MC-CDMA multicarrier code-division multiple-access.
University of Southampton, Southampton, SO17 1BJ, UK (e-mail: {lh, MED minimum Euclidean distance.
shaoshi.yang}@ecs.soton.ac.uk). MF matched filter.

1553-877X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/COMST.2015.2475242, IEEE Communications Surveys & Tutorials
SUBMITTED TO IEEE COMMUNICATIONS SURVEYS & TUTORIALS, 2015 2

MFB matched filter bound.

MFSK multiple frequency-shift keying.
MGS mixed Gibbs sampling.
MIC multistage interference cancellation.
ML maximum likelihood.
MMF multimode fibre.
MMSE minimum mean-square error.
MR multiple restart.
MS mobile station.
MSDSD multiple symbol differential sphere decoder.
MSE mean-square error.
MSI multiple-stream interference.
MUD multiuser detection/detector.
MUI multiuser interference.
NFV network function virtualization.
NP-complete nondeterministic polynomial-time complete.
NP-hard nondeterministic polynomial-time hard.
OFDM orthogonal frequency-division multiplexing.
OFDMA orthogonal frequency-division multiple-access.
OSIC ordered successive interference cancellation.
PAM pulse-amplitude modulation.
PDA probabilistic data association.
PDF probability density function.
PER packet-error rate.
PI-SDPR polynomial-inspired semidefinite programming relaxation. Fig. 1. Cisco VNI: global mobile data traffic forecast, 2013-2018.
PIC parallel interference cancellation.
PSK phase-shift keying.
PSO particle swarm optimization.
QAM quadrature amplitude modulation.
(VNI) forecast [6], [7]. More specifically,1 the global mobile
QPSK quadrature phase-shift keying. data traffic grew 81% in 2013, up from 0.82 exabytes (EB),
QRD QR-decomposition. i.e. 0.82 × 1018 bytes per month at the end of 2012 to 1.5
R-MCMC randomized Markov chain Monte Carlo.
RPDA real-valued probabilistic data association.
EB per month at the end of 2013; furthermore, as predicted
RS randomized search. in Fig. 1, it will increase nearly 11-fold between 2013 and
RTS reactive tabu search. 2018, which translates to a compound annual growth rate
SA simulated annealing.
SC-FDMA single-carrier frequency-division multiple-access.
(CAGR) of 61% for the period spanning from 2013 to 2018,
SD sphere decoding/decoder. reaching 15.9 EB per month by 2018 [6], [7]. As seen from
SDM space-division multiplexing. Fig. 2, this explosive growth is mainly fuelled by the preva-
SDMA space-division multiple-access.
SDN software-defined network.
lence of smartphones, laptops and tablets, as well as by the
SDP semidefinite programming. emergence of machine-to-machine (M2M) communications
SDPR semidefinite programming relaxation. [8]–[19]. Additionally, the design of wireless communication
SDR software-defined radio.
SE Schnorr-Euchner.
systems is highly constrained by the paucity of radio spectrum,
SER symbol-error rate. which is exemplified by the overcrowded frequency allocation
SIC successive interference cancellation. chart of the United States [20]. As a consequence of the
SINR signal-to-interference-plus-noise ratio.
SIR signal-to-interference ratio.
combined effect of the mobile data traffic growth trend and the
SNR signal-to-noise ratio. scarcity of favorable radio spectrum in the low-loss frequency-
STBC space-time block code/coded. range, the forthcoming fifth generation (5G) communication
STSK space-time shift keying.
SUMF single-user matched filter.
systems have to resort to the employment of massive/large-
SUMIS subspace marginalization aided interference suppression. scale multiple-input multiple-output (LS-MIMO) transmission
TDM time-division multiplexing. techniques, which invoke a large number of antenna elements
TDMA time-division multiple-access.
UEP unequal error protection.
at the transmitter and/or receiver for achieving a high spectral-
VA-SDPR virtually antipodal semidefinite programming relaxation. efficiency [21]–[29] and high energy-efficiency [25], [30]–
VB Viterbo-Boutros. [32].
VBLAST vertical Bell Laboratories layered space-time.
VER vector-error rate.
A range of other fundamental technologies conceived for
VLSI very-large-scale integration. 5G communications are closely related to LS-MIMO. For
VNI visual network index. example, both millimetre wave communications [33] and LS-
WLS weighted least-squares.
ZF zero-forcing.
MIMOs may be regarded as enabling techniques facilitating
high-dimensional physical-layer communication technologies.
Their difference is that LS-MIMOs achieve high dimension-
ality in the spatial domain, while millimetre wave communi-
I. I NTRODUCTION cation systems achieve a high dimensionality in the frequency
domain by operating at frequencies ranging from about 30
A. Why are Large-Scale MIMOs Important? GHz to 300 GHz, which is much higher than the operating
The multimedia data traffic conveyed by the global mobile frequencies of contemporary third generation (3G)/4G systems
networks has been soaring [1]–[5], and this trend is set to 1 Fig. 1 and Fig. 2 are reprinted from the Cisco VNI white paper [6], with
continue, as indicated by Cisco’s visual networking index permission of Cisco.

Fig. 2. Cisco VNI: global mobile devices and connections growth forecast, 2013-2018.

Inputs s Noise n Outputs y Estimates ŝ

Small−Cell s1 n1 y1 ŝ1
based s2 n2 y2 MIMO ŝ2
Wireless Channel Matrix
Heterogeneous Networks .. .. .. Detector ..
. . . .
H
s NI n NO y NO ? ŝNI

Network Architecture

Fig. 4. Conceptual illustration of the MIMO detection problem.

Large−Scale MIMO Millimetre Wave

based based
Page 7 of
Signal Processing Wireless Communications
[34]. Compared to conventional single-input single-output sys-
tems, e.g. the single-antenna point-to-point system, in MIMO
Physical Layer Technologies
systems we have multiple interfering messages/symbols trans-
mitted concurrently, and at the receiver these symbols are
expected to be detected/decoded subject to the contamination
of random noise or interference, as shown in Fig. 4. The
multiple symbols may be detected separately or jointly. As
opposed to separate detection, in joint detection each symbol
has to be detected taking into account the characteristics of the
Fig. 3. There is a natural marriage amongst LS-MIMO, millimetre wave,
and small-cell based HetNet, which constitute three fundamental technologies other symbols. As a beneficial result, typically joint detection
for 5G wireless communications. is capable of achieving a significantly better performance than
separate detection, although joint detection imposes a higher
computational complexity.
(from 450 MHz to 3.5 GHz). Furthermore, owing to the
much shorter wavelength, millimetre wave technologies may The joint detection of multiple symbols in MIMO systems
facilitate compacting a large number of antenna elements in is of central importance for the sake of realizing the substantial
a relatively small space. Additionally, the coverage area of a benefits of various MIMO techniques. This is because the co-
single cell of millimetre wave communication systems may channel interference (CCI) routinely encountered in MIMO-
be significantly smaller than a single cell of 3G/4G systems. based communication systems constitutes a fundamental lim-
As a result, small-cell based heterogeneous network (HetNet) iting characteristic [24], [35]–[44]. Unfortunately, the opti-
architecture is required. Therefore, as shown in Fig. 3, there is mum MIMO detection problem was proven non-deterministic
a natural marriage amongst LS-MIMO, millimetre wave and polynomial-time hard (NP-hard) [45]–[47], thus all known
small-cell technologies. algorithms conceived for solving the problem optimally have
a complexity exponentially increasing with the number of
decision variables. As a result, the computational complexity
B. Why is MIMO Detection Important and Challenging? of the optimum maximum-likelihood (ML) criterion or the
As Claude Shannon pointed out, “The fundamental problem maximum a posteriori (MAP) criterion based MIMO detec-
of communication is that of reproducing at one point either tion algorithms quickly become excessive as the number of
exactly or approximately a message selected at another point” decision variables increases. Thanks to the rapid develop-

ment of the semiconductor industry, the hardware computing has a significantly higher number of antennas than the other.
power has been dramatically increasing over the years, and in By contrast, only limited attention has been dedicated to
some cases a “not-so-extreme” computational complexity is the uplink of general LS-MIMOs. Hence, our goal is to fill
no longer regarded as a bottleneck of practical applications. these gaps in the open literature. For the sake of clarity, the
However, it should be noted that while transistors get faster organization of this paper is shown in Fig. 5.
and smaller, supply voltages cannot be reduced significantly in
modern complementary metal-oxide semiconductor (CMOS)
II. T HE NATURE OF C O -C HANNEL I NTERFERENCE
processes. Therefore, virtually all modern integrated circuits
(ICs) encounter an integration density limit owing to the max- To gain profound insights into the intricacies of the MIMO
imum tolerable internal temperature imposed by the excessive detection problem, let us briefly reflect on the nature of the
power consumption or power density. In other words, this CCI in this section. The nature of CCI depends on the specific
power bottleneck still limits today’s IC development. As a context. In this paper, it is defined in its most generic form
consequence, one cannot simply rely on Moore’s law, and as the interfering signal imposed by multiple transmissions
even modest-complexity MIMO detection algorithms could be taking place on channels which are mutually non-orthogonal.
too power hungry for battery-powered devices. Hence low- Mathematically, CCI may also be interpreted as interfering
complexity, yet high-performance suboptimum MIMO detec- signals that span a subspace having a “non-empty” intersection
tion algorithms are needed for practical MIMO applications. with the subspace spanned by the desired signals. The channel-
induced non-orthogonality may be observed in the frequency,
C. The Contributions of This Paper time and/or space domain, as shown in Fig. 6. To recover
the desired signal at the receiver, the desired signal has to be
In this paper, an extensive review of the family of repre- distinguishable from the interference in at least one domain.
sentative MIMO detection methods invented during the past In the extreme case, if the multiple transmissions are highly
fifty years is presented in a unified mathematical model2 , non-orthogonal in all domains, then it may become impossible
although practical MIMO schemes have various subtleties. Our to recover the desired signal by any means.
particular focus is on complexity-scalable MIMO detection In essence, the CCI originates from signal-feature-
algorithms potentially applicable to LS-MIMO systems [25]. overlapping of multiple transmissions. For example, in
The algorithms surveyed include the classic linear MIMO de- spectrum-efficient communication systems such as the code-
tection, the interference-cancellation based MIMO detection, division multiplexing /multiple-access (CDM/CDMA) systems
the tree-search based MIMO detection, the lattice-reduction [149]–[151] and the space-division multiplexing /multiple-
(LR) aided MIMO detection, the probabilistic data association access (SDM/SDMA) systems [21], [44], [152]–[158], mul-
(PDA) based MIMO detection, and the semidefinite program- tiple transmissions are often deliberately arranged to take
ming relaxation (SDPR) based MIMO detection. Several high- place simultaneously over the same frequency band. These
quality books or reviews were published on MIMO detection “frequency sharing” and “time sharing” strategies result in
[48]–[52]. They were predominantly dedicated to CDMA a “frequency-overlapping” and a “time-overlapping” phe-
systems in the 1990s [48]–[50] or to conventional small- nomenon, respectively. It is worth pointing out that as far as
/medium-scale MIMO systems [51], [52], whereas LS-MIMOs radio waves are concerned, rigorously the CCI always tends to
just became a hot research topic at the time of writing [25]. exist in the frequency, time and space domains. For example,
On the other hand, they mainly covered the most common when no deliberate frequency-overlapping is arranged, the
suboptimum MIMO detection methods, such as the linear zero- “frequency-overlapping” is due to the underlying fact that for
forcing (ZF) detector, the linear minimum mean-square error all realizable, time-limited radio waveforms, their absolute
(MMSE) detector and various interference cancellation based bandwidth is infinite [159], [160], as shown in Fig. 7. In
detectors [48]–[51], or focused largely on a single type of other words, every active radio transmitter has an impact
MIMO detector [52]. In comparison, there is a paucity of on every operating radio receiver. Similarly, for a strictly
reviews on more advanced MIMO detection methods, such bandwidth-limited signal, its time duration has to be infinite.
as the tree-search based MIMO detectors (the sphere decoder With respect to the space domain, it is well known that
(SD) constitutes an instance of the tree-search based MIMO the propagation of electromagnetic energy in free space is
detectors) [53]–[82], the LR based MIMO detectors [83]– determined by the inverse square law [21], [158], [161], i.e.
[103], as well as the PDA [104]–[133] and the SDPR [134]– we have S = Pt /4πd2 , where S is the power per unit area or
[147] based detectors etc., although a concise tutorial on power spatial density (in Watts per metre-squared) at distance
some of these detectors was given in [148]. Additionally, d, and Pt is the total power transmitted (in Watts). Hence,
most of the existing research on LS-MIMO is focused on theoretically, the radio signals cannot be stopped, they are only
the precoding/beamforming based downlink of a special case attenuated in the frequency, time and space domains.
of LS-MIMO, where one side of the communication link In engineering practice, fortunately, by using well-designed
2 This means that the algorithms conceived for the equalization, multiuser filters [162], [163], typically the waveform of the time-limited
detection and multi-antenna detection problems can be treated under the same signal can be shaped so that most energy of the signal can
umbrella of the MIMO detection model of (1). More discussions on the be kept within a given limited frequency-band, and thus the
similarities and differences amongst these three problems are provided in
Section III and Section IV, as well as are found in the first two paragraphs signal energy leakage outside the target frequency-band can
of Section VIII and the last but one paragraph of Section VIII-B. be reduced to a sufficiently low level. Similarly, in the space

Sec. I − Introduction
Sec. II − The Nature of CCI

Sec. III − Concept and Generality of MIMO Detection

Sec. IV − Formal Definition of the MIMO Detection Problem

Part I: Fundamentals of MIMO Detection
Sec. V − MIMO System Model for Linear Memoryless Channels

Sec. VI − MIMO System Model for Dispersive Channels Exhibiting Memory

Sec. VII − Complex−Valued versus Real−Valued MIMO System Models

Optimum MIMO Detector

Linear MIMO Detectors

Interference Cancellation Aided MIMO Detectors

Tree−Search Based MIMO Detectors

Lattice−Reduction Aided MIMO Detectors

Part II: Sec. VIII − History and State−of−the−Art of MIMO Detection
Probabilistic Data Association (PDA) Based MIMO Detectors

Semidefinite Programming Relaxation (SDPR) Based MIMO Detectors

Detection in Rank−Deficient and Overloaded MIMO Systems

Impact of Soft Decision and Transmit Preprocessing on MIMO Detection

Guidelines on Choosing the Right MIMO Detectors

Detection in Single−Cell/Noncooperative Multi−Cell LS−MIMO Systems

Detection in Cooperative Multi−Cell LS−MIMO Systems

Part III: Sec. IX − Detection in LS−MIMO Systems Applicability of Existing MIMO Detection Algorithms to LS−MIMO Systems

Recent Advances in LS−MIMO Detection

Applications of MIMO Detection Techniques in Other Areas

Sec. X − Summary and Conclusions

Fig. 5. The organization of this paper.

in these three domains is a result of a deliberate design. In

this context, the signals are made as much distinguishable as
Space
possible in one domain, and as much overlapping as possible
in the remaining domains. Our task is to recover the desired
signal based on this deliberate arrangement.
Since the frequency, time and space domains represent the
fundamental physical features of signal transmission, each
of them corresponds to a distinct multiplexing/multiple-
access scheme [164], namely the frequency-division
multiplexing/multiple-access (FDM/FDMA), time-
division multiplexing/multiple-access (TDM/TDMA), and
SDM/SDMA, respectively. All of these multiple-access
techniques aim for avoiding CCI by orthogonalizing the
channel access in a certain domain. It is worth noting
Frequency that compared to these three fundamental domains, the
spreading code sequences used in CDM/CDMA systems do
Time not constitute an independent domain. This is because the
orthogonality of the spreading code sequences is essentially a
Fig. 6. The multiple signals have to be distinguishable in at least one of the special case of time-domain orthogonality. In systems using
three fundamental domains of time, frequency and space. spreading codes, in principle we pursue to transmit orthogonal
code sequences to minimize the inter-code interference.
Although it is mathematically possible to construct perfectly
domain, two transmissions taking place at a sufficiently far dis- orthogonal code sequences, the orthogonality of these code
tance can also be regarded as non-interfering with each other. sequences is typically degraded in practical transmissions
Therefore, despite the fact that the signal-feature-overlapping [151]. Moreover, since the number of theoretically orthogonal
in frequency, time and space domains is inevitable from the code sequences is rather limited, often quasi-orthogonal code
theoretic point of view, in practical spectrum-efficient systems sequences are adopted in practice [151]. Therefore, typically
we can typically assume that the signal-feature-overlapping substantial interference is imposed by the non-orthogonality

Rectangular Aperiodic Pulse is imposed by the non-orthogonality of “spreading codes” in

1.2
practical space-division systems [21], [44], [152]–[158]. This
is similar to the case in CDM/CDMA systems and explains
1
why MIMO detection typically represents a more significant
0.8
problem in CDM/CDMA systems and SDM/SDMA systems
than in FDM/FDMA and TDM/TDMA systems.
Additionally, there are more advanced multicarrier based
Amplitude

0.6
orthogonal multiple-access techniques, such as orthogo-
0.4 nal frequency-division multiple-access (OFDMA) [166]–
[168], single-carrier frequency-division multiple-access (SC-
0.2 FDMA)3 [169], and multicarrier CDMA [170]–[177]. Despite
their potential advantages in averaging interference over dif-
0 ferent subcarriers for different users, they usually suffer high
sensitivity to frequency offset, which leads to intercarrier in-
−0.2 terference. Therefore, judicious frequency offset compensation
−2 −1 0 1 2
Time (sec) scheme and frequency reuse scheme have to be designed to
minimize the intercarrier interference.
(a) Time-domain waveform In this paper, the CCI considered mainly refers to the
Sinc Function: Frequency Response of Rectangular Aperiodic Pulse interference in SDM/SDMA or CDM/CDMA systems, where
multiple transmissions often take place simultaneously, or
1
partially simultaneously over the same frequency. Depending
on specific applications, CCI is often alternatively termed as
0.8
intersymbol interference (ISI), interchannel interference (ICI),
0.6
interantenna interference (IAI), multiuser interference (MUI),
multiple-access interference (MAI), and multiple-stream inter-
Amplitude

0.4 ference (MSI) etc.

0.2 III. C ONCEPT AND G ENERALITY OF MIMO D ETECTION

0 ··· ··· As a family of general techniques for physical-layer CCI
management, MIMO detection deals with the joint detection
−0.2 of several information-bearing symbols transmitted over a
communication channel having multiple inputs and multiple
−0.4
−10 −5 0 5 10
outputs. This problem is of fundamental importance for mod-
Frequency (Hz) ern high-throughput digital communications. Rigorously, the
MIMO detection problem arises if and only if the respective
(b) Frequency-domain response
subchannels of the multiple inputs are not orthogonal to
Fig. 7. All realizable, duration-time limited waveforms has a infinite each other, and hence there exists interference between the
frequency band. outputs. As a generic mathematical model, the MIMO detec-
tion problem underpins numerous relevant applications, while
the physical meaning of the inputs and outputs herein may
of spreading code sequences in practical CDM/CDMA vary in different contexts. For instance, in band-limited ISI
systems [165]. channels requiring equalization, the inputs refer to a sequence
of symbols interfering with each other, and the outputs are
In analogy to CDM/CDMA, the conventional frequency-
the signals received within a given observation window in
division pattern, time-division pattern and space-division pat-
the time/frequency domain [178]. In single-user SDM-MIMO
tern can also be regarded as a special case of “spreading
systems equipped with multiple transmit and receive antennas
codes” in the frequency domain, time domain and space
[179]–[181], the inputs refer to the vector of modulated
domain, respectively. Note, however, that there exist a certain
symbols that are transmitted from multiple collocated transmit
degree of differences in terms of their multiplexing/multiple-
antennas, while the outputs refer to the vector of received
access resolution in these three domains. More specifically, in
signals recorded at multiple collocated receive antennas, as
practice, by using guard intervals in the corresponding domain,
shown in Fig. 8. This is indeed a canonical scenario of inves-
a good resolution of frequency-division and time-division may
tigating MIMO detection algorithms [148]. A third example
be readily maintained – in other words, the orthogonality of
is the uplink of multiuser multiple-antenna systems [182],
“spreading codes” in frequency- and time-division systems
[183], where the inputs may be multiple transmitted symbols
may be relatively easy to obtain [164]. By contrast, the
belonging to a cluster of geographically distributed single-
resolution of space-division tends to be undermined by the
physical size of transmitters/receivers and by the random 3 Note that the so-called SC-FDMA is actually a multicarrier multiple-access
propagation channel, hence typically substantial interference technique, although somewhat misleadingly it has “single-carrier” in its name.

antenna mobile stations (MSs), and the outputs may be the Noise

signals received at the serving base station (BS) equipped

multiple outputs
multiple inputs
with multiple collocated antennas, as shown in Fig. 9. This is

Distributed
Colocated
actually the so-called SDMA system [21], [44], [152]–[158]. ... Channel ...
Yet another important example represented by Fig. 9 is the
uplink of CDMA systems [48]–[51], where the inputs are the
transmitted symbols of distributed single-antenna MSs, and the
outputs are typically generated by filtering the signal received Fig. 10. MIMO broadcast channel, which is not suitable for MIMO detection.
at the single-antenna BS with a bank of matched filters (MFs),
whose impulse responses are matched to a set of a priori
Noise
known user-signature waveforms.
Here, it should be emphasized that whether the multiple

multiple outputs
multiple inputs
W [1] Ŵ [1]

Distributed

Distributed
inputs and/or the multiple outputs are “collocated or not” W [2] Ŵ [2]
is extremely important in determining the signal processing ... Channel ...
[K]
techniques to be used. If multiple inputs/outputs are collocated, W Ŵ [K]
the cooperative joint encoding/decoding of the inputs/outputs
can be conducted [21], [44], [158], [184]–[190], which renders
Fig. 11. MIMO interference channel, where W [k] represents the message
joint MIMO transmission/detection feasible. For example, the that originates at transmitter i and is intended for its dedicated receiver i,
single-user MIMO system shown in Fig. 8 has both its transmit while Ŵ [i] denotes the recovered version of W [k] , k = 1, · · · , K.
and receive antennas collocated, hence it enjoys the privilege
of performing both joint encoding and joint decoding. As
a benefit, both simultaneous transmission and simultaneous geographically distributed, the MIMO channel turns into either
reception can be attained relatively simply. By contrast, the an interference channel [197]–[205] or an X channel [206]–
multiple-access MIMO system of Fig. 9 is typically not [209], which are shown in Fig. 11 and Fig. 12, respectively.
capable of joint encoding at the user side, hence the uplink An interference channel characterizes a situation where each
transmissions of both CDMA and SDMA systems are asyn- transmitter, potentially equipped with multiple antennas, only
chronous by nature. wants to communicate with its dedicated receiver, and each
Additionally, as far as the downlink of multiuser MIMO receiver, possibly equipped with multiple antennas as well,
systems, namely the multiuser MIMO broadcast channel of only cares about the information arriving from the correspond-
Fig. 10 is concerned, typically most of the sophisticated signal ing transmitter. There is a strict one-to-one correspondence
processing tasks are conducted in the form of transmit pre- between the multiple transmitters and the multiple receivers.
processing (i.e. precoding) at the BS, where collocated inputs Therefore, each transmission link interferes with the others. By
are available for cooperative joint encoding [191]–[196]. As a comparison, in the MIMO X channel relying on L transmitters
result, detection at the user becomes less challenging. Since and K receivers, each transmitter has an independent message
the investigation of MIMO transmit preprocessing techniques for each receiver. Hence, there are a total of KL independent
is beyond the scope of this paper, we will not elaborate on it messages to deliver. The MIMO X channel is a more gener-
in the sequel. alized model, which encompasses the MIMO multiple access
Finally, when both the transmitters and the receivers are channel of Fig. 9, the MIMO broadcast channel of Fig. 10 and
the MIMO interference channel of Fig. 11 as its special cases.
Despite their difference, the MIMO interference channel and
Noise
X channel share a key common feature, namely they both have
distributed transmitters and receivers. The distributed nature of
multiple outputs
multiple inputs
Distributed

transmitters and receivers makes the signal processing required

Colocated

... Channel ... for mitigating the detrimental effects of the MIMO interference
channel and X channel far more challenging compared to
the single-user MIMO channel. In fact, the capacity analysis
and the signal processing techniques for MIMO interference
Fig. 9. MIMO multiple-access channel. channel and X channel still constitute a largely open field,

Noise
multiple outputs
multiple inputs
Colocated

Colocated

... Channel ...

Fig. 8. Point-to-point MIMO channel.

Noise {am |m = 1, · · · , M } and the a priori probability of selecting

each vector from ANI is identical. The set A is usually referred

multiple outputs
multiple inputs

[W [11], · · · , W [K1]]T [Ŵ [11], · · · , Ŵ [1L]]T

to as the constellation alphabet, whose elements can take either
Distributed

Distributed
[W [12], · · · , W [K2]]T [Ŵ [21], · · · , Ŵ [2L]]T
... Channel ... real or complex values. Additionally, sn , n = 1, · · · , M NI ,
[W [1L], · · · , W [KL]]T [Ŵ [K1], · · · , Ŵ [KL]]T represents the realizations of s, hence they are the elements of
ANI . Then the relationship between the inputs and the outputs
of this linear system can be characterized by5
Fig. 12. MIMO X channel, where W [kl] represents the message that
originates at transmitter l and is intended for receiver k, while Ŵ [kl] denotes y = Hs + n, (1)
the recovered version of W [kl] , k = 1, · · · , K, l = 1, · · · , L.
where y ∈ FNO is the received signal vector, H ∈ FNO ×NI
is the transfer function/channel matrix of the system, and n ∈
and most of existing efforts have aimed for transforming the FNO represents the additive random noise vector. Depending
MIMO interference channel and X channel so that coopera- on the specific applications considered, F can be either the
tion at the transmitter/receiver side can be exploited to some field of real numbers, R, or the field of complex numbers,
degree, at least in some specific scenarios. For example, in C. Concisely speaking, any system having multiple inputs,
multicell systems, BS cooperation [131], [194], [210]–[219], multiple outputs and subject to additive random noise can be
also known as joint multicell processing [24], [194], has been regarded as a MIMO system, but the MIMO detection problem
advocated for the sake of transforming the MIMO interference considered herein is only confronted in MIMO systems whose
channel and X channel to a number of cooperative multiuser channel matrix is non-orthogonal in columns. It is worth noting
MIMO channels. Additionally, the recent advances in the that the constellation alphabet A, the number of inputs NI and
capacity analysis of the MIMO interference channel and X the number of outputs NO are typically regarded as constant
channel have stimulated significant interests in interference quantities6 for a given system. Hence, they are assumed to
alignment [207], [208], [208], [209], [220]–[225], which is be known by default, although this will not be explicitly
essentially constituted by a family of precoding/beamforming emphasized, unless necessary. As a further note, when the
techniques for the MIMO interference channel and X channel. input symbol vectors of multiple consecutive time slots are
The problems related to interference alignment are also beyond associated with each other via space-time coding [266]–[274],
the scope of this paper and will not be discussed in detail. the MIMO system model is given by
Y = HC + N, (2)
IV. F ORMAL D EFINITION OF THE MIMO D ETECTION
P ROBLEM where Y is a matrix denoting the signal received in multiple
Despite the fact that similar problems have been known time slots, C is a matrix representing the space-time codeword,
for a while [46], [50], [51], [226]–[261], the term “MIMO and N is the corresponding noise matrix. We can obtain (1)
detection” became widespread mainly with the advent of from (2) by setting the number of time slots considered to
multiple-antenna techniques during the mid-1990s4 [152]– one. In this regard, (2) is more general than (1). However,
[156], [179]–[181], [262]–[273]. As a result, in the narrow (2) is mainly used for characterizing space-time coding aided
sense, MIMO detection usually refers to the symbol detection MIMO systems, where typically the MIMO detection problem
problem encountered in narrow-band SDM based multiple- defined in this paper does not exist. This is because the optimal
antenna systems, such as the vertical Bell Laboratories lay- ML decoding can be simply implemented using the separate
ered space-time (VBLAST) system [179]–[181]. However, we symbol-by-symbol decoding strategy [for orthogonal space-
emphasize that as a family of important signal processing time block codes (STBCs)] or the pairwise decoding strategy
techniques, MIMO detection should be interpreted based on a (for quasi-orthogonal STBCs) [274]. Therefore, in most cases
generic mathematical model, as detailed below. associated with MIMO detection, we rely on the system model
In the generic sense, the MIMO detection problem can be (1).
defined for an NI -input linear system whose transfer function Based on the generic mathematical model of (1), the basic
is described by a matrix having non-orthogonal columns and task of MIMO detection is to estimate the input vector s
its NO outputs are contaminated by additive random noise. relying on the knowledge of the received signal vector y
Note that the noise does not necessarily obey the Gaussian and the channel matrix H. Note that for y, typically its
distribution. The multiple inputs can be denoted as a vector s, exact value has to be known, while for H, sometimes only
which is randomly drawn from the set ANI composed by NI - the knowledge of its statistical parameters is available. To
element vectors, whose components are from a finite set A = elaborate a little further, if the instantaneous value of H is
4 Note that the first multi-antenna based MIMO system was attributed to
5 In the additive random noise contaminated ISI channels, the received signal
a patent granted to Paulraj and Kailath in 1994 [262]. Gerlach [153], Roy PMh
[152], [155] and Ottersten [152], [154] initiated the earliest research on SDMA is given by yn = i=0 hi sn−i + wn , n = 0, 1, · · · , L − 1, or by y =
systems. The earliest contribution demonstrating the huge capacity of multi- Hs + w, where y = [y0 , y1 , · · · , yL−1 ]T , w = [w0 , w1 , · · · , wL−1 ]T , L
antenna based MIMO systems may be attributed to Telatar [263], [264], as is the number of symbols observed, and Mh is the memory length of the ISI
well as Foschini [179], [265] and Gans [265], followed by other members of channel.
the team at Bell Labs [180], [181]. On the other hand, Tarokh, Jafarkhani, 6 However, in an adaptive system both the constellation alphabet A and
Calderbank, Naguib et al. [266]–[272] as well as Alamouti [273] are pioneers the number of inputs NI might be varying. But this adaptation is typically
of space-time code design. constrained by a discrete size-limited codebook.

h1,1 system [179]–[181] communicating over flat fading channels,

n1
h2,1 h1,2 as shown in Fig. 13. Because the system’s outputs at the
s1 y1
current time interval are independent of the system’s inputs
h2,2 n2 at previous time intervals, its baseband equivalent discrete-
Transmitter

time (i.e. sampled) system model, representing an instance of

Receiver
s2 y2
.. .. .. .. .. the generic model (1), can be written as
h1,Nt hNr ,1 
y1
 
h1,1 h1,2 · · · h1,Nt

s1

hNr ,2  y   h
h2,Nt n Nr  2   2,1 h2,2 · · · h2,Nt    s2 
 
=
 ..   .. .. .. ..   .. 
  
sN t hNr ,Nt yNr
 .   . . . .  . 
Fig. 13. Schematic of VBLAST-style SDM-MIMO systems communicating yNr hNr ,1 hNr ,2 ··· hNr ,Nt s Nt
over flat fading channels. (4)
n1
 
 n 
known from explicit channel estimation, the detection of s is  2 
said to rely on coherent detection. By contrast, if the explicit +
 ..
.

 .
estimation of the instantaneous channel state is avoided, the

detection of s belongs to the family of noncoherent detection nNr
schemes. In the latter case, the channel estimation is either In this specific application, we have Nt = NI and Nr = NO ,
performed implicitly in signal detection, or it is completely which represent the number of transmit and receive antennas,
avoided, whereas typically the statistical knowledge of the respectively. Furthermore, hj,i denotes the (complex-valued)
channel matrix H is invoked for supporting signal detec- impulse response between the ith transmit antenna and the jth
tion. Additionally, the noncoherent MIMO detection schemes receive antenna, with i = 1, 2, · · · , Nt and j = 1, 2, · · · , Nr .
usually require that the input symbols are subject to some Another example is the multiple-antenna aided orthogonal
form of differential encoding, which imposes correlation on frequency-division multiplexing (OFDM) system [283] com-
the input symbols, and as a result, typically a block-by-block municating over frequency-selective channels, where each
based sequence detection has to be employed. This is the subcarrier subjected to a specific frequency-domain attenuation
so-called multiple-symbol differential detection [274]–[283], is narrowband and the model (4) applies to each subcarrier.
which usually leads to higher computational complexity than Notably, for this linear memoryless MIMO channel, the one-
the symbol-by-symbol based detectors of coherent MIMO shot detection which relies only on a single received signal
systems. Moreover, the noncoherent detectors typically exhibit vector y = [y1 , y2 , · · · , yNr ]T is adequate. Additionally, for
degraded power efficiency, which results in an inherent perfor- the sake of fair comparison with the single-input single-output
mance loss compared to their coherent counterparts, unless the systems, typically the energy normalization of E(si ) = 1 or
block size is sufficiently large. Therefore, we have to consider E(s) = 1 is imposed on the transmitted symbols.
the performance-versus-complexity tradeoff in choosing the
proper block size. However, similar to the coherent detection VI. MIMO S YSTEM M ODEL FOR D ISPERSIVE C HANNELS
of STBCs, there exist simple symbol-by-symbol or pairwise E XHIBITING M EMORY
noncoherent detection schemes [276], [277] for differential On the other hand, when considering the stand-alone
space-time modulation. As a result, the decoding complexity wideband VBLAST system7 communicating over frequency-
of the differential space-time modulation increases linearly, selective MIMO channels [112], [285]–[288], the link between
instead of exponentially, with the number of antennas [274]. each input-output pair may be modelled by a linear finite
In this paper, we focus our attention on coherent MIMO de- impulse response (FIR) dispersive channel, whose sampled
tection. Then, from the perspective of mathematical mapping, version can be denoted as the (possibly complex-valued)
a coherent MIMO detector is defined as: vector hj,i = (h0j,i , h1j,i , · · · , hL−1 T
j,i ) . Here L is the maximum
ŝ = D(y, H) : FNO × FNO ×NI 7→ ANI , (3) number of multipath components in each link, and it is also
known as the channel memory length. In this case, the one-shot
where ŝ is the estimate of s. detection which utilizes a single Nr -element received signal
vector is not optimal. Instead, the sequence detection using
V. MIMO S YSTEM M ODEL FOR L INEAR M EMORYLESS multiple Nr -element received vectors has to be used.
C HANNELS We assume that a block-based transmission structure relying
Bearing in mind specific applications, the system model of on zero-padding for eliminating the interblock interference
(1) may be established either in the time domain or in the is used, which is beneficial for alleviating the performance
frequency domain, and may be applied to both memoryless degradation imposed by noise enhancement or error propa-
channels and dispersive channels exhibiting memory [21], gation [289]. Following zero-padding, a transmission block
[44], [50], [51], [158], [261]. With respect to linear mem- 7 Multicarrier techniques, such as OFDM, are not used in this system.
oryless MIMO channel, a canonical example is the narrow- However, when MIMO-OFDM systems [284] are considered, the MIMO
band single-carrier synchronous VBLAST-style SDM-MIMO detection is carried out on each subcarrier separately.

becomes a frame which occupies K = N + P sampling asynchronous CDMA systems can also be characterized by (1)
intervals, where N is the number of sampling intervals oc- in the z domain [50].
cupied by information-bearing symbol vectors in the frame,
while P ≥ L − 1 represents the number of sampling intervals VII. C OMPLEX -VALUED VERSUS R EAL -VALUED MIMO
during which P consecutive Nt -element zero vectors are S YSTEM M ODEL
inserted at the tail of the frame. Here we set P = L − 1. As we mentioned in Section IV, the generic MIMO system
Given the above-mentioned transmitted frame, the entire re- model of (1) can be defined both in the field of real num-
ceived signal vector may be generated by a concatenation bers, R, and in the field of complex numbers, C. Since the
of K noise-contaminated sampled received signal vectors, complex-valued modulation constellations, such as quadrature
namely, y = (yT [0], yT [1], · · · , yT [K − 1])T , where y[k] = amplitude modulation (QAM) and phase-shift keying (PSK),
(y1 [k], y2 [k], · · · , yNr [k])T represents the Nr outputs at the are often employed in digital communications, the complex-
kth sampling instant, k = 0, 1, · · · , K −1. Then, the baseband valued MIMO system model is typically a natural and more
signal received by the jth receive antenna at the kth sampling concise choice for the formulation and performance analysis
instant is given by of the algorithms considered.
Nt L−1
X X The complex-valued and the real-valued MIMO system
yj [k] = hlj,i si [k − l] + nj [k], (5) models are often mutually convertible. More specifically, if we
i=1 l=0 assume that the generic MIMO system model of (1) is defined
in C, and assume that the real part and the imaginary part of
where hlj,i , the lth element of hj,i , l = 0, 1, · · · , L − 1,
s are uncorrelated,8 then the complex-valued MIMO system
denotes the channel gain of the lth path between the ith
model of (1) can be transformed to an equivalent real-valued
transmit antenna and the jth receive antenna. Furthermore,
system model of
si [k] is the symbol transmitted from the ith transmit antenna
ỹ = H̃s̃ + ñ, (8)
at the kth sampling instant, and nj [k] represents the noise " # " # " #
imposed on the jth receive antenna at the kth sampling instant. ℜ(y) ℜ(s) ℜ(n)
Similar to y[k], we define s[k] = (s1 [k], s2 [k], · · · , sNt [k])T where ỹ = , s̃ = , ñ = , and
ℑ(y) ℑ(s) ℑ(n)
and n[k] = (n1 [k], n2 [k], · · · , nNr [k])T . Additionally, similar " #
ℜ(H) −ℑ(H)
to y, we may construct s = (sT [0], sT [1], · · · , sT [N − 1])T H̃ = .
and n = (nT [0], nT [1], · · · , nT [K − 1])T . Then, the received ℑ(H) ℜ(H)
signal corresponding to a transmitted frame can also be written However, the above real-valued decomposition is only appli-
following the matrix notation of (1), where the size of y and n cable to MIMO systems employing real-valued constellations
is NO = KNr = (N +L−1)Nr , while that of s is NI = N Nt , or rectangular QAM constellations (but not for PSK, star
and the MIMO channel matrix H exhibits the banded Toeplitz QAM [290]–[293] and near-Gaussian constellations [294],
structure [289] of: [295] etc.), which severely limits its applicability. Furthermore,
in many applications, complex-valued operations are more
H0
 
0 ··· 0 efficient for hardware implementation. The reason for this
 .. .. .. 
fact is twofold. Firstly, decomposing an (Nt × Nr )-element
. H0 . .
 
 
  MIMO system into a real-valued system requires storage of
.
.. .. the (2Nr ×2Nt )-element real-valued channel matrix H̃, which
 
 HL−1 . 0 
H= , (6) is twice larger than having 2Nt Nr real-valued elements that
 
 L−1 . . 0

 0 H . H  would be needed for a standard representation of the original
 
 . . . ..  complex-valued channel matrix H. Secondly, implementing
 .. .. .. .

  complex-valued arithmetics in hardware (e.g. very-large-scale
L−1 integration (VLSI) based circuits) is straightforward and does
0 0 ··· H
not result in more complex hardware. For example, a complex-
whose dimension is (KNr × N Nt ), and each entry Hl of (6) valued multiplier can be built using 4 real multipliers and
is an (Nr × Nt )-element matrix containing the channel gains 2 real adders, because we have (a + jb)(c + jd) = ac −
between all pairs of transmit and receive antennas for the lth bd + j(ad + bc), or using 3 real multiplications and 5 real
path, i.e. we have additions, because alternatively we have (a + jb)(c + jd) =
 l
h1,1 hl1,2 · · · hl1,Nt
 ac − bd + j[(a + b)(c + d) − ac − bd], which is known
 hl hl2,2 · · · hl2,Nt  as the Gaussian technique of multiplying complex numbers
 2,1 [296]. As a result, the complex-valued model imposes a
Hl = 

.. .. .. . (7)

 . . ··· .

 lower silicon complexity than that required by the real-valued
decomposition based model. Therefore, in many cases the real-
hlNr ,1 hlNr ,2 ··· hlNr ,Nt
valued decomposition can be detrimental and hence MIMO
It is worth noting that the linear memoryless MIMO model 8 For rectangular QAM, this uncorrelatedness assumption is almost always

and the linear MIMO model exhibiting memory may also adopted for ease of decoding, even in single-input single-output systems. In
channel-coded systems, where the channel codes may introduce correlation
be used for characterizing the family of synchronous and between the coded bits. However, in such systems, typically an interleaver is
asynchronous CDMA systems, respectively. Additionally, the invoked after the encoder, which mitigates the correlation.

detector designers are typically in favor of the complex-valued

system model, owing to its flexibility concerning the choice Symbol detection in large−scale
late 2000s −− present
of constellations and its efficiency in VLSI implementations. multi−antenna systems

On the other hand, the real-valued MIMO system model

may also enjoy some advantages, such as the increased free-
dom of manipulation in signal processing. To elaborate a little Symbol detection in small−/medium−scale
mid−1990s −− early 2000s
multi−antenna systems
further, as far as the achievable performance is concerned, in
most cases signal processing algorithms based on the complex-
valued model of (1) and the real-valued MIMO system model
of (8) deliver an equivalent performance. For example, [264] Multiuser detection in CDMA systems 1980 −− 1990s
showed the equivalence between the complex-valued and the
real-valued MIMO system models in the derivation of the op-
timal ML detector and the MIMO channel capacity. However,
this equivalence does not always hold. For example, it was
Combating crosstalk in FDM/TDM systems 1960 −− 1970s
shown in [297] that the real-valued VBLAST detector outper-
forms its complex-valued counterpart, owing to its additional
freedom in selecting the optimum detection ordering. Hence, a
beneficial performance gain may be gleaned from transforming Fig. 14. The four historical periods in the development of MIMO detection.
the complex-valued system model to the double-dimensional
real-valued system model. This is also true for the tree-search
based MIMO detectors, which will be introduced in Section a circularly (symmetric) complex-valued random variable has
VIII-D, when they invoke symbol detection ordering. More a probability distribution that is invariant under rotation in
generally, the key insight inferred here is that for all MIMO the complex plane, namely the distribution of x must be the
detection algorithms whose performance is related to detection same as the distribution of ejθ x, where we have θ ∈ [0, 2π).
ordering, the real-valued system model based formulation is The conventional signal processing techniques often assume,
capable of providing a better performance than its complex- usually implicitly, that the complex-valued random signals are
valued counterpart. This gain is achieved at the expense of proper or circularly symmetric9 . However, these assumptions
extra redundancy in storage of the channel matrix H̃, and may not be justified in some applications, hence the complex-
if the symbol detection ordering technique is invoked, this valued signal processing techniques may in certain circum-
redundancy cannot be avoided. stances achieve a better performance [120], [130]–[133]. For
more details on the complex-valued signal processing, please
Additionally, it is worth noting that the real-valued formula-
refer to [301]–[303].
tion of the complex-valued MIMO system model is not unique.
For example, a pairwise real-valued MIMO system model was
used in [298], [299], which was shown to result in a reduced VIII. H ISTORY AND S TATE - OF - THE -A RT OF MIMO
complexity compared to the conventional real-valued MIMO DETECTION
system model of (8). A more comprehensive investigation of The research of MIMO detection is a broad and vibrant area.
the complex-valued versus real-valued MIMO detectors was Its embryonic concept dates back to the 1960s. The earliest
presented in [300]. contribution on MIMO detection was sparked off in 1967
Finally, we emphasize that if the complex-valued random [226], when Shnidman considered the equalization problem
signals considered are improper or noncircular, the more ad- of a bandwidth-limited pulse modulation system. This system
vanced complex-valued signal processing techniques of [301]– was modelled with the aid of M waveforms, each of which
[303] that rely on additional statistics and tools for fully is amplitude-scaled and simultaneously transmitted over a
characterizing the complex-valued random signals have to be single physical channel, which has M outputs corresponding
used. More specifically, for a complex-valued random vector to each signal waveform. In order to eliminate both the ISI
x, in addition to the conventional covariance matrix of between the pulse train and the interference between different
waveforms (also known as crosstalk), Shnidman formulated
Cxx = E[(x − µx )(x − µx )H ], (9) a generalized Nyquist criterion and proposed an optimum
another second-order statistic, namely the pseudo-covariance linear receiver. This landmark contribution was essentially
matrix defined as inspired by the classic Nyquist’s problem [304], which aims
for the joint optimization of the transmitter and receiver
C̄xx = E[(x − µx )(x − µx )T ], (10) for the sake of combating the ISI when communicating
over a conventional single-input single-output channel. Since
has to be introduced for fully describing the complex-valued
then, the MIMO detection problem has been studied in the
random vector. For a proper complex-valued random vector,
context of diverse applications and under possibly different
the pseudo-covariance matrix vanishes, which is formulated
names. This half-century history can be roughly divided
as C̄xx = 0. This results in the fact that for a proper
complex-valued random scalar, the real and imaginary parts 9 For complex-valued Gaussian random vector Z, circular symmetry implies
must have the same variance and be uncorrelated. Additionally, that Z is zero mean and proper.

into four periods, as seen in Fig. 14, namely the period of 1) Matched Filter (MF) versus Optimum MIMO Detector:
combating crosstalk in the context of the early single-user Although it is widely recognized at the time of writing
FDM/TDM systems (1960s – 1970s) [226]–[229], [232], the that MIMO detection provides significant performance gains
period of multiuser detection (MUD) during the prevalence of compared to conventional single-stream detection, there was a
CDM/CDMA systems (1980s – 1990s) [46], [50], [51], [230], widespread misconception until the early 1980s that the MUI
[231], [233]–[261], the period of joint symbol detection in the can be accurately modelled as a white Gaussian random pro-
small-/medium-scale multiple-antenna systems (mid-1990s cess, and thus the conventional single-user MF (SUMF) based
– mid-2000s) [53], [54], [57]–[64], [68]–[72], [80], [82], detector, as illustrated in Fig. 16, was believed to be essentially
[104]–[126], [128], [129], [134]–[146], [179]–[181], [261], optimum. In 1983, this conventional wisdom was explicitly
[305]–[324], and the period of symbol detection in the proven wrong by Verdú [235], [236] with the introduction of
large-scale multiple-antenna systems [118], [128], [176], the optimal MUD in the context of asynchronous/synchronous
[325]–[341]. Diverse MIMO detectors have been proposed Gaussian multiple-access channels shared by K users. The full
for meeting the requirements imposed by a multiplicity analysis and derivation of the optimum MUD was reported
of applications. These MIMO detectors can be categorised later in [237], [354], demonstrating that there is, in general, a
from various perspectives, such as optimum/suboptimum, substantial gap between the performance of the conventional
linear/nonlinear, sequential/one-shot, adaptive/non-adaptive, SUMF and the optimal MUD performance. Additionally, upon
hard-decision/soft-decision, blind/non-blind, iterative/non- identifying the non-Gaussian nature of the MUI, Poor and
iterative, synchronous/asynchronous, coded/uncoded etc. Note Verdú [355] also designed nonlinear single-user detectors for
that a detailed discourse on the application of soft-decision CDMA systems operating in diverse scenarios such as weak
MIMO detectors in near-capacity turbo/iterative receivers interferers, high spreading gains and high signal-to-noise ratio
was provided in [342], while coherent and noncoherent (SNR). The performance of the ML based optimum MIMO
MIMO detectors in the context of the emerging “space-time detector has been analyzed in [237], [354], [356]–[360].
shift keying (STSK)” based multicarrier MIMO systems There does exist some situations where a bona fide applica-
was presented in [177]. The representative MIMO detectors tion of the central limit theorem is feasible and hence the MUI
considered in this paper are summarized in Fig. 15. can be rigorously proven to be asymptotically Gaussian.11
Owing to the similarities between the classic equalization However, even if the MUI may be accurately modelled as a
problem encountered in channels imposing ISI and the generic Gaussian variable, the SUMF is still not the optimal receiver.
MIMO detection problem defined by (1) and (3), it is not This is because the output of the MF for the desired user
surprising that the techniques, which were found to be effective does not constitute a sufficient statistic in the presence of MUI
in combating ISI were also often extended to the context of [361]. In other words, the SUMF is optimal only in the context
MIMO detection problems [343]. Some of the equalization of the single-user channel contaminated by additive white
algorithms which have been adapted for MIMO detection in- Gaussian noise (AWGN). By contrast, in multiple-access sys-
clude, but not limited to, the ML sequence estimation (Viterbi tems, unless the multiplexed signals (after passing through the
algorithm) [344]–[348] based equalization, linear ZF equaliza- channel) are orthogonal, the outputs of the MFs corresponding
tion [161], linear MMSE equalization [161], ZF/MMSE aided to the interfering users contain valuable information which
decision-feedback equalization [161], adaptive equalization may be exploited for the detection of the symbol of interest,
[349], [350], blind equalization [351], [352] etc, as detailed and hence more intelligent joint detection strategies capable of
below. exploiting all MFs’ outputs are required for achieving better
detection performance.
2) Optimum Decision Criteria: When designing optimal
A. Optimum MIMO Detector detectors/receivers for communication systems, it is usually
necessary to clarify in what sense the word “optimum” is
The earliest work on optimum MIMO detectors dates back referred to. This is because the specific choice of an optimal
to 1976, when van Etten [229] derived an ML sequence esti- detector/receiver is strongly dependent on the specific assump-
mation based receiver for combating both ISI and interchannel tions and criteria of “goodness”. An optimal detector/receiver
interference (ICI) in multiple-channel transmission systems. is the one that best satisfies the given criterion of goodness
Explicitly, he demonstrated that under certain conditions, the under a given set of assumptions. If either the criterion or
performance of the ML receiver asymptotically approaches the assumptions change, typically the choice of the optimal
that of the optimum receiver of the idealized system which is detector/receiver also changes. If the assumptions used in the
free from both ISI and ICI. The significance of this work was theoretical analysis are inconsistent with the conditions of
not fully recognized until the research interests in commercial the realistic environment considered, then it is possible that
CDMA systems and multiple-antenna systems intensified.10 the theoretically optimal detector/receiver obtained fails to
provide valid insights and results for the practically achievable
10 In fact, van Etten’s pioneering companion papers on the optimum MIMO performance and designs. Special attention has to be paid to
detector [229] and on the optimum linear MIMO detector [228] were included
in the book The best of the best: Fifty years of communications and networking 11 A specific example of such a situation is that an infinite-population
research [353], which was compiled by the IEEE Communications Society in multiuser signal model with the individual amplitudes going to zero at the
2007, and he is the only researcher who has two sole-author papers included appropriate speed – in other words, when the overall interference power is
in this selection. fixed and the number of equal-power interferers tends to infinity [361].

ML/MAP optimum

MIMO Detectors

suboptimum

Linear Nonlinear

Metaheuristic
MF ZF MBER MMSE IC Tree Search PDA SDPR LR
Algorithms

SIC/DFD PIC/MIC ACO PSO GA SA ···

Depth−First Breadth−First Best−First
(SD) (QRD−M) (A∗)

Fig. 15. Overview of representative MIMO detectors considered in this paper. All the acronyms used have been defined in our Glossary. Note that the tree-
search based detector is in general suboptimum, but it has the flexibility to strike different tradeoffs between the achievable performance and the computational
complexity. Even the optimum ML performance may be achieved by tree-search based detectors in certain scenarios.

the definition of “optimum”, since the theoretically optimal Using Bayes’ rule, the a posteriori probability (APP) in (11)
results obtained are mainly invoked as a benchmark or bound, may be expressed as
against which any other results can be compared. There p(y|s) Pr(s) p(y|s) Pr(s)
are many criteria of goodness. As far as the performance Pr(s|y) = = P , (12)
p(y) p(y|s′ ) Pr(s′ )
of the detectors/receivers used in communication systems s′ ∈ANI
is concerned, the minimum error probability criterion is of
primary interest, and hypothesis testing as well as likelihood where Pr(s) is the a priori probability of s, and p(y|s) is the
ratios are of great importance. In Bayesian inference, the conditional probability density function (PDF) of the observed
optimum decision criterion which minimizes the error proba- signal vector y given s. The MAP criterion can be simplified
bility based only on the observed signals and a given set of when each vector in ANI has an identical a priori probability,
hypotheses is the maximum a posteriori (MAP) criterion. The i.e. we have Pr(s) = 1/|A|NI for all realizations of s, where
error probability in communication systems can be measured |A| represents the number of elements, i.e. the cardinality of
in multiple scales, such as bit-error rate (BER), symbol- the constellation alphabet A. Furthermore, considering the fact
error rate (SER), and packet-/frame-/block-/vector-error rate that p(y) is independent of which particular signal vector is
(PER/FER/BLER/VER). When considering the MIMO system transmitted, then the MAP detector of (11) becomes equivalent
model of (1), the MAP criterion based MIMO detector which to the ML detector of
is optimal in the sense of minimum VER is formulated as DML : ŝ = arg max p(y|s). (13)
s∈ANI
DMAP : ŝ = arg max Pr(s|y). (11)
s∈ANI Therefore, the MAP criterion is usually used in the iterative
detection and decoding (IDD) aided receiver of forward-error-

Matched Filter .. b̂1

User 1

y Matched Filter .. b̂2

User 2

Matched Filter .. b̂K

User K

Fig. 16. An implementation example of the MF detector for synchronous

CDMA systems employing BPSK modulation.

Hs1
correction (FEC)-coded systems, where the a priori proba-
bilities of the transmitted symbols, Pr(s), may be obtained Hs2
with the aid of a backward-and-forward oriented iterative
information exchange between the signal detector and the
y
channel decoder. By contrast, the ML criterion is usually
used in FEC-uncoded systems, where the a priori probabilities
of the transmitted symbols cannot be made available by the
channel decoder. If n is AWGN, then we have
Hs3
2
p(y|s) ∝ exp(− ky − Hsk2 ), (14)
where the symbol ∝ represents the relationship “is propor-
tional to”. Consequently, we have
Hs4
2
max p(y|s) ⇔ min ky − Hsk2 , (15)
s∈ANI s∈ANI Fig. 17. Example of the optimal ML based MIMO detector in the context
where the symbol ⇔ represents the relationship “is equivalent of NI = 2 and BPSK modulation.
to”. Therefore, the ML detection problem for the system model
of (1) can be reformulated as the finite-set constrained least-
squares (LS) optimization problem of More explicitly, the optimum MUD proposed in [237] for
asynchronous CDMA systems consists of a bank of MFs
2
ŝML = arg min ky − Hsk2 , (16) followed by a dynamic programming algorithm of the forward
s∈ANI
(Viterbi) type [344]–[348] (for ML criterion based detection)
which can also be interpreted as the minimum Euclidean or of the backward-forward type [364]–[368] (for minimum
distance (MED) criterion. error probability criterion based detection). As mentioned in
Note, however, that the above-mentioned MAP, ML and Section VI, asynchronous CDMA systems can be modelled
MED criterion based MIMO detectors all aim for minimizing relying on the MIMO system model given in Section VI for
the VER, but do not guarantee achieving the minimum BER transmission over linear dispersive channels exhibiting mem-
and minimum SER, which are two metrics of particular im- ory. Therefore, the optimum MUD conceived for asynchronous
portance in many applications, such as in FEC-coded systems. CDMA constitutes a sequence detector, while the optimum
There are other frequently used criteria in MIMO detector MUD of synchronous CDMA is a one-shot detector, and
design. The linear MF criterion is optimal for maximizing the as such it is a special case of the asynchronous optimum
received SNR in the presence of additive stochastic noise. MUD. The optimum MUD relying on brute-force search [237]
The linear ZF criterion is optimal for maximizing the re- requires that the transmitted energies of each user were known
ceived signal-to-interference ratio (SIR). By contrast, the linear to the receiver. More critically, the computational complexity
MMSE criterion based detector is optimal amongst all linear of the optimum decision algorithms suggested in [237], [367],
detectors12 in terms of achieving the MMSE, and in essence [368] increases exponentially with the number of active users,
it is also optimal for maximizing the received signal-to- i.e. it is on the order of O(2K ) per bit for asynchronous trans-
interference-plus-noise ratio (SINR) amongst linear detectors mission and O(2K /K) per bit for synchronous transmission,
[362], [363]. Additionally, the linear minimum bit-error rate where K is the number of active users. This is because the
(MBER) criterion based detector achieves the lowest BER optimum MUD of both the synchronous and asynchronous
amongst all linear detectors, as detailed in Section VIII-B4. CDMA scenarios was proven by Verdú13 to be an NP-hard and
3) Computational Complexity: The optimization problem a non-deterministic polynomial-time complete (NP-complete)
of (16) can be solved by “brute-force” search over ANI , result- problem [46], [354]. Thus, all known algorithms designed
ing in an exponentially increasing computational complexity for solving this problem optimally exhibit an exponentially
of |A|NI . To elaborate a little further, let us consider the increasing computational complexity in the number of decision
example shown in Fig. 17, where binary phase-shift keying variables. Therefore, the optimum MUD becomes computa-
(BPSK) modulation (M = 2) and NI = 2 are employed. tionally intractable for a large number of active users.
Hence, there are a total of M NI = 4 possible realizations It should be noted that the optimal MIMO detection problem
for the transmitted symbol vector s, and they are denoted as would only have a polynomially increasing complexity if
s1 = [1, 1]T , s2 = [1, −1]T , s3 = [−1, −1]T , s4 = [−1, 1]T . and only if a polynomial-time solution could be found for
To gain deeper understanding of the computational com- any NP-complete problem, such as the famous travelling
plexity of the optimum MIMO detector formulated in (13), let salesman problem and the integer linear programming problem
us examine its implementations in practical CDMA systems. which have been so far widely believed insolvable within
12 In general, the MMSE detector and the linear MMSE detector are not polynomial time. However, the question of whether there exists
necessarily the same. The former only aims at minimizing mean-square error
(MSE) and does not impose any constraint on the form of the MMSE 13 In fact, the optimum MIMO detection problem of (16) constitutes an
estimator. The latter assumes that the MMSE estimator is a linear function instance of the general closest lattice-point search (CLPS) problem, whose
of the observed signal vector y. If y and the transmitted signal vector s are complexity had been analyzed earlier by Boas [45] in 1981, showing that this
jointly Gaussian, then the MMSE estimator is linear. In this case, for finding problem is NP-hard. Additionally, Micciancio [47] provided a simpler proof
the MMSE estimator, it is sufficient to find the linear MMSE estimator. for the hardness of the CLPS problem in 2001.

filter matrix

T?
NP−hard NP−hard
y ŝ
demodulator

Complexity
NP−complete

Fig. 19. Conceptual illustration of linear MIMO detectors.

NP P = NP = NP-complete

include the linear detectors, the interference cancellation aided

P
detectors, the tree-search based detectors, the PDA based de-
tectors, the SDPR based detectors and the LR based detectors
if P 6= NP if P = NP etc., as seen in Fig. 15 and detailed below.

B. Linear MIMO Detectors

Fig. 18. Euler diagram for P, NP, NP-complete, NP-hard set of problems
under both P6=NP and P=NP assumptions [369], [370]. The linear MIMO detectors of Fig. 15 are based on a linear
transformation of the output signal vector y. In general, they
are known for their appealingly low complexity, but suffer
a polynomial-time solution for NP-complete problems has from a considerable performance loss in comparison to the
not been answered by a rigorous proof to date. It is widely ML detector. More explicitly, the decision statistics of linear
recognized that in computational complexity theory, the com- MIMO detectors may be expressed as
plexity class of “P” represents one of the most fundamental
d = Ty, (17)
complexity classes, and it contains all decision problems that
can be solved by a deterministic Turing machine using a where T is the linear transformation (or filtering) matrix to
polynomially increasing amount of computation time (this is be designed using various criteria. A conceptual illustration
conventionally abbreviated to the parlance of “polynomial- of the linear MIMO detectors is given in Fig. 19.
time” for convenience). In fact, the most important open 1) MF Detector: For the sake of illuminating the phi-
question in computational complexity theory [369], [370] has losophy of linear MIMO detectors, let us rely on (4) and
been the formal proof of “Is P = NP?”, which explicitly poses continue by considering the MF detector, which has the lowest
the dilemma whether polynomial-time algorithms actually do computational complexity among all MIMO detectors and its
exist for NP-complete problems, and by corollary, for all NP linear transformation matrix is given by
problems. Fig. 18 concisely depicts the Euler diagram char-
acterizing the relationships amongst the P, NP, NP-complete, TMF = HH . (18)
and NP-hard set of problems under both the P6=NP and P=NP Upon using the MF detector of (18), we obtain
assumptions.
Additionally, it is worth mentioning that for some algo- d = HH Hs + HH n. (19)
rithms, such as the tree-search based MIMO detectors to be The MF detector is well known as the optimal linear filter
detailed in Section VIII-D, the computational complexity may designed for maximizing the output SNR in the presence of ad-
vary in different scenarios. As such, the average computational ditive stochastic noise. In Section VIII-A1, we have provided
complexity, the worst-case computational complexity and the some discussions regarding the MF detector in order to justify
distribution of computational complexity become important the motivations of developing joint detection based MIMO
metrics to examine. Finally, in practical algorithm implementa- detectors. To elaborate a little further, the MF detector had
tions, it is also important to consider the hardware complexity, been widely used before the concept of MIMO detection was
which, in simplest form, can be measured by the silicon born, and it is essentially based on the single-user detection
area and the number of NAND2 gates required in the IC philosophy. Hence, strictly speaking, it does not belong to the
implementation. joint detection based MIMO detection family, and typically it
4) Milestone Contributions: For the sake of clarity, the exhibits a poor performance in CCI-limited MIMO systems.
main contributions to the development of the optimum MIMO However, in certain LS-MIMO contexts [25], [373], the MF
detector are summarized in Table I. detector is capable of approaching the performance of the
The substantial performance and complexity differences optimal ML detector, as it will be further discussed in Section
between the optimum MIMO detector and the conventional IX.
SUMF detector stimulated a lot of interests in the development 2) Linear ZF Detector: Assuming that the noise vector is
of suboptimum MIMO detection algorithms that are capable zero, (4) becomes a system of linear equations, and the MIMO
of achieving good performance at a low computational cost. detection problem becomes equivalent to “finding the solution
Some representative classes of suboptimum MIMO detectors for Nt unknown variables subject to Nr linear equations”.

TABLE I
M ILESTONES IN THE DEVELOPMENT OF THE OPTIMAL MIMO DETECTOR

Year Authors Contributions

1976 van Etten [229] Derived an ML sequence estimation based receiver for combating both the ISI and ICI in
multiple-channel transmission systems and demonstrated that under certain conditions, the
performance of the ML receiver is asymptotically as good as if both the ISI and ICI were
absent.
1981 Boas [45] Analyzed the complexity of the generic problem of “closest point search in an NI -dimensional
lattice”, which is identical to the optimum MIMO detection problem, as a function of the
dimension NI of the decision-variable vector, and proved that this problem is NP-hard. Thus, all
known algorithms conceived for solving the generic MIMO detection problem optimally have
an exponentially increasing computational complexity.
1983 - 1986 Verdú [235]–[237], First presented a full derivation and analysis of the ML based multiuser detector for asyn-
[354] chronous/synchronous CDMA systems; showed that there is, in general, a huge gap between the
performance of the conventional SUMF and the optimal attainable performance; showed that
the infamous near-far problem was not an inherent flaw of CDMA but a consequence of the
inability of the SUMF to exploit the structure of the MUI; introduced the performance measure
of multiuser asymptotic efficiency, which was later widely used in the asymptotic analysis of
multiuser detectors at the high-SNR region.
1984 - 1989 Verdú [46], [354] Independently proved that the optimum MUD problem in CDMA systems is NP-hard and NP-
complete.
2001 Micciancio [47] Presented a simpler proof of the NP-hardness of the problem of closest point search in an
NI -dimensional lattice.
2003 Garrett et al. [371] Proposed the first VLSI implementation of a soft-output ML detector having a 19.2 Mbps
uncoded data rate supporting up to 4 × 4 QPSK MIMO.
2003 Burg et al. [372] Presented an efficient VLSI implementation of hard-decision optimum ML detector for QPSK
MIMO. The proposed method does not compromise optimality of the ML detector. Instead it
uses the special properties of QPSK modulation, together with algebraic transformations and
architectural optimizations, to achieve low hardware complexity and high speed up to 50 Mbps.

Therefore, if H is a square matrix (i.e. Nr = Nt ) and of full that if the transmitted energies of each user are unknown to
rank, the solution of this system of linear equations is given the receiver, then both the ML amplitude estimates and the
by s = H−1 y. To generalize this problem a little further, if ML decisions on the transmitted bits are obtained by the ZF
the matrix H satisfies Nr > Nt and has a full column rank of detector, regardless of the values of the received energies of
Nt , we have s = H† y, where H† = (HH H)−1 HH is the left- each user. As a beneficial result, the ZF detector achieves the
multiplying Moore-Penrose pseudoinverse of H. This example same degree of resistance to the infamous near-far problem
actually conveys the essential idea of the ZF criterion based as the optimum ML detector, despite its significantly reduced
MIMO detector, for which the linear transformation matrix is computational complexity. The insight that the near-far prob-
given by lem was not an inherent flaw of CDMA but a consequence
TZF = H† , (20) of the SUMF’s inability to exploit the non-Gaussian structure
of the MUI [237], and the fact that the joint detection based
and if H is invertible, the left-multiplying pseudoinverse H†
MUDs, including its linear versions, achieve a significantly
and the inverse coincides, i.e. we have H† = H−1 . Upon
better near-far resistance [239], [240], [374], [375] became
using the ZF detector, we have d = s + H† n, which indicates
another major incentive for the subsequent research activities
that the interference amongst the multiple inputs is completely
dedicated to MUD in CDMA. Additionally, with the advent
eliminated, albeit the noise power is augmented.
of the multiple-antenna technologies conceived during the
Similar to the case of the optimum ML-based MIMO mid-1990s, the ZF detector was first studied in the SDM-
detector, the ZF criterion based linear MIMO detector of Fig. based VBLAST systems by Foschini, Wolniansky, Golden and
15 was also first proposed by van Etten [228] in 1975 for a Valenzuela [179]–[181].
multiple-channel multiplexing transmission system subjected
to both ISI and ICI. As far as CDMA systems are concerned, 3) Linear MMSE Detector: As seen in Fig. 15, the linear
this solution was first proposed by Schneider [231] in 1979 transformation matrix T of (17) can also be designed accord-
for synchronous CDMA systems transmitting equal-energy ing to the MMSE criterion, which minimizes the mean-square
multiuser signals, where he sought to minimize the probability error between the actual transmitted data and the channel’s
of bit error, but erroneously arrived at the ZF detector. From output data after using the linear transformation matrix T. To
1986 to 1990, Lupas and Verdú systematically investigated this be more specific, T is obtained by solving the optimization
detector in the context of both synchronous [239], [374] and problem of
asynchronous [240], [375] CDMA systems. They referred to
2

it as the linear decorrelating multiuser detector. It was shown TMMSE = arg min E ks − Tyk2 . (21)
T

Using the orthogonality principle [376], we have by a single-user system to achieve the same asymptotic
H error probability) and the actual SNR of the desired user,
E[(s − Ty)y ] = 0, (22)
when the noise power tends to zero. Furthermore, the
then TMMSE may be derived as linear MAME detector was designed by exploiting the
assumption that the individual transmitted energies of all
TMMSE = (HH H + 2σ 2 I)−1 HH , (23) the users are fixed and known to the receiver. By contrast,
where σ 2 is the noise power per real dimension, and E(s) = 1 the ZF detector does not require the knowledge of the
is assumed. Compared to the linear ZF detector, the linear transmitted energies of the users.
MMSE detector achieves a better balance between the MUI • Additionally, since a common disadvantage of the lin-

elimination and noise enhancement by jointly minimizing the ear ZF and MMSE detectors is that their estimates of
total error imposed by both the MUI and the noise. Hence, the the transmitted symbols are biased, Xie, Rushforth and
linear MMSE detector achieves a better performance at low Short [253], [377] proposed the so-called weighted least-
SNRs than the ZF detector. squares (WLS) linear detector, which is capable of pro-
The MMSE criterion based linear MIMO detector was viding an unbiased estimate of the transmitted symbols.
first proposed by Shnidman [226] in 1967, and hence it It is worth pointing out that except for the linear MF and
is the oldest MIMO detector found in the literature. The ZF detectors, other linear MIMO detectors – including
generalized Nyquist criterion formulated by Shnidman first the linear MMSE detector, the linear MAME detector and
indicates that the ISI and crosstalk14 between multiplexed the linear WLS detector – were typically derived under
signals essentially represent identical phenomena. Then, re- the assumption that the system parameters such as the
lying on this insight, he proposed a linear receiver that is signal’s phase, power and delay are known. As a result,
optimal in the sense of the MMSE criterion for combating both in practice these parameters must be estimated and the
the ISI and crosstalk in single-channel multiple-waveform- receiver’s structure has to be regularly modified to reflect
multiplexed pulse-amplitude modulation (PAM) systems. In the updated estimates.
1970, Kaye and George [227] explicitly extended the MMSE • Another important class of linear MIMO detectors are

receiver of [226] to the family of general multiple-channel based on the MBER criterion. The linear MBER detector
systems transmitting multiplexed PAM signals and/or pro- is capable of outperforming the linear MMSE detector
viding diversity. The MMSE criterion based linear detector when either the signature cross-correlation is high or
for CDMA systems was proposed by Xie, Rushforth and the background noise is non-Gaussian [392]. Again, the
Short in 1989 [253], [377]. A decade later, it was also MBER based MIMO detector was first considered by van
revisited by Foschini, Wolniansky, Golden and Valenzuela in Etten [228] in 1975 in the context of a multiple-channel
the context of SDM-based multi-antenna systems [179]–[181]. multiplexing transmission system subjected to both ISI
The performance of linear ZF/MMSE based MIMO detectors and ICI. This MBER criterion was later studied in the
depends on the SINR experienced at the output of these context of CDMA systems [386]–[389], [392]–[397] and
detectors, which was first analyzed by Poor and Verdú [378] in multi-antenna systems [308], [309].
1997, and investigated in more depth later from various other • Finally, we would like to mention that the linear MIMO

perspectives, such as the error probability, outage probability, detector can also be designed from the perspective of
diversity-multiplexing tradeoff (DMT) [379], [380], as well a linear equalizer [253], [377], since the mathematical
as asymptotic distribution of the SINR (in terms of antenna models of the MIMO detection problem and of the equal-
number and high/low SNR regimes) [362], [363], [381]–[385]. ization problem are similar [343]. To elaborate a little
further, in MIMO systems each symbol’s interference is
4) Other Linear Detectors: As observed in the family-tree imposed by other simultaneous transmissions, while in
of Fig. 15, there are a range of other criteria for designing the the band-limited ISI channels requiring equalization, the
linear transformation matrix T. interference of a particular symbol is due to other symbols
that are transmitted sequentially in the time domain.
• For example, in [239], [374], Lupas and Verdú also
proposed a maximum asymptotic-multiuser-efficiency The main contributions to the development of linear MIMO
(MAME) based linear detector, which is capable of detectors are summarized in Table II.
minimizing the probability of bit errors in the limit as
the noise approaches zero. The asymptotic-multiuser- C. Interference Cancellation Aided MIMO Detectors
efficiency (AME) is a metric which characterizes the Another important class of suboptimum MIMO detectors
performance of the MUD in the high-SNR region. It portrayed in Fig. 15 are constituted by the interference cancel-
implies the performance loss of the desired user in the lation based MIMO detectors, which are nonlinear and gener-
high-SNR region due to the interference imposed by other ally achieve a better performance than linear MIMO detectors.
active users. To be more specific, it is defined as the limit The concept of interference cancellation was first studied in
of the ratio between the effective SNR (that is required 1974 by Bergmans and Cover [398], [399], as well as by
14 Crosstalk may be interpreted as a special case of ICI. For example, as
Carleial [197] in 1975, in their information-theoretic studies
mentioned before, in [226], crosstalk means the interference between the of broadcast channels and of interference channels, respec-
multiplexed different waveforms. tively. In the context of CDMA and multi-antenna systems,

TABLE II
M ILESTONES IN THE DEVELOPMENT OF LINEAR MIMO DETECTORS

Year Authors Contributions

1967 Shnidman [226] First formulated a generalized Nyquist criterion, which pointed out that the ISI and crosstalk
between multiplexed signals are essentially identical phenomena; he then proposed a linear
MMSE receiver for combating both ISI and crosstalk in single-channel multiple-waveform-
multiplexed PAM systems.
1970 Kaye et al. [227] Extended the MMSE receiver of [226] to the general multiple-channel systems transmitting
multiplexed PAM signals and/or providing diversity.

1975 van Etten [228] Developed linear receivers based on both the ZF criterion and the minimum error probability
criterion for a multiple-channel transmission system similar to that of [227]; these two
detectors heralded the linear ZF and the linear MBER multiuser detectors of CDMA systems.
1975 Horwood et al. [230] Proposed two linear signal-correlation based detectors for synchronous digital multiple-access
systems; one of them assumes that each user only knows its own signature, while the other
assumes that each user knows all users’ signatures; this is the first attempt in multiple-access
systems to exploit the structure of the signals simultaneously sent, which is the key idea of
MUD in CDMA systems.
1979 Schneider [231] First made an attempt to conceive MUD for CDMA systems; he proposed the linear
decorrelating detector, namely the linear ZF detector, which represents one of the mainstream
MUD approaches conceived for CDMA systems; this detector was also extended to the
scenario of combating crosstalk in M -ary multiplexed transmission systems in 1980 [232].
1986-1990 Lupas et al. [239], Systematically investigated the linear ZF MUD in the context of both synchronous [239], [374]
[240], [374], [375] and asynchronous [240], [375] CDMA systems; they showed that the ZF detector achieves
exactly the same degree of resistance to the infamous near-far problem as the optimum ML
detector, despite its much lower computational and implementation complexity; they also first
proposed a linear MAME MUD, which is capable of equivalently minimizing the probability
of bit error in the limit as the noise approaches zero.
1989 - 1990 Xie et al. [253], [377] First proposed the MMSE criterion based linear MUD, the modified linear equalizer based
MUD, and the WLS linear MUD for CDMA systems. In contrast to the linear ZF detector,
to the linear MMSE detector, and to the modified linear equalizer based detector, the linear
WLS detector is capable of providing an unbiased estimate of the transmitted symbols.
1993 - 1997 Mandayam et al. First proposed the MBER criterion based linear MIMO detectors for CDMA systems; the
[386]–[389] linear MBER detector is capable of outperforming the linear MMSE detector when either the
signature cross-correlation is high or the background noise is non-Gaussian.
1996 - 1999 Foschini et al. [179]– First discussed the application of linear ZF/MMSE detectors in multiple-antenna aided SDM-
[181] MIMO systems.
2006 Chen et al. [308] Proposed the MBER criterion based linear detector for multi-antenna aided MIMO systems.
2006 Burg et al. [390] Presented an algorithm and a corresponding VLSI architecture for the implementation of linear
MMSE detection in packet-based MIMO-OFDM communication systems. The algorithm also
supports the extraction of soft information for channel decoding.

2009 Yoshizawa et al. Reported a VLSI implementation for a 4 × 4 MIMO-OFDM transceiver relying on linear
[391] MMSE, which achieves a target data rate of 1 Gbps.
2014 Yin et al. [341] Presented the first application-specific integrated circuit (ASIC) implementation for the soft-
output linear MMSE detector based large-scale MIMO system which uses 128 BS antennas
to support 8 users, and a sum-rate of 3.8 Gbps was achieved.

this class of MIMO detectors have numerous variants due tectors is that they often suffer from error propagation. Hence
to the associated design flexibility, including the successive their performance only approaches that of the optimum ML
interference cancellation (SIC) detector [179]–[181], [251], based MUD when the interfering users have a much stronger
[400], the parallel interference cancellation (PIC) detector signal strength than the desired user. From this perspective,
[241], [248], [318], the multistage interference cancellation the weakest user benefits most from the employment of the
(MIC) detector [234], [244], [245], [401], and the decision- interference cancellation detector.
feedback detector (DFD) [253], [255], [256], [258] etc. The • SIC: In the most popular SIC based MIMO detector, a
interference cancellation based MIMO detectors are typically single symbol si is detected at a time. Then the inter-
capable of providing a significantly better performance than ference imposed by this particular symbol on the other
their linear counterparts at the expense of a higher complexity, symbols {sk }k6=i yet to be detected is subtracted after
especially in the absence of channel coding [260], albeit this recreating the interference upon generating the modulated
is not necessarily always the case. In practice, a common signal corresponding to this symbol. In this scheme, it
drawback of the interference cancellation based MIMO de- is most important to cancel the effect of the strongest

interfering signal before detecting the weaker signals. emphasized that although theoretically the SIC method
Therefore, the specific symbol detection ordering, which achieves the Shannon capacity in the multiple-access
can be designed according to various criteria, is quite channel by assuming perfectly error-free detection (hence
critical for the SIC detector’s performance. Some of the avoiding decision error propagation), this is not necessar-
typical ordering criteria for ordered SIC (OSIC) include ily true in practice, because the SIC method is sensitive
the decreasing signal-to-noise ratio (DSNR) criterion to decision error propagation, and hence MIMO detectors
[310], [311], the greatest signal-to-noise ratio (GSNR) that are more robust to decision error propagation might
criterion [312], the increasing mean-square error (IMSE) outperform the SIC detector in practice. Another fact
criterion [311], and the least mean-square error (LMSE) worth mentioning is that the performance degradation
criterion [313]–[315]. The SIC method performs well imposed by error propagation in the SIC detector can
when there is a substantial difference in the received be mitigated by accurate power control [408].
signal strength of the multiple simultaneously transmitted • PIC: Alternatively, in the PIC based MIMO detector, all
symbols. However, this condition is not always satisfied symbols are detected simultaneously. For each symbol,
in practical applications, which renders the SIC detector the coarse initial estimate of the interfering symbols can
potentially sensitive to decision error propagation. There- be used for regenerating the interference and then for
fore, the SIC detector is well-suited for multiple-access deducting it from each of the composite received signals.
systems suffering from the near-far problem, such as the Then this PIC detection process may be repeated for
family of CDMA or SDMA systems. In the SIC detector, several iterations. Therefore, sometimes the PIC detection
there is a need for detection reordering at each iteration of is also regarded as a MIC technique, or vice versa.
the SIC detector, and the number of detection iterations Compared to the SIC detector, the PIC detector has lower
increases linearly with the number of symbols in s. processing delay, and is more robust to inter-stream error
Therefore, for a system which has a high-dimensional propagation. However, its near-far resistance is inferior
transmitted symbol vector s, the SIC technique imposes to that of the SIC detector, because some users might
a substantial complexity, which ultimately increases the have much weaker received signal strength than others.
processing delay. The SIC detector designed for CDMA Hence, the PIC is suitable for similar-power signals, while
systems was first proposed by Viterbi [251]. Later it was the SIC performs better for different-power streams. In
studied extensively in [257], [259], [400], [402]–[407]. In the context of CDMA systems, the earliest contribution
the context of multi-antenna based SDM systems, the SIC to PIC may be attributed to Kohno et al. [241]–[243].
scheme was first studied by Foschini, Wolniansky, Golden Later significant contributions to PIC were also attributed
and Valenzuela in [179]–[181], and it was later studied to Yoon [247], [409], Divsalar [248], Buehrer [410] and
more comprehensively by numerous other researchers in Guo [411] et al.. In the context of multi-antenna MIMO
[305]–[307]. Among these schemes, Viterbi [251] pro- systems, the PIC detector was studied mainly in [316]–
posed an SIC scheme for a convolutionally coded direct- [318].
sequence CDMA (DS-CDMA) system and revealed that • MIC: In the MIC based MIMO detector, the first stage can
with the aid of the SIC based receiver, the aggregate be the conventional SUMF detector, the linear ZF/MMSE
data rate of all simultaneous users may approach the detector, the SIC detector or any other suboptimum
Shannon capacity of the Gaussian channel. It should be detector. The decisions made for all symbols s by the

Update residual
y b̂
signal vector Linear detector Decoder

yk+1 = yk − uk
Channel matrix
H
Re−encode and
uk re−modulate

Fig. 20. The basic principle of the SIC/DFD based MIMO detectors.

y1 IC for y1′ Final ŝ1

y2 user 1 y2′ decision

y, H Suboptimal .. .. Suboptimal .. ..
′
y ···
detector y detector y ′
k−1 k−1
yK IC for yK ′
Final ŝK
user K decision

Fig. 21. The basic principle of the PIC/MIC based MIMO detectors.

(n−1)th stage are employed as the input of the nth stage ML decoding Sphere decoding

for the sake of cancelling the MUI. Note that historically,

the MIC detector was developed independently of the
PIC, although they share similar concepts. The MIC de-
tector was first proposed by Timor for frequency-hopped R

CDMA (FH-CDMA) systems [233], [234]. Then, it was

extensively studied in the context of both asynchronous
DS-CDMA systems [244], [412] and synchronous DS-
CDMA systems [245], [401], [413]. An analytical frame-
work was proposed for adaptive MIC in [249].
Fig. 22. The basic principle of the SD based MIMO detectors. In short, it
• DFD: The concept of DFD is based on the same premise
solves the problem of finding the “closest” lattice point to a given vector y
as that of the family of decision-feedback equalizers in the skewed and rotated (caused by the MIMO channel) lattice Hs.
[417], [418]. Although DFD also relies on the SIC idea,
its emphasis is on the receiver filter’s optimization, which
consists of a feedforward filter and a feedback filter of their counterparts used in trellis decoding [423]–[436].
optimization. The first DFD scheme was proposed by Xie However, these tree-search based detectors did not attract as
et al. [253], [377] for asynchronous DS-CDMA systems. much attention as the linear detectors and the interference
Other major contributors of the subject of DFD include cancellation aided detectors in the era of CDMA systems.
Duel-Hallen [255], [256], [414], [415], who comprehen- The research interests related to tree-search based MIMO
sively investigated decision-feedback MUDs designed for detectors were largely stimulated by the seminal work of
both synchronous [255], [414] and asynchronous [256], Viterbo et al. [53], [319], who first applied the depth-first
[415] CDMA systems. Furthermore, Varanasi [258] ana- SD algorithm to the ML detection of multidimensional con-
lyzed the performance of a general class of DFDs using stellations transmitted over single-antenna fading channels.
a new performance metric constituted by the probability The basic principle of the SD algorithm is illustrated in Fig.
that at least one user is detected erroneously, and also 22. Compared to the optimal brute-force search based ML
proposed algorithms for determining the most beneficial decoding, the SD algorithm aims for reducing the computa-
detection ordering. tional complexity by only searching over the noiseless received
The basic principles of the SIC/DFD detectors and the signals that lie within a hypersphere of radius R around the
PIC/MIC detectors are illustrated in Fig. 20 and Fig. 21, received signal. Note that before it was applied in digital
respectively. The main contributions to the development of communications, the SD algorithm, also known as the Fincke-
the interference cancellation based MIMO detectors are sum- Pohst algorithm, had been reported in [437], [438]. Later,
marized in Table III. A more comprehensive exposition of the Agrell et al. [54] proposed to employ the Schnorr-Euchner
above-mentioned MIMO detectors developed in the context of (SE) refinement [439] of the Fincke-Pohst algorithm [437],
CDMA systems can be found in [48]–[51], [261], [419]. [438] for solving the CLPS problem, and they concluded that
the SE enumeration is more efficient than the Viterbo-Boutros
(VB) implementation [53] of the SD algorithm. More recently,
D. Tree-Search Based MIMO Detectors Damen et al. [57] proposed two improved SD algorithms for
The tree-search based MIMO detectors are arguably the finding the closest lattice point, both of which were shown
most popular detectors investigated in the era of multi-antenna to offer a significant complexity reduction compared to the
MIMO systems. This is because 1) the introduction of the VB-SD [53] and to the SE-SD [54], respectively. There exist
powerful SD algorithm happened to coincide with the devel- a number of other variants of the tree-search based MIMO
opment of multi-antenna MIMO techniques; 2) some profound detectors, which mainly fall into three categories: the depth-
research results on the CLPS problem showed that the tree- first tree-search detector [53], [54], [57]–[61], [80], [82], [319],
search MIMO detectors enjoy significant design flexibility in the breadth-first tree-search detector [62]–[64], [320] and the
terms of striking an attractive tradeoff between approaching best-first tree-search detector [68]–[72], [321], [322], [440].
the optimum ML performance and reducing the computational The major momentum which propels the enormous research
complexity. activities on tree-search based MIMO detectors is that they
Indeed, some tree-search based MUDs had been reported were shown to be capable of achieving near-ML performance,
earlier in the context of CDMA systems [252], [254], [420]– or even exact ML performance at the expense of significantly
[422]. For example, the so-called (depth-first) stack sequen- reduced complexity [58], [59], [441]–[443]. However, we
tial detection was proposed by Xie in [252], [420], while would like to emphasize that this claim is shown not true
the (breadth-first) K-best tree-search detection was proposed, in general [60], [444]. More specifically, Hassibi and Vikalo
again, by Xie in [254], [421], which was then further studied [58], [59], [441]–[443] first studied the expected complexity,
by Wei et al. in [422]. Looking back to the earlier history, averaged over the noise and over the lattice, of the Fincke-
because of the convertibility between the trellis structure and Pohst SD based MIMO detectors. It was claimed that although
the tree structure, the tree-search detection methods proposed the worst-case complexity of the SD algorithm is exponential,
for CDMA systems, including the classic M -algorithm [423], the expected complexity of the SD algorithm is polynomial,
[424] and T -algorithm [425]–[428], were actually extensions in fact, often roughly cubic, for a wide range of SNRs and

TABLE III
M ILESTONES IN THE DEVELOPMENT OF INTERFERENCE CANCELLATION MIMO DETECTORS

Year Authors Contributions

1974 - 1975 Bergmans and Cover [398], First demonstrated the effectiveness of the SIC concept from an information-theoretic
[399] perspective for broadcast channels.
1975 Carleial [197] First characterized the effectiveness of the SIC principle from an information-theoretic
perspective for interference channels.
1980-1981 Timor [233], [234] First proposed a two-stage [233] MUD and a multistage [234] MUD for FH-CDMA
systems employing multiple frequency-shift keying (MFSK) modulation; showed that
the mutual interference between the users of a FH-CDMA system may be significantly
reduced by making use of the well-defined algebraic structure of the users’ signature
waveforms, and that introducing an extra stage of interference cancellation may further
improve the detector’s performance.
1990 Viterbi [251] First conceived an SIC scheme for a convolutionally coded DS-CDMA system, and
revealed that with the aid of the SIC based receiver the aggregate data rate of all
simultaneous users can approach the Shannon capacity of the Gaussian channel.
1983 - 1990 Kohno et al. [241]–[243] First proposed a PIC based MUD for CDMA systems.

1988 - 1991 Varanasi et al. [244], [245], Designed and systematically characterized the MIC MUDs for both asynchronous and
[412], [413] synchronous CDMA systems.
1989 - 1990 Xie et al. [253], [377] First proposed a DFD based MUD for asynchronous DS-CDMA systems.
1991 - 1995 Duel-Hallen et al. [255], Systematically investigated DFD MUDs conceived for both synchronous [255], [414] and
[256], [414], [415] asynchronous [256], [415] CDMA systems from a receiver filter optimization perspective.
1996 - 1999 Foschini et al. [179]–[181] First discussed the ZF based SIC detector conceived for multiple-antenna aided SDM-
MIMO systems.
2002 Chin et al. [316] Extended the PIC detector to the multiple-antenna aided SDM-MIMO systems.
2003 Wübben et al. [310] Proposed a QR-decomposition (QRD) based MMSE-SIC detector for multiple-antenna
aided SDM-MIMO systems.
2003 Guo et al. [416] Presented a VLSI implementation of the V-BLAST detector for a 4 × 4 MIMO system
employing QPSK, and a detecting throughput of 128 Mbps was achieved.
2011 Studer et al. [318] Reported an ASIC implementation of a soft-input soft-output MMSE based PIC detector
for multiple-antenna aided SDM-MIMO systems.

number of antennas. Contrary to this claim, Jaldén and Ot- E. Lattice-Reduction Aided Detectors
tersten [60], [444] demonstrated that the expected complexity Lattice-reduction (LR) aided detectors constitute another
of SD based MIMO detectors is given by O(M βNt ), where important class of MIMO detectors, which rely on the alge-
β ∈ (0, 1] is a small factor depending on the value of braic concept of “lattice” originating from classic geometry
SNR. In other words, the expected complexity of the SD and group theory15 . A lattice in Rn is a discrete subgroup
algorithm is still exponential for fixed SNR values. Therefore, of Rn , which spans the real-valued vector space Rn . Each
in general the tree-search based MIMO detectors are not lattice in Rn can be generated from a basis of the vector space
efficient for MIMO systems which operate under low-SNR by forming all linear combinations with integer coefficients.
condition and/or have a large number of inputs. Notably, in In the MIMO transmission model of (1), the received signal
order to avoid the varying-complexity characteristic of tree- vector y is the noisy observation of the vector Hs, which is
search based MIMO detectors, recently a suboptimal fixed- in the lattice spanned by the column vectors H, since both
complexity SD (FCSD) was proposed for MIMO systems [80]. the real and imaginary parts of all the elements of s may be
It was shown that the FCSD √ achieves a near-ML performance transformed to integers by shifting and scaling.
with a complexity of O(M Nt ) [82] regardless of the specific A lattice typically has multiple sets of basis vectors. Some
SNR encountered, which represents an attractive solution of bases spanning the same lattice as H might be “closer” to
facilitating an efficient hardware implementation compared orthogonality than H itself. The process of finding a basis
to the conventional exponential-complexity SD. The main closer to orthogonality is referred to as LR. Theoretically,
contributions in the development of the depth-first tree-search finding the optimal (closest to orthogonality) set of basis
MIMO detectors, the breadth-first tree-search MIMO detectors vectors is computationally expensive. Therefore, in practice
and the best-first tree-search MIMO detectors are summarized LR algorithms typically aim for finding a “better” channel
in Table IV, Table V and Table VI, respectively.
15 Recall that the lattice perspective – many detection problems can be
interpreted as the problem of finding the closest lattice point [54], [57], [453]
– is also the foundation for the tree-search based MIMO detectors described
in Section VIII-D.

TABLE IV
M ILESTONES IN THE DEVELOPMENT OF THE TREE - SEARCH MIMO DETECTORS : D EPTH - FIRST TYPE

Year Authors Contributions

1981 - 1985 Pohst and Fincke [437], First proposed the SD algorithm, which is hence known as the Fincke-Pohst algorithm,
[438] for calculating vectors of short length in a lattice at a reduced complexity; this work
laid the mathematical foundation of applying the SD algorithm to the MIMO detection
problem.
1988 - 1990 Xie et al. [252], [420] First proposed a stack sequential decoding based MUD for asynchronous CDMA
systems; this detector is essentially a depth-first tree-search MIMO detector.
1993 - 1999 Viterbo et al. [53], [319] Applied the depth-first SD algorithm to the ML detection of multidimensional constel-
lations transmitted over single-antenna fading channels, which largely stimulated the
research interests of tree-search based MIMO detectors.
1994 Schnorr and Euchner [439] Proposed a more efficient variation, known as SE refinement, of the Fincke-Pohst SD
algorithm; the SE-SD algorithm was based on the lattice basis reduction philosophy and
represents a popular solution to the MIMO detection problem.
2001 - 2003 Hochwald et al. [74], [445] Proposed a complex-valued SD and the list-SD (LSD) for a FEC-coded MIMO using
IDD receiver, showing that a near-capacity performance can be achieved with the aid
of a soft-SD based IDD receiver.
2002 Agrell et al. [54] First proposed to use the SE refinement [439] of the Fincke-Pohst SD algorithm [437],
[438] to the CLPS problem, and concluded that the SE enumeration technique is more
efficient than the VB implementation [53] of the SD algorithm designed for MIMO
detection.
2003 Damen et al. [57] Proposed a pair of improved SD algorithms for finding the closest lattice point, both of
which were shown to offer a significant complexity reduction compared to the VB-SD
of [53] and to the SE-SD of [54].
2001 - 2005 Hassibi and Vikalo [58], Analyzed the expected complexity of the SD algorithm, and concluded that the expected
[59], [441]–[443] complexity of SD algorithm is dependent on both the problem size and the SNR; showed
that when the SNR is high, the expected complexity of SD can be approximated by a
polynomial function for a small problem size.
2004 - 2005 Jaldén and Ottersten [60], Further analyzed the expected complexity of the SD algorithm, and demonstrated that
[444] the expected complexity of the SD algorithm increases exponentially for a fixed SNR
with a search-space, which contradicts previous claims; therefore, strictly speaking, the
SD algorithm has an exponential lower bound in terms of both the expected complexity
as well as the worst-case complexity, although it can be efficient at high SNRs and for
problems of moderate size.
2004 Garrett et al. [446] First reported the VLSI implementation of a soft-output depth-first SD based detector
for a 4 × 4 16QAM MIMO system, achieving 38.8 Mbps over a 5-MHz channel.

2005 Burg et al. [61] Proposed two ASIC implementations of depth-first MIMO SD. The first ASIC attains
the ML performance with an average throughput of 73 Mb/s at an SNR of 20 dB; the
second ASIC achieves a throughput of 170 Mb/s at the same SNR with only a negligible
BER degradation. The proposed implementations rely on four key contributing factors
to achieve high throughput and low complexity: depth-first tree traversal with radius
reduction, implemented in a one-node-per-cycle architecture, the use of the l ∞ -instead
of l 2 -norm, and an efficient implementation of the enumeration approach.
2006 - 2008 Barbero et al. [80], [447] Proposed a noise-level independent fixed-complexity tree-search MIMO detector, which
overcomes the two main limitations of the SD from an implementation point of view:
its variable complexity and its sequential nature.
2008 Studer et al. [65] Presented the VLSI implementation of a soft-output depth-first SD based MIMO
detector, which demonstrated that single tree-search, sorted QR-decomposition, channel
matrix regularization, log-likelihood ratio clipping, and imposing runtime constraints
are the key ingredients for realizing soft-output MIMO detectors with near max-log
performance.
2009 Jaldén et al. [82] Presented analytical study of the error probability of the fixed-complexity SD in MIMO
systems having an arbitrary number of antennas, proving that it achieves the same
diversity order as the ML detector, regardless of the constellation size used.

TABLE V
M ILESTONES IN THE DEVELOPMENT OF THE TREE - SEARCH MIMO DETECTORS : B READTH - FIRST TYPE

Year Authors Contributions

1990 - 1993 Xie et al. [254], [421], [422] First conceived a breadth-first K-best tree search MUD for asynchronous CDMA
systems; proposed a joint signal detection and parameter estimation scheme based on
their breadth-first tree search MUD.
1997 Wei et al. [422] Studied both the M -algorithm and the T -algorithm based breath-first tree-search MUD
in the context of CDMA systems operating in fading channels.
2002 Wong et al. [62] Proposed and implemented a breadth-first K-best tree-search MIMO detector using a
VLSI architecture, which is capable of achieving a decoding throughput of 10 Mb/s at
100 MHz clock frequency in a 16-QAM aided (4 × 4)-element SDM-MIMO system.
2004 - 2006 Guo et al. [63], [448] Proposed and implemented both hard and soft SE-strategy based K-best tree-search
MIMO detectors, which are capable of supporting up to 53.3 Mb/s throughput at 100
MHz clock frequency for a 16-QAM aided (4 × 4)-element SDM-MIMO system.

2006 Wenk et al. [449] Presented a new VLSI architecture for the implementation of the K-best algorithm,
which relies on a more parallel approach and the ASIC design achieves up to 424
Mbps throughput.
2007 Chen et al. [64] Reported a VLSI implementation of a soft-output breadth-first tree search aided MIMO
detector for a (4 × 4)-element MIMO system employing 64-QAM, which is capable of
achieving a throughput of above 100 Mb/s.
2010 Patel et al. [450] Presented a VLSI architecture of a novel soft-output K-Best MIMO detector. This
implementation attains a peak throughput of 655 Mbps for a 4 × 4 64-QAM MIMO
system with 0.13um CMOS. Synthesis results in 65nm CMOS show the potential to
support a sustained throughput up to 2 Gbps, which may meet the requirements of for
mobile WiMAX and LTE-A standards.

TABLE VI
M ILESTONES IN THE DEVELOPMENT OF THE TREE - SEARCH MIMO DETECTORS : BEST- FIRST TYPE

Year Authors Contributions

2004 Fukatani et al. [321] Applied Dijkstra’s algorithm [451] for reducing the complexity of the SD based MIMO detector at
the expense of an increased storage size.
2004 Xu et al. [452] Applied the stack algorithm [432] to the best-first tree search based MIMO detector.
2006 Murugan et al. [66] Proposed a unified framework for tree search decoding, which encompasses all existing SDs as
special cases, hence unifying the depth-first search, the breadth-first search and the best-first search
based on the proposed framework.
2012 Chang et al. [72] A generalization of Dijkstra’s algorithm was developed as a unified tree-search detection framework;
the proposed framework incorporates a parameter triplet that allows the configuration of the memory
usage, detection complexity and the sorting dynamic associated with the tree-search algorithm; by
tuning the different parameters, beneficial performance-complexity tradeoffs are attained and a fixed-
complexity version can be conceived.
2012 Chang et al. [71] First applied the A* algorithm to the best-first tree-search based MIMO detection problem.
2012 Shen et al. [440] Proposed the algorithms and VLSI architectures for both the best-first soft- and hard-decision tree-
search based MIMO decoders in the context of a 4 × 4 64-QAM system using 65-nm CMOS
technology at 333 MHz clock frequency.

matrix H = HL, where the real and imaginary parts of Seysen’s algorithm [90], [91], [461] and Brun’s algorithm [92],
all the entries of the unimodular matrix L are integers and [462], [463], which are all suboptimal. The most popular LR
the determinant of L is ±1 or ±j. As a result, the LR algorithm is the LLL algorithm, which does not guarantee to
preprocessing technique is capable of improving the “quality” find the optimal basis, but it guarantees to find a basis within
of the MIMO channel matrices. a factor to the optimal one in polynomial time [88], [89],
There is a variety of LR algorithms developed by mathe- [457], [464]. For example, it was formally proved in [88] that
maticians [454]. Some of them, such as Gaussian reduction an upper bound on the average computational complexity of
[455], Minkowski reduction and Korkine-Zolotareff (KZ) re- LLL is O(Nt3 log Nt ), where Nt is the size of s. Furthermore,
Nt
duction [456], are capable of finding the optimal basis for a a tighter upper bound of O(Nt2 log Nr −N t +1
) was found in
lattice, but they are computationally prohibitive for communi- [89], where Nr is the size of y. Note that the worst-case
cations systems [456]–[459]. Other well-known LR algorithms computational complexity of the LLL algorithm can be infinite
include the Lenstra-Lenstra-Lovász (LLL) algorithm [460], [89], [457]. However, in practice the probability of the scenario

which leads to this worst-case complexity is zero. There are The PDA approach may also be applied for solving chal-
mainly two variants for the LLL algorithm, namely the real- lenging problems in digital communications. For example, it
valued LLL [84], [85] and the complex-valued LLL [95]– may be developed as a reduced-complexity design alternative
[98]. The real-valued LLL is applied to the real-valued MIMO to the optimal soft-decision aided MAP detectors/equalizers
system model of Section VII, while the complex-valued LLL of MIMO channels [104]–[133], and it is also applicable to
is designed for directly using it in the complex-valued MIMO channel estimation of MIMO systems [486], [487]. Since we
system model. Additionally, the authors of [465] proposed an mainly focus on MIMO detection in this paper, a more detailed
LLL algorithm which is not only applicable to the complex- discussion of the PDA-based MIMO channel estimation will
valued model, but also applicable to the Euclidean ring in not be included in the sequel. As far as the PDA-based
general. MIMO detection is concerned, it is Luo et al. [104] who first
In principle, LR can be combined with virtually all the applied the PDA approach to the MUD problem of BPSK-
other MIMO detectors to further improve their performance. modulated synchronous CDMA systems in 2001, showing
For example, the LR technique was used in conjunction with a near-optimum performance at a significantly lower com-
traditional linear ZF and nonlinear ZF-SIC detectors in [83], as putational complexity than the ML detector. Thereafter, the
well as with linear MMSE and nonlinear MMSE-SIC detectors PDA-based detector was naturally extended to the scenario of
in [85], [466], both achieving a substantial performance gain BPSK-modulated asynchronous CDMA systems [106], [107].
with little additional computational complexity. As a further Recently, it was also extended to the symbol detection of
advance over [83], a real-valued LLL-based LR algorithm was QAM-aided SDM-MIMO systems [111]–[113], [116], [120],
used in [84], which enables the application of the algorithm [130], striking an attractive tradeoff between the attainable
in MIMO systems with arbitrary numbers of dimensions. performance and the complexity imposed. More specifically,
In addition, it was shown in [84], [86] that LR can also in [111] a real-valued PDA (RPDA) was formulated for M -
be favorably applied in MIMO systems that use precoding. QAM constellations, which is based on the equivalent real-
The LLL based LR algorithm was shown to be capable of valued MIMO signal model previously discussed in Section
achieving the maximum attainable/full receive diversity in VII. Additionally, in [112] an approximate complex-valued
MIMO decoding [87]. The VLSI implementation of the LR PDA (A-CPDA) detector was proposed, in which the complex-
technique aided precoder and of the K-best MIMO detector valued Gaussian distribution is approximately characterized
was reported in [92] and [93], respectively. LR-aided soft- by a matched mean and a matched covariance only. Fur-
decision MIMO detectors are studied in [99]–[102]. Recently, thermore, the pseudo-covariance, as defined by Neeser and
element-based LR algorithms, which reduce the diagonal Massey in [301], was employed in [120] to fully characterize
elements of the noise covariance matrix of linear detectors and the complex-valued Gaussian distribution, and the resultant
thus enhance the asymptotic performance of linear detectors, formulation of complex-valued PDA (CPDA) [120] was shown
were proposed for large-scale MIMO systems [103]. The main to outperform both the RPDA [111] and the A-CPDA [112].
contributions in the development of LR-aided MIMO detection In these PDA-based MIMO detectors/equalizers, the prob-
are summarized in Table VII. abilities of the potential candidate symbols serve as the soft
input/output information and are typically estimated relying
on a self-iterative process. The key operation in this process is
F. Probabilistic Data Association Based Detectors the iterative approximation of the interference-plus-noise term
The PDA filter technique is a statistical approach originally obeying a multimodal Gaussian mixture distribution by an
invented by Bar-Shalom [467] in 1975 for the problem of ever-updated multivariate Gaussian distribution [104], [111]–
target tracking and surveillance in a cluttered environment, [113], [129]. Therefore, the performance of the PDA based
where measurements are unlabelled and may be spurious. To MIMO detectors is largely determined by the accuracy of
elaborate a little further, it was developed for solving the the iterative Gaussian approximation, whose impact on the
problem of plot-to-track association in a radar tracker. In this performance of the PDA based detectors was investigated
context, all of the potential candidates for association to a in [116]. In order to further improve the accuracy of the
specific track are combined into a single statistically most Gaussian approximation, the authors of [127] proposed a PDA
likely update, taking account of the statistical distributions of detector for correlated source bits using the joint detection of
both the tracking errors and the clutter, while assuming that multiple consecutive symbol vectors. Additionally, in [130],
only one of the candidates is the desired target with the rest of [488] a unified direct bit-based PDA approach was proposed
them representing false alarms. A major extension of the PDA for detecting linear mapping based high-order rectangular
filter is the joint probabilistic data association (JPDA) filter QAM symbols, which achieves a better performance at a
[468], [469], which takes account of the situation that multiple lower computational complexity than the CPDA detector of
targets are present out of all the potential candidates, and [120]. Furthermore, the PDA based iterative receiver design of
hence seeks to compute the joint decision probabilities for the FEC-coded MIMO systems was investigated in [132], [133],
multiple targets. In addition to their wide applications in radar, [489], [490], where it was revealed that the outputs of the
sonar, electro-optical sensor networks and navigation systems conventional PDA detectors in [104], [111], [112], [120],
[467]–[481], the PDA techniques have also been applied in the [127], [130] are indeed the normalized symbol likelihoods,
field of computer vision for solving the visual target tracking rather than the true APPs. Based on this insight, a pair of
problem [482]–[485]. PDA based MIMO iterative receivers, namely the approximate

TABLE VII
M ILESTONES IN THE DEVELOPMENT OF THE LR- BASED MIMO DETECTORS

Year Authors Contributions

1982 Lenstra et al. [460] First proposed the LLL algorithm for LR, which becomes the most popular LR algorithm used in
practice.

2002 Yao et al. [83] First applied the LR technique in conjunction with the traditional linear ZF and nonlinear ZF-SIC
detector, showing a substantial performance gain at a modest additional computational complexity.
2003-2004 Windpassinger and Presented LR-aided MIMO detectors relying on real-valued LLL algorithms.
Wübben et al. [84], [85]
2003-2004 Windpassingeret al. Proposed a real-valued LLL-based LR algorithm, which enables the application of the algorithm in
[84], [86] MIMO systems having arbitrary numbers of dimensions. It was also shown that LR can be favorably
applied in MIMO systems that use precoding.
2004 Wübben et al. [85], Extended the LR-aided linear ZF and nonlinear ZF-SIC MIMO detectors to their MMSE based
[466] counterparts.

2007 Taherzadeh et al. [87] Demonstrated that the LLL based LR algorithm is capable of achieving full receive diversity of
MIMO decoding.
2007-2008 Ling and Jaldén et al. Provided upper bounds for the average computational complexity of the LLL algorithm, namely
Nt
[88], [89] O(Nt3 log Nt ) and O(Nt2 log N −N +1
), respectively.
r t

2007-2008 Seethaler and Zhang et Studied the performance of the Seysen’s algorithm based LR techniques in MIMO detection problems.
al. [90], [91]
2007 Burg et al. [92] The first VLSI implementation of the LR technique relying on Brun’s algorithm was reported.
2008 Shabany et al. [93] Presented a VLSI implementation of the LR-aided K-best MIMO detector.

2008 Gestner et al. [94] The first VLSI implementation of the LR technique relying on the complex-valued LLL algorithm
was reported.
2005-2009 Gan and Ma et al. [95]– Proposed a number of complex-valued LLL algorithms which can be directly used in the complex-
[98] valued MIMO system model.
2006 - 2010 Silvola, Qi, Studied a range of LR-aided soft-output MIMO detectors, including LR aided K-best [100], LR-
Ponnampalam and aided MAP [99], LR-aided fixed radius algorithm, fixed candidates algorithm, fixed memory-usage
Zhang et al. [99]–[102] algorithm etc. [102].
2013 Zhou et al. [103] Proposed a class of element-based LR algorithms, which reduce the diagonal elements of the noise
covariance matrix of linear detectors and thus enhance the asymptotic performance of linear detectors,
in large-scale MIMO systems having hundreds of BS antennas.

and the exact Bayesian theorem based iterative PDA receivers or users, the better its performance, provided that the
were proposed in [132], [489] and [133], [490], respectively. channel is not overloaded (NI > NO ) or rank-deficient
Additionally, a distributed soft combining based PDA receiver [116]. However, due to its nature of approximation and
was conceived in [131], [218] for BS cooperation aided multi- iteration, the PDA based MIMO detector has not been
cell multiuser MIMO systems. well-understood compared to other mature MIMO detec-
The advantages of the PDA based detectors are summarized tors.
as follows. For the sake of more explicitly clarifying the fundamental
• First, it may achieve a near-optimal detection perfor- principle of the PDA based MIMO detector, its Gaussian
mance in certain circumstances, for example in the con- approximation process is conceptually illustrated in Fig. 23,
text of FEC-uncoded CDMA systems [104]–[107]. which is based on the assumption that the interference-plus-
• Second, it has a low complexity that increases no faster noise term to be processed by the PDA detector obeys a single-
than O Mi NI 4 per symbol vector, where Mi is the variate multimodal (four-modal) Gaussian mixture distribution
number of PDA iterations, while NI represents either the of pM (x) = p1 × f1 (x) + p2 × f2 (x) + p3 × f3 (x) + p4 × f4 (x).
number of users in CDMA [104]–[107], or the number of Here, the “single variate” assumption indicates that only a
transmit antennas in multi-antenna aided MIMO systems single interfering symbol, say si , exists for the other symbol
[111], [112], [120]. to be detected. In other words, a (2 × 2)-element VBLAST
• Third, it is inherently an soft-input soft-output algorithm, system is assumed. More specifically, the four-modal distri-
which is eminently applicable in combination with FEC bution observed in Fig. 23 stands for the case of a 4PAM-
codes such as convolutional codes, turbo codes [491], like scenario, which is a simplified real-valued example for
[492] or low-density parity-check (LDPC) codes [493], M -QAM. More specifically, pM (x) is constructed by a mix-
[494]. ture of four constituent Gaussian distributions f1 (x), f2 (x),
• Furthermore, the higher the number of transmit antennas f3 (x), f4 (x) having the same variance, but different means

0.25 0.25
(p1 , p2 , p3 , p4 ) = (0.25, 0.25, 0.25, 0.25) (p1 , p2 , p3 , p4 ) = (0.4, 0.1, 0.25, 0.25)

0.2 0.2

0.15 0.15

PDF

PDF
0.1 0.1

0.05 0.05

0 0
-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
x x

(a) Initial distribution (b) Updated distribution: 1st stage

0.25 0.25
(p1 , p2 , p3 , p4 ) = (0.6, 0.05, 0.15, 0.2) (p1 , p2 , p3 , p4 ) = (0.9, 0.03, 0.05, 0.02)

0.2 0.2

0.15 0.15
PDF

PDF
0.1 0.1

0.05 0.05

0 0
-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
x x

Fig. 23. The basic principle of the PDA based MIMO detectors: An example process of approximating a single-variate four-modal Gaussian
mixture distribution by a single Gaussian distribution.

of m1 = −3, m2 = −1, m3 = +1, m4 = +3 and different convex optimization problems are efficiently solvable, whereas
constituent probabilities of p1 , p2 , p3 , p4 . The four constituent nonconvex optimization problems are generally difficult to
probabilities correspond to the different probabilities that an solve. Convex optimization has a range of other important
interfering symbol has the value si = −3, si = −1, si = +1 properties. For example, in convex optimization problems,
and si = +3, respectively. every locally optimal solution constitutes the globally optimal
The main contributions to the development of the PDA solution, hence there is no risk of being trapped in a local
based MIMO detectors are summarized in Table VIII. optimum. Additionally, a rigorous optimality condition and
a duality theory exist for verifying the optimal nature of a
solution in convex optimization problems. For more details of
G. Semidefinite Programming Relaxation Based Detectors convex optimization, please refer to [324], [496].
In contrast to other MIMO detectors, the SDPR approach The SDPR based MIMO detectors have recently received
relies on a relaxation of the optimum MIMO detection problem substantial research attention [134]–[146]. The most attractive
to the mathematical model of semidefinite programming (SDP) characteristic of the SDPR-aided detectors is that they guar-
[323], [324], [495], which is a subfield of convex optimization antee a so-called polynomial-time16 worst-case computational
theory [496]. complexity, while achieving a high performance in certain
Convex optimization constitutes a subfield of the generic circumstances. Most of the existing SDPR detectors are depen-
mathematical optimization problem. It studies the minimiza- dent on the specific modulation constellation. To elaborate a
tion of a convex objective function over convex sets. Fig. little further, SDPR was first proposed for a BPSK-modulated
24 illustrates the basic framework of solving mathemati- CDMA system [134], [135], [498]–[501], and then it was
cal optimization problems using convex optimization. If a extended to quadrature phase-shift keying (QPSK) [136].
mathematical optimization problem is identified as a convex Simulation results showed that the SDPR detector is capable
optimization problem, it is mathematically regarded as an of achieving a near-ML BER performance, when using BPSK
easy problem, because powerful numerical algorithms, such [134] and QPSK [136]. The numerical and analytical results
as the interior-point methods [497], exist for efficiently find- of [137], [138] confirmed that the SDPR detector achieves
ing the optimal solution of convex problems. Therefore, in the maximum possible diversity order, when using BPSK for
mathematical optimization theory, the dividing line between transmission over a real-valued fading MIMO channel. The
the family of easy and difficult problems is convex versus
nonconvex, rather than linear versus nonlinear. In other words, 16 The computational complexity increases as a polynomial function of NI .

TABLE VIII
M ILESTONES IN THE DEVELOPMENT OF THE PDA- BASED MIMO DETECTORS

Year Authors Contributions

2001 Luo et al. [104] First applied the PDA filter technique to the MUD problem of synchronous CDMA
systems, showing a near-optimum performance at a significantly reduced complexity.
2002 Pham et al. [106] Proposed a PDA-Kalman MUD approach for asynchronous CDMA systems.
2003 Luo et al. [107] Conceived a sliding-window PDA based MUD approach for asynchronous CDMA
systems.
2003 Tan et al. [108] Designed a PDA-based IDD receiver for a coded CDMA system using BPSK modulation.
2004 Pham et al. [111] Extended the PDA detector to SDM-MIMO systems based on a real-valued signal model.
2004 Liu et al. [112] Proposed a PDA-based soft equalization scheme for frequency-selective MIMO channels.
2005 Liu et al. [113] Extended the PDA-Kalman MUD approach of [106] to the soft equalization of frequency-
selective MIMO channels.
2005 Latsoudas et al. Proposed a hybrid MIMO detector that combined the SD and the PDA detectors.
[114]
2005 Fricke et al. [116] Studied the impact of Gaussian approximation on the performance of the PDA based
MIMO detector.
2006 Jia et al. [120] Proposed a complex-valued PDA (CPDA) detector which takes the pseudo-covariance
into account during the derivation of the complex-valued PDA detector.
2008 Kim et al. [126] Applied the PDA method as a component of an iterative receiver designed for non-
coherent MIMO systems.
2008 Yang et al. [127] Proposed a PDA detector for correlated source bits using joint detection of multiple
consecutive symbol vectors.
2009 Mohammed et al. Applied the PDA algorithm to the problem of decoding large non-orthogonal space-time
[128] block codes (STBCs).
2011 Yang et al. [130] Proposed a unified direct bit-based PDA approach for detecting linear mapping based
high-order rectangular QAM symbols, achieving a better performance at a lower
computational complexity than the CPDA detector of [120].
2011 Yang et al. [131] Proposed a distributed soft combining based PDA receiver for BS cooperation aided
multi-cell multiuser MIMO systems.
2013 Yang et al. [132], Investigated the PDA based iterative receiver design for FEC-coded MIMO systems:
[133] revealed that the outputs of the conventional PDA detectors are indeed the normalized
symbol likelihoods rather than the true APPs; proposed a pair of PDA based MIMO
iterative receivers, namely the approximate and the exact Bayesian theorem based iterative
PDA receivers.

SDPR approach was also further developed for high-order capable of directly detecting the nonlinear Gray mapping aided
modulation schemes, such as for M -ary PSK scenario in rectangular high-order QAM symbols was proposed in [147],
[139], [140], and for high-order rectangular QAM in [141]– where the unequal error protection property (UEP) of QAM
[145]. As for the high-order QAM scenario, it was recently bits was exploited and the resultant SDPR detector was shown
shown in [146] that the so-called polynomial-inspired SDPR to outperform that of [145]. It should be noted, however, that
(PI-SDPR) [141], the bound-constrained SDPR (BC-SDPR) for high-order modulation scenarios, the performance of the
[143] and the virtually antipodal SDPR (VA-SDPR) [145] SDPR detectors is not as promising as that of the BPSK/QPSK
are actually equivalent in the sense that they arrive at the scenario. Therefore, there is a need to further improve the
same symbol decisions, and hence they exhibit an identical performance of the SDPR based MIMO detector designed
SER performance.17 Furthermore, a bit-based SDPR detector for high-order QAM constellations, while maintaining its
appealingly low computational complexity. The basic principle
17 More specifically, the solution equivalence of the PI-SDPR and BC- of SDPR based detectors is illustrated in Fig. 25, where the
SDPR schemes holds for 16-QAM and 64-QAM, while that between the blue boxes represent the technical challenges. Furthermore,
BC-SDPR and VA-SDPR techniques holds for any 4q -QAM scheme, where
q is a positive integer. The SDPR QAM detector of [144] exhibits a
the main contributions to the development of the SDPR based
better performance than that of [141], [143], [145], but has a much higher MIMO detectors are summarized in Table IX.
complexity.

TABLE IX
M ILESTONES IN THE DEVELOPMENT OF THE SDPR- BASED MIMO DETECTORS

Year Authors Contributions

2001 -2003 Tan et al. [134], [498], These authors independently proposed a SDPR based MUD for BPSK-
Ma et al. [135], [499], modulated synchronous CDMA systems; the eigen-decomposition based
and Wang et al. [500], method of [134], [498], [500], [501] and the randomization method of [135],
[501] [499] were proposed for converting the continuous-valued solution of the SDP
problem into the binary decision output. Additionally, a cutting plane method
was introduced for further improving the performance of the SDPR detector
for systems supporting a large number of users [134], [498]; it was shown that
the classic MUDs, such as the linear ZF/MMSE detector, can be interpreted
as degenerate forms of the SDPR based MUD [135], [499].
2003 Steingrimsson et al. Proposed a soft SDPR detector for an IDD receiver of QPSK-aided MIMO
[502] systems employing LDPC codes.
2004 Ma et al. [136] Conceived a SDPR based MUD for BPSK/QPSK aided asynchronous CDMA
systems with multiple receive antennas in frequency-selective fading environ-
ments; based on a flexible block alternating likelihood maximization (BALM)
principle, the large-scale ML detection problem was decomposed into smaller
subproblems, and each subproblem was solved by the SDPR detector.
2003 -2004 Luo et al. [139] and Ma Proposed SDPR detectors for general M -PSK aided synchronous CDMA
et al. [140] systems.
2005 Kisialiou et al. [137] Provided the first analytical study of the SDPR detector for BPSK-aided
MIMO systems; it was shown that the SDPR detector is capable of achieving
the same BER performance as that of the ML detector in high-SNR scenarios,
while at the low SNR region, the SDPR detector serves as a constant factor
approximation to the ML detector in large systems.
2005 Wiesel et al. [141] Designed a PI-SDPR detector for 16-QAM aided MIMO systems, which can
be extended to high-order M -QAM scenarios.
2006 Sidiropoulos et al. [143] Advocated a BC-SDPR detector for employment in high-order M -QAM aided
MIMO systems.
2007 Mao et al. [145] Proposed a VA-SDPR detector for M -QAM aided multicarrier CDMA(MC-
CDMA) systems; the method can directly operate at the bit-level in the context
of linear mapping based M -QAM.
2007 Mobasher et al. [144] Studied several variants of the SDPR detectors, and showed that it is possible
to further improve the SDPR detector’s performance by increasing their
complexity.
2008 Jaldén et al. [138] Analytically demonstrated that the SDPR based detector is capable of achiev-
ing full receive diversity order in BPSK-aided real-valued MIMO channels.
2009 Ma et al. [146] Demonstrated that the PI-SDPR of [141], the BC-SDPR of [143], and the
VA-SDPR of [145] are actually equivalent in the sense that they obtain the
same symbol decisions, and hence exhibit an identical SER performance.
2013 Yang et al. [147] Proposed a bit-based SDPR detector capable of directly detecting the nonlinear
Gray mapping aided rectangular high-order QAM symbols, where the unequal
error protection property (UEP) of QAM bits was exploited and the resultant
SDPR detector outperforms that of [145].

H. Detection in Rank-Deficient and Overloaded MIMO Sys- tenna pair. Then, the full-rank requirement may also be sat-
tems isfied. However, in some propagation scenarios, the channel
matrix H may not be of full-rank. For example, if the spatial
For MIMO detection, typically it is preferable to have a separation between the antenna elements of the transmitter
full-rank channel matrix, namely rank(H) = NI or NO , and/or the receiver is not large enough and hence the angular
In CDMA systems, this requirement may be satisfied by spread is small, the strong correlation between the antenna
using well-designed spreading codes. In multi-antenna SDM elements results in a rank-deficient channel matrix, i.e. we
systems, when an ideal rich scattering multipath environment have rank(H) < min(Nt , Nr ). Hence, the spatial degrees
is assumed, typically independently fading communications of freedom available are reduced, which translates into a
channels are encountered between each transmit/receive an- decreased MIMO capacity [503]–[508]. Furthermore, even if

Input modified SDPR detection [147], [526], [528], the metaheuris-

Convex Arguments Efficient Numerical Solver
System Model Optimization Problem e.g. interior point algorithms tics based detection [529]–[550] etc. It is possible to design
Problem Modelling

various scenario-dependent MIMO detectors for the rank-

Approximation
"Key Efforts"

Easy
deficient and overloaded MIMO systems. However, it seems
Hard
that the group detection strategy [514]–[519] and the search-
Problem Nonconvex
Output Solution based detection, regardless of the ML detector, the generalized
Formulation Optimization Problem
SD detector [56], [57], [520]–[527] and the metaheuristics
"Tricks"
based detector [529]–[550], are particularly suitable for rank-
deficient and overloaded MIMO scenarios.
Fig. 24. Framework of solving problems using convex optimization.

Fig I. Impact of Soft-Decision and Transmit Preprocessing on

H MIMO Detection
In previous sections we aim for understanding the funda-
y Construct input matrix
Q x b̂
mental properties of MIMO detection algorithms. However,
for standard Invoke SDP
solver
Post−processing if we look at the entire process of communication, the
SDP solver
assumption that only the receiver is responsible for signal
recovery represents a passive and incomplete strategy. In fact,
almost all practical systems invoke some form of encoding or
Mathematical formulation and relaxation
transmit preprocessing, such as FEC, space-time coding and
precoding/beamforming, to actively improve the performance
Fig. 25. The basic principle of the SDPR based MIMO detectors. of signal recovery or to reduce the detector’s computational
complexity from the transmitter side.
To elaborate a little further, when FEC is used, tremen-
the spatial separation between antenna elements is sufficiently dous efforts have been devoted to designing soft-input soft-
large, it is still possible that H is rank-deficient. This is due output MIMO detectors that can fit into the powerful “turbo
to the so-called “keyhole/pinhole effects” [508]–[513], which processing principle” [491], [492], [551]–[553] based IDD
may be simply understood as a diffraction phenomenon, where receiver architecture conceived for achieving near-optimum
a large obstacle with a small keyhole punched through it is performance. All the MIMO detectors reviewed in Section
placed between the MIMO transmitter and receiver, hence the VIII have their soft-decision versions to fit into IDD receivers.
only channel the radio wave can propagate through to the The representative contributions to iterative MIMO detection
receiver is the keyhole. Due to this effect, the channel matrix and decoding include: the optimal MAP detector based itera-
H is degenerate and has only a single degree of freedom, tive receiver [554]–[558], the expectation-maximization (EM)
i.e. we have rank(H) = 1, even though the entries of H are algorithm based soft-decision MUD [559], the soft-decision
uncorrelated. MMSE detector assisted iterative receiver [560]–[562], the
Another preferable condition for the detection in CDMA soft-decision SD based MIMO iterative receiver [65], [73],
and SDM-MIMO systems is that the system is not overloaded. [74], [81], [525], [563], the PDA detector based iterative
Then, the channel matrix is “fat” and does not have full receiver [108], [125], [132], [133], the soft-decision SDPR
column-rank (but it may still have full row-rank.). In the detector aided MIMO iterative receiver [502], [564], the soft-
multi-antenna scenario, this means that Nt ≤ Nr , while in decision multiple symbol differential SD (MSDSD) detector
CDMA systems, it means that the number of users is higher based non-coherent iterative receiver [565], and the soft-
than the dimension of the signal space/the processing gain decision iterative receiver for LS-MIMO systems [339], [440]
of the system. As far as MIMO detection is considered, both discussed in Section IX. Yet another important contribution to
the rank-deficient scenario discussed above and the overloaded IDD design is the extrinsic information transfer (EXIT) chart
scenario face the common challenge that the standard versions invented by ten Brink [566], [567], which is a powerful tool
of most of the representative MIMO detectors, such as the conceived not only for analyzing the convergence behavior of
linear ZF/MMSE detector, the ZF/MMSE-SIC detector and the iterative receivers, but also for assisting near-capacity wireless
original linear-decorrelation based PDA detector [104], [111] system design [74], [77], [342], [568]–[576]. For more details
that invoke the inverse of H, and the standard SD detector that on designing iterative MIMO receivers, please refer to [74],
invokes standard QRD or Cholesky’s factorization [53], usu- [342], [555], [557], [560], [577]–[580].
ally provide an unacceptably poor performance, because they Additionally, when space-time coding is employed, as we
are invoked for finding the solution of an under-determined discussed in Section IV, the optimal ML decoding can be
linear system subject to random noise. implemented with a simple separate symbol-by-symbol de-
Several strategies have been proposed to circumvent this coding strategy for orthogonal STBCs and with a linear-
predicament, such as the “pseudo-inverse” based linear de- complexity pairwise decoding strategy for quasi-orthogonal
tection [180], [239], the group detection [514]–[519], the STBCs [274]. As a result, the MIMO detection problem
generalized SD detector [56], [57], [520]–[527], the modified does not constitute a grave challenge for STBC aided MIMO
non-decorrelated PDA detection [112], [116], [121], [130], the systems. Similarly, when precoding/beamforming techniques

[84], [581]–[587] are employed in SDM-MIMO systems, the to estimate the noise variance. Furthermore, as indicated
interference between the transmit antennas may be signifi- in Section VIII-A2, the linear ZF detector is preferable in
cantly mitigated or even completely removed (when using ZF- interference-dominated scenarios, the linear MF detector
based linear precoding). As a result, the signal detection task of is preferable in noise-dominated scenarios, while the lin-
a precoded MIMO system becomes less challenging compared ear MMSE detector provides the highest SINR amongst
to that of SDM-MIMO systems invoking no preprocessing. all linear detectors, which makes it preferable in scenarios
The key insight gained here is that we can design an encoder where the noise and the interference have a comparable
or precoder to improve the performance or to reduce the level. Finally, the linear ZF and MMSE detectors exhibit
computational complexity of decoders/detectors. an inadequate performance in rank-deficient/overloaded
systems, where the number of independent inputs is
higher than the dimension of the received signals, while
J. Guidelines on Choosing the Right MIMO Detectors
the linear MF detector remains applicable.
As we mentioned in Section VIII-A2, the optimality of • The interference cancellation based MIMO detectors have
MIMO detectors strongly depends both on the criteria of a computational complexity between O(Nt3 ) and O(Nt4 ),
“goodness” and on the assumptions made for specific appli- and typically they have a much more attractive error
cation scenarios. Each type of MIMO detector has a different probability performance than the linear detectors. The-
performance-and-complexity profile18 , and each of them has oretically, the SIC/DFD based detectors are capable of
its own pros and cons. Therefore, in general there is no simple approaching the Shannon capacity, provided that there is
answer as to which algorithm is the best. In what follows, we no error propagation at any of the decision stages. By
first provide a qualitative comparison of the performance and contrast, the PIC/MIC based detectors do not have this
complexity characteristics of the MIMO detectors reviewed, property. Compared to PIC/MIC, the SIC/DFD detectors
and then summarize their analytical performance and com- are more sensitive to error propagation. However, this
plexity results in Table X. makes them preferable in the “near-far” scenario, where
• The MAP/ML based MIMO detectors relying on brute- the powers of different users are significantly different,
force search have the optimal VER performance (not such as those of the cell-center user and cell-edge user.
necessarily optimal BER or SER) and a computational Furthermore, the SIC/DFD detectors may have a higher
complexity which increases exponentially with the sys- processing delay than the PIC detectors. Additionally,
tem’s dimension (e.g. the number of transmit antennas or similar to the linear ZF and MMSE detectors, the in-
users). Naturally, their computational complexity order terference cancellation based detectors are not generally
O(M Nt ) is the highest amongst all the MIMO detec- applicable to the rank-deficient/overloaded scenarios.
tors. Additionally, the MAP/ML algorithms have to be • The tree-search based MIMO detectors, especially the
aware of the amplitudes of the transmitted symbols for K-best detectors, have the flexibility to achieve different
calculating the decision statistics. However, the MAP/ML error probability versus computational complexity trade-
detector is insensitive to channel imperfections and rank- offs. They are even capable of attaining the optimum
deficiency/overloading, and it has the best possible error ML performance at a reduced complexity. In contrast to
probability performance across the entire SNR region. other types of MIMO detectors, the tree-search based
When the system’s dimension is not too large, it remains detectors typically have a non-deterministic complex-
possible to implement the exact MAP/ML algorithm in ity, which is a challenge for hardware implementation,
practical systems with the aid of state-of-the-art VLSI albeit it is possible to design fixed-complexity tree-
technologies. search detectors. Therefore, the average computational
• The linear MIMO detectors typically have the lowest complexity, worst-case computational complexity and
computational complexity between O(Nt2 ) and O(Nt3 ), even the computational complexity distribution become
although there exist subtle differences amongst the com- important complexity metrics to consider. Note, however,
putational complexities of different linear detectors. Nat- that theoretically the tree-search based MIMO detectors
urally, in general they have the least attractive error prob- still have an exponentially increasing worst-case/average
ability performance. However, in some scenarios, such as computational complexity, in which case the exponent
the large-scale MIMO systems to be detailed in Section depends on different system parameters, such as the noise
IX, where the receiver side has a significantly higher variance. As a result, the tree-search based detectors may
number of antennas than the transmitter side, the linear not be suitable for low-SNR scenarios. Additionally, it
MF, ZF, MMSE, MBER etc. based MIMO detectors may may be possible to design tree-search based detectors
achieve a near-ML error probability. Additionally, the for rank-deficient/overloaded scenarios. Furthermore, the
linear MF and ZF detectors only have to know the channel tree-search based detectors rely on specific enumeration
matrix H, but the linear MMSE detector additionally has strategies, which by nature are not suitable for large-scale
MIMO systems that have a high number of inputs.
18 Generally, “performance” and “complexity” may be interpreted in various • The LR algorithms constitute a family of powerful
ways. For example, the “performance” can be error probability, robustness preprocessing techniques conceived for improving the
to system imperfections, configuration flexibility, application generality etc.,
while the “complexity” could be computational complexity, hardware/silicon “quality” of the effective channel matrix. They can be
complexity and so on. used in conjunction with all the other MIMO detectors.

TABLE X
P ERFORMANCE AND COMPLEXITY COMPARISON OF VARIOUS HARD - DECISION MIMO DETECTORS IN UNCODED SDM-MIMO SYSTEMS , WHERE
Nr ≥ Nt .

Detector Receive diversity order Error probability/DMT/asymptotic SNR penalty Worst-case computational

at high SNR analysis complexity order

per symbol vector

MAP/ML Nr [356], [358] See [237], [354], [356]–[360] Zero O(M Nt )

Linear ZF Nr − Nt + 1 [588] See [381], [384], [385], [588] High Between O(Nt2 ) and O(Nt3 )

Linear MMSE Nr − Nt + 1 [362], [363], [588] See [378], [588], Lower than that Between O(Nt2 ) and O(Nt3 )

[362], [363], [381]–[384] of linear ZF

ZF/MMSE-SIC Nr − Nt + 1 [44], [385], [589]–[591] See [44], [385], [589]–[591] Lower than that of Between O(Nt3 ) and O(Nt4 )

linear ZF/MMSE

ZF/MMSE-OSIC Nr − Nt + 1 [385], [589], [591], [592] See [385], [589], [591], [592] Lower than that of Between O(Nt3 ) and O(Nt4 )

ZF/MMSE-SIC

SD Nr [593] The same as that of ML, (Can be) zero O(M βNt ), β ∈ (0, 1]

if it is used for obtaining in general [58]–[60], [594], [595]

the exact ML solution.

√
2 Nt
FCSD min[Nr , (Nr − Nt )(p + 1) + (p + 1) ], See [82], [596] Approach zero O(M ) [82], [596]

where the first p levels at high-SNR

experience full search [82], [596].

K-best SD Nr − Nt + 1 to Nr , unknown for arbitrary K, flexible Between that of SIC and ML, Between that of SIC and ML,

depending on the value of K and suitable for VLSI implementation depending on K [63], [449] depending on K

LLL-LR-ZF/MMSE/SIC Nr [87], [96]–[98], [597] See [87], [96]–[98], [597]–[601] Can approach zero, Infinite in general (there

but the actual gap is no universal upper bound

depends on how well on the number of LLL

the particular channel iterations) [89], [457], but

can be reduced. can be substantially reduced

in many cases [600], [601]

PDA unknown See [116], [118], [249] Approach zero for large O(Mi Nt3 ) to O(Mi Nt4 ),

[121], [132], [602] Nt [104], [116], [118] where Mi is the number

of PDA iterations

SDPR Nr only for BPSK transmission over See [135], [137], [603] Typically near-ML Constellation-dependent:

real-valued Gaussian fading channels [138] [138], [146], [604] for BPSK/QPSK, but not O[(1 + Nt log2 M )3.5 ]

for high-order constellations [140], [502], [605]

unless complexity is [143], [147], [323], [606] to

√
substantially increased [144]. O[( M (2Nt + 1))6.5 ]

[143], [144], [146]

Since practically usable LR algorithms, such as the LLL and/or Nr when SNR increases. Additionally, on a fast-fading
algorithm, have a polynomially increasing computational MIMO channel, the achievable rate scales as
complexity, the LR-aided MIMO detectors do not have
min{Nt , Nr } log2 (1 + SNR), (25)
a significantly increased total computational complexity.
Hence, LR may be particularly useful for designing high- which indicates that the achievable rate of a MIMO system
performance MIMO detectors maintaining a low com- scales linearly with min{Nt , Nr }, and hence it is possible to
plexity, which is critical in numerous practical implemen- attain a high data rate using a large Nt and Nr . In conclusion,
tations. However, the LR techniques do not fundamentally fundamentally, using more antennas grants us higher degrees
change the pros and cons of their baseline detectors. of freedom in the spatial domain without increasing the
• Compared to the other MIMO detectors mentioned above, bandwidth occupied.
the SDPR and PDA based MIMO detectors are not well-
The LS-MIMO systems can be implemented in a variety
understood at the time of writing and they have not
of ways. For example, in the operational 3G/4G wireless
achieved the same degree of practical success, which is
communication systems, a point-to-point LS-MIMO system
partially indicated by the lack of VLSI implementations
might be constructed to provide high-throughput wireless
of these two types of detectors. Although SDPR detec-
backhaul connectivity between the BSs by using a large
tors have a favorable worst-case polynomial complexity,
number of antennas at each BS. However, apart from this
which
√ is roughly between O[(1 + Nt log2 M )3.5 ] and
6.5 particular application, it is typically quite challenging to con-
O[( M (2Nt + 1)) ], their achievable error probability
struct a point-to-point LS-MIMO system where the antenna
performance becomes less attractive for high-order mod-
elements can have a sufficiently high spatial separation to
ulations (but they may achieve near-ML performance for
guarantee a well-conditioned channel matrix. Furthermore,
BPSK and QPSK constellations). The Gaussian-mixture
achieving the attractive multiplexing gains promised by point-
approximation based PDA detectors operate in a way
to-point LS-MIMO schemes requires a high SNR. On the
similar to the classic soft interference cancellation, hence
other hand, a multiuser LS-MIMO system [608], [609] can
their computational complexity is similar to that of the
be envisaged, where the BS may be equipped with hundreds
soft SIC detectors, i.e. typically on the order between
of antenna elements and serves dozens of MSs each having
O(Mi Nt3 ) and O(Mi Nt4 ). As a result, the PDA detectors
only a few antennas. Additionally, the LS-MIMO may be
are also sensitive to error propagation, whilst exhibiting
implemented in the extremely high frequency (EHF) band (i.e.
the nice property of preferring a large number of inputs,
at millimeter wave (MMW) frequencies ranging from 30 to
provided that the receive dimensions are no less than
300 GHz and having wavelengths spanning from ten to one
that of the inputs. Hence, for certain large-scale MIMO
millimeter [33]). They may also be considered in the optical
scenarios, both SDPR and PDA based detectors may be
band for frequencies ranging from 300 GHz to 300 PHz and
attractive. Finally, for large-scale MIMO systems which
including the infrared, the visible and the ultraviolet band
have a similarly large number of transmit and receive
[610], [611]. Due to the adverse propagation properties, the
antennas, it might be valuable to resort to metaheuristics
coverage of the LS-MIMO systems operating in these high-
based algorithms, since all the other MIMO detectors
frequency bands might be significantly limited, hence they are
might either be excessively complex or fail to provide
more applicable to indoor environments [610] or small-cell
a high performance. Some of the metaheuristics based
scenarios [612]. For the sake of more explicit clarity, several
large-scale MIMO detectors are described in Section
typical antenna configurations and deployment scenarios of
IX-D.
LS-MIMOs are illustrated in Fig. 26 [613]. To elaborate
a little further, the simplest linear array propagates signals
IX. D ETECTION IN LS-MIMO S YSTEMS on the two-dimensional plane and it typically occupies a
Having reviewed the representative families of MIMO large physical area. By contrast, the rectangular, cylindrical
detection algorithms in Section VIII, let us now shift our and spherical arrays are capable of radiating signals to any
attention to the detection problem encountered in the emerg- directions in the three-dimensional space. These antenna arrays
ing massive/LS-MIMO systems [25], [373], [607], where are more complex, but also more compact, hence occupying a
dozens or even hundreds of antennas may be invoked and smaller physical area. Additionally, a virtual LS-MIMO may
an unprecedented spectral efficiency/diversity order may be be constructed relying on distributed antenna arrays, which
achieved. The major benefits of LS-MIMOs can be deduced may be exploited to enhance the indoor coverage or outdoor
from the following well-known results. For transmission over cooperation [613].
a quasi-static channel where a codeword occupies only a single As pointed out in Section VIII-A, the key motivation of
coherence-time and coherence-bandwidth interval, the outage studying the fundamental MIMO detection problem is that the
probability of a point-to-point MIMO link scales according to computational complexity of the optimum ML/MAP MIMO
detection increases exponentially with the problem size. There-
Proutage ∝ SNR−Nt Nr , (24)
fore, in principle the MIMO detection problem has intrinsically
which indicates that potentially a diversity order of (Nt × Nr ) embedded the “large-scale” concept. In this regard, people
may be achieved. In other words, the MIMO link’s reliability may argue that the detection in LS-MIMO systems is not
quantified in terms of its error rate falls exponentially with Nt a novel problem, and consequently the detectors conceived

for LS-MIMO systems might have no significant difference ..

.
..
.
..
.

with respect to the existing MIMO detectors, except for the

associated larger problem size. At first glance, this seems to be
Large−Scale
true. However, due to the limitations of practical applications, Wireless Channel Matrix

in the past large-scale MIMO systems have been regarded as

being impractical and most of the research focused on small-
scale MIMO systems. Nonetheless, in addition to their signif- ··· ···
icant link reliability and throughput benefits, the LS-MIMO
systems have been shown to enjoy some distinct advantages
that are not available in small-scale MIMO systems. These
benefits are mainly attributed to a range of relevant results
in random matrix theory [614], [615], and might be capable
of circumventing signal processing problems in LS-MIMO
systems. As a result, insights drawn from the detection in
small-scale MIMO systems might have to be carefully adapted Fig. 27. An example of Type-I (point-to-point) LS-MIMO systems.
for the large-scale MIMO environments. Depending on the
application scenarios considered, the detection problem of
large-scale MIMO systems may be categorized as follows.
ically, the antenna configuration of the Type-I system may be
A. Detection in Single-Cell/Noncooperative Multi-Cell LS- characterized by
Nt
MIMO Systems lim =c (26)
Nt ,Nr →∞ Nr
In a single-cell/noncooperative multi-cell MIMO system
with c being a positive constant. (26) indicates that both Nt
the BS is not concerned about the CCI imposed by the
and Nr tend to infinity at the same rate. By comparison,
transmissions of other cells. In this scenario, as pointed out in
in the Type-II system, only the receiver is equipped with a
Section III, the detection problems encountered in both the
large number of collocated antennas, while the total number
point-to-point MIMOs (see Fig. 8) and the multiple-access
of active antennas at the transmit side is significantly smaller.
MIMOs (see Fig. 9) can be characterized using the same
Hence, the antenna configuration of this system may be
received signal model of (1). From the antenna configuration
characterized as
point of view, there are two types of single-cell/noncooperative Nt
multi-cell LS-MIMO systems. As shown in Fig. 27, in the lim = 0. (27)
Nr →∞ Nr
Type-I system, a large number of collocated antennas may be
mounted on the receiver, and also a large number of collocated For the Type-I system, it has been shown that the empirical
or distributed antennas are used at the transmitter. Mathemat- distribution of the singular values of the random channel

linear

cylindrical

rectangular
spherical

distributed

Fig. 26. Typical antenna array configurations and deployment scenarios of

LS-MIMO systems [613].

.. .. ..
BS antennas is always beneficial, even when the SNR is
low and the channel estimate is poor. Furthermore, when the
. . .

number of BS antennas tends to infinity, the effects of both the

small-scale fast fading and uncorrelated noise are mitigated.
In other words, a large number of BS antennas, regardless
of whether the uplink or the downlink is considered, may
be exploited to trade for relevant performance improvements,
such as compensating for the low SNR and/or poor channel
estimates [25], [373], [607], [618].
However, in the noncooperative multi-cell scenario, due
to the so-called “pilot contamination” problem21 [25], [607],
the interference emanating from other cells does exist and
Fig. 28. An example of Type-II (multiuser) LS-MIMO systems. becomes the major limiting factor of the achievable perfor-
mance [25], [607]. Therefore, in order to further enhance
the achievable performance, the BS cooperation based multi-
matrix H converges to a deterministic limiting distribution19 cell joint processing philosophy has to be adopted [131], as
for almost all realizations of H, which is a result of the detailed below.
Marčenko and Pastur law [616]. In other words, as H becomes
larger (in terms of both Nt and Nr ), its singular values become
B. Detection in Cooperative Multi-cell Multiuser LS-MIMO
less sensitive to the actual distributions of the i.i.d. entries of
Systems
H, and the channel becomes more and more deterministic.
The Marčenko and Pastur law also shows that as the size of H As pointed out in Section III, the multi-cell transmission
increases, the diagonal entries of HH H become increasingly scenario is characterized by the so-called “MIMO interference
larger in magnitude than the off-diagonal entries. This is channels” of Fig. 11. Fundamentally, in order to cope with the
the so-called “channel-hardening” behavior, which may be interference, it may be desirable to transform the distributed
exploited for large-scale MIMO detection. To be more specific, model (such as the BSs of multiple cells) to a centralized
the matrix inversion invoked by many MIMO detectors such model. This may be achieved by centralized/distributed BS
as the ZF-aided detector, the MMSE-aided detector and the cooperation [24], [131], where multiple BSs of adjacent cells
PDA-aided detector etc., may be conveniently approximated may be connected via high-capacity optic fiber or microwave
using the series expansion technique for large-dimensional links, as shown in Fig. 29. As a result, effectively a phys-
random matrices [401]. Additionally, the channel-hardening ical/virtual super-BS is constructed to serve the cluster of
phenomenon may allow low-complexity detection algorithms collaborative cells, and this physical/virtual centralized model
to achieve a good performance for large-scale MIMO systems provides the performance upper bound of the original dis-
[617]. tributed system model. As far as detection is concerned, in
The Type-II system essentially deals with the MIMO de- principle most of the detection algorithms developed for the
tection problem encountered on an underloaded uplink,20 as single-cell/noncooperative multi-cell scenarios may be adapted
shown in Fig. 28. On the one hand, since the number of BS to the uplink of the cooperative multi-cell LS-MIMO system.
antennas may be significantly higher than the total number of The BS cooperation aided network MIMO detectors may be
active MS antennas, a very unbalanced antenna configuration designed based on two distinct philosophies, namely using
is encountered, which results in a high receive diversity order. either interference cancellation [215] or data fusion [131].
In the extreme case shown by (27), the receive diversity gain However, the employment of BS cooperation might result
obtained is so high that the impact of both the MUI and the in substantially increased backhaul traffic, which represents
noise diminishes. Additionally, the channel vectors associated one of the major challenges facing the BS cooperation aided
with distinct MSs may become asymptotically orthogonal. network MIMO.
Furthermore, another beneficial result of the Marčenko and
Pastur law [616] is that very tall (with large Nr ) and very
wide (with large Nt ) channel matrices H are very well condi- C. Applicability of Existing MIMO Detection Algorithms to
tioned. Therefore, in the Type-II system, even the simplest MF LS-MIMO
detector is capable of achieving a near-optimum performance An inherent characteristic of LS-MIMO systems is their
[373], [607]. Similarly, when considering the precoding based large dimension. Before investigating the applicability of ex-
downlink of the single-cell/noncooperative multi-cell TDD isting MIMO detection algorithms in the LS-MIMO context,
system, it was also revealed that increasing the number of we have to identify which specific type of LS-MIMO systems
is considered. On the one hand, in general most of the
19 This limiting distribution is the so-called quarter circle law [44, Chapter
existing MIMO detectors would be applicable to a Type-II
8.2].
20 On the downlink, large-scale MIMO precoding techniques may be em- LS-MIMO system, where it is possible that low-complexity
ployed, which facilitates the employment of simple receivers at each MS,
because the precoder is capable of eliminating the IAI at the transmitter with 21 This is essentially the interference caused by reusing pilot sequences in
the aid of accurate channel knowledge. adjacent cells.

Central Signal Processing Unit

w av
e 3Com CoreBuilder 5000
TM
Switching Hub

M
multivariate Gaussian approximation of the MUI for large-
cro icr
M
i
mgt fb fb fb fb tpl6 tpl6 fb fb fb fb 5302m 5302m
ow
av scale CDMA systems. This approach provided an alternative
e .. .. ..
..
.
..
.
..
. cmgt cmgt
. . .
analytical justification for the structure of the PDA based

re
ib
detectors. The associated performance analysis showed that

O
lF

pt
a

ic
ic

a
pt
the BER performance of the PDA detectors can be accurately

lF
O

ib
re
predicted and is close to the optimal detector’s performance
for large CDMA systems. Also in 2006, Liang et al. [325] pro-
posed a block-iterative generalized decision feedback equalizer
(a) Centralized BS cooperation based LS-MIMO cellular network. (BI-GDFE) for LS-MIMO systems using PSK constellations.
Their asymptotic performance analysis demonstrated that the
e
ow
av M
icr BI-GDFE closely approaches the single-user matched-filter
icr ow
.. .. ..
M av
e ..
.
..
.
..
.
bound (MFB) for large random MIMO channels, provided that
. . .

the received SNR is sufficiently high [326]. Furthermore, in

2007 Liang et al. derived both the limit and the asymptotic
distribution of the SINR for a class of MMSE receivers
invoked in large-scale CDMA systems supporting unequal-
power users. Their solution relied on the random matrix
Optical Fibre theory. They also proved that the limiting SINR converges
(b) Distributed BS cooperation based LS-MIMO cellular network. to a deterministic value when limK,N →∞ K N = c, where
K is the number of users, N is the number of degrees
Fig. 29. Centralized and distributed BS cooperation based multi-cell
multiuser based LS-MIMO systems, which is also known as network-MIMO. of freedom and c is a positive constant. Recall that this
insight is the same as that discussed in the context of (26).
Additionally, they proved that the SINR of each particular user
linear MIMO detectors might be capable of achieving near- is asymptotically Gaussian for large N and derived the closed-
optimum performance. Hence, the employment of more so- form expressions of the variance for the SINR variable under
phisticated MIMO detectors, such as the SD detector, may both real-valued spreading and complex-valued spreading. As
become unnecessary. On the other hand, some existing MIMO a further advance, in 2008 Liang et al. [327] investigated
detection algorithms that have been specifically tailored for the relationship between the MMSE-SIC receiver and the
conventional small-/medium-scale MIMO systems might not BI-GDFE receiver. The asymptotic performance of the two
be applicable to the Type-I LS-MIMO systems. To elaborate receivers was compared for large random MIMO channel
a little further, the family of tree-search based MIMO detec- matrices, and it was shown that the two receivers have a similar
tors, such as the popular SD detector that has a worst-case convergence behavior, and that both of them are capable of
computational complexity increasing exponentially with the achieving a BER performance approaching the single-user
number of transmit antennas (see Section VIII-D for more MFB for sufficiently high SNRs.
details), will become less feasible in the Type-I LS-MIMO Chockalingam et al. also made significant contributions to
systems. Nonetheless, it might still be invoked in the Type-II the LS-MIMO detection problem, mainly using a variety of
LS-MIMO systems. By contrast, the PDA algorithm [130]– metaheuristics based local search algorithms invoked from ma-
[133], which invokes the central limit theorem to perform chine learning/artificial intelligence [330]. More specifically,
stochastic interference modelling and imposes a polynomial- in 2008 they extended the low-complexity likelihood ascent
time worst-case computational complexity, will achieve an search (LAS) based MUD [619]–[621] to the Type-I LS-
attractive performance versus complexity tradeoff in the Type- MIMO system having up to 600 transmit and receive antennas
I LS-MIMO systems. Similarly, the convex optimization22 [176]. This detector relies on the local neighborhood search
based SDPR detectors, which also exhibit a polynomial-time and has its roots in the family of Hopfield neural network
worst-case complexity as a function of the number of transmit (HNN)23 based MUD algorithms [624]–[629]. It was shown
antennas, might potentially be applicable to the Type-I LS- that the LAS detector24 is capable of achieving near single-
MIMO systems [147]. input single-output AWGN performance in a fading LS-MIMO
environment at an average per-bit complexity of O(Nt Nr )
D. Recent Advances in LS-MIMO Detection [176]. Subsequently, they applied another local neighborhood
search based algorithm, namely the reactive tabu search (RTS)
The LS-MIMO systems have become a hot research topic
algorithm, to the detection of LS-MIMO systems. The RTS
following Marzetta’s seminal work [607]. However, in terms
detector was shown to perform better than the LAS detector,
of detection, several earlier works had touched upon this
because it relied on an efficient local minima exit strategy
topic from either a large system analysis or an asymptotic
[128]. Additionally, a class of belief propagation (BP) LS-
performance analysis perspective. To elaborate a little fur-
MIMO detectors relying on graphical models were proposed in
ther, in 2006 Tan and Rasmussen [118] derived a class of
asymptotically optimal nonlinear MMSE MUDs based on a 23 The HNN algorithms were also proposed for the restoration of large image
[622], [623].
22 Note that it is quite common to solve hundreds of unknown variables in 24 A multiple-output selection based LAS detector, namely the list LAS
a convex optimization problem. detector, was also proposed in [630] for LS-MIMO systems.

[331], [332], [337]25 . A range of other detectors were studied

by Chockalingam and his team in the context of LS-MIMO
systems, including the randomized Markov chain Monte Carlo
(R-MCMC) detector [333], the randomized search (RS) de-
tector [333], the Monte-Carlo-Sampling based detector which
jointly relies on a mixed Gibbs sampling (MGS) strategy
combined with a multiple restart (MR) strategy [334], and
the LR based26 detector [335]. Additionally, they applied AMC

40602

various detectors, including the MMSE detector, the PDA 602

U.S. AIR FORCE

detector, the LAS detector and the RTS detector, in high-rate

non-orthogonal STBC aided LS-MIMO systems [128], [328],
[329]. Furthermore, it was shown that non-binary LDPC coded
LS-MIMO systems are capable of achieving a near-capacity
performance with MMSE detection [336]. It should be noted
that in principle a variety of other metaheuristics based MUDs,
such as the genetic algorithm (GA) based MUD [529]–[535],
the ant colony optimization (ACO) based MUD [537]–[541],
the particle swarm optimization (PSO) aided MUD [545]–
[547], and the simulated annealing (SA) assisted MUD [548],
[549], may also be extended to the LS-MIMO context. Fig. 30. Conceptual illustration of MIMO radar systems.
Finally, some soft-input soft-output LS-MIMO detectors
having a relatively low complexity were proposed in [338], Modulator 1 Coherent
receiver 1
[339], which rely on the subspace marginalization aided inter- MMF

ference suppression (SUMIS) technique and an approximate ... ... MIMO

Source ... ... ... ...
message passing algorithm, respectively. The first ASIC design laser ... ...
detection
of an LS-MIMO detector invoking the truncated Neumann
Coherent
series expansion technique was reported in [340], [341], which Modulator N Local receiver N
oscillator
achieves a data rate of 3.8 Gb/s for a 3GPP Long Term
Evolution-Advanced (LTE-A) based LS-MIMO system having
Fig. 31. An example of MIMO optical fiber system.
128 BS antennas and supporting 8 users.

E. Applications of MIMO Detection Techniques in Other the so-called “MIMO radar”, as illustrated in Fig. 30, has been
Areas a hot research topic since the 2000s [646]–[652]. Additionally,
MIMO detection techniques may also be utilized in more MIMO signal processing techniques are also instrumental in
advanced scenarios. For example, as a promising technique mode-division multiplexing (MDM) based multimode fiber
to utilize the precious radio spectrum more efficiently and (MMF) optical communication systems, as shown in Fig.
flexibly, cognitive radio (CR) [631]–[633] has stimulated sub- 31. For more details on MIMO aided high-speed optical
stantial research interests over the past decade. Relying on communications, please refer to [653]–[659].
the software-defined radio (SDR) concept, CR is defined as
an intelligent wireless communication system that is capable X. S UMMARY AND C ONCLUSIONS
of learning from the environment and adapting to statistical
variations of the environment. Aiming for gleaning the benefits The concept of LS-MIMO systems may be regarded as
of both the CR and MIMO techniques, MIMO cognitive radio a paradigm shift in the wireless communication and signal
has also been studied from various perspectives [634]–[644]. processing community. In this large dimensional context, the
Furthermore, MIMO techniques may also be integrated with MIMO detection problem becomes even more challenging
the SDR or software-defined network (SDN) for 5G wireless and important. To facilitate a better understanding of MIMO
communication systems, where a network function virtual- detection techniques, in this survey, we provided a detailed
ization (NFV) based novel network architecture is envisaged clarification of the MIMO detection fundamentals, and recited
[645]. the half-a-century history of MIMO detection. We also pro-
Apart from their dominant applications in wireless com- vided concise discussions on the distinct detection strategies
munications, MIMO detection techniques also significantly for different types of LS-MIMO systems and concluded with
benefit a range of other research areas. For example, the idea the recent advances in LS-MIMO detection. Relevant insights
of MIMO signal processing was extended to radar design, and and lessons were extracted from the rich heritage of small-
/medium-scale MIMO detection. We note that when consider-
25 Very recently, Wu et al. also proposed an approximate message passing
ing the design of LS-MIMO detectors, it is necessary to first
algorithm based iterative detector for FEC-coded large-scale MIMO-OFDM identify which type of LS-MIMO system is considered. Specif-
systems [339].
26 The application of LR detectors in LS-MIMO systems was also investi- ically, the employment of several popular MIMO detectors,
gated by Zhou et al. in [103]. such as the SD based MIMO detectors, may become infeasible

in Type-I LS-MIMO systems, while some low-complexity lin- [17] G. Wu, S. Talwar, K. Johnsson, N. Himayat, and K. D. Johnson, “M2M:
ear MIMO detectors may achieve near-optimum performance from mobile to embedded internet,” IEEE Communications Magazine,
vol. 49, no. 4, pp. 36–43, Apr. 2011.
in Type-II LS-MIMO systems. Additionally, it was reported [18] Y. Zhang, R. Yu, S. Xie, W. Yao, Y. Xiao, and M. Guizani, “Home
that in the LS-MIMO context, local neighborhood search based M2M networks: architectures, standards, and QoS improvement,” IEEE
metaheuristics, Bayesian based message passing methods as Communications Magazine, vol. 49, no. 4, pp. 44–52, Apr. 2011.
[19] S.-Y. Lien and K.-C. Chen, “Massive access management for QoS
well as convex optimization based methods may strike a guarantees in 3GPP machine-to-machine communications,” IEEE Com-
promising performance versus complexity tradeoff. munications Letters, vol. 15, no. 3, pp. 311–313, Mar. 2011.
[20] United States Frequency Allocation Chart: August 2011 Edition.
National Telecommunications and Information Administration, United
ACKNOWLEDGMENT States Department of Commerce. [Online]. Available: http://www.ntia.
doc.gov/files/ntia/publications/spectrum wall chart aug2011.pdf
The authors would like to thank Prof. Hamid Jafarkhani [21] A. Paulraj, R. Nabar, and D. Gore, Introduction to Space-Time Wireless
(University of California, Irvine), Prof. Arogyaswami Paulraj Communications. Cambridge University Press, 2003.
(Stanford University) and anonymous reviewers for their in- [22] D. Gesbert, M. Kountouris, R. W. Heath, C.-B. Chae, and T. Salzer,
“Shifting the MIMO paradigm: from single-user to multiuser communi-
sightful comments and suggestions. cations,” IEEE Signal Processing Magazine, vol. 24, no. 5, pp. 36–46,
Oct. 2007.
R EFERENCES [23] J. Mietzner, R. Schober, L. Lampe, W. H. Gerstacker, and P. A.
Hoeher, “Multiple-antenna techniques for wireless communications –
[1] D. Mcqueen, “The momentum behind LTE adoption,” IEEE Commu- a comprehensive literature survey,” IEEE Communications Surveys &
nications Magazine, vol. 47, no. 2, pp. 44–45, Feb. 2009. Tutorials, vol. 11, no. 2, pp. 87–105, Second Quarter 2009.
[2] K. Lee, J. Lee, Y. Yi, I. Rhee, and S. Chong, “Mobile data offloading: [24] D. Gesbert, S. V. Hanly, H. Huang, S. Shamai (Shitz), O. Simeone,
how much can WiFi deliver?” in Proc. ACM 6th International and W. Yu, “Multi-cell MIMO cooperative networks: a new look at
Conference on emerging Networking EXperiments and Technologies interference,” IEEE Journal on Selected Areas in Communications,
(CoNEXT’10), Philadelphia, Pennsylvania, USA, Dec. 2010, pp. vol. 28, no. 9, pp. 1380–1408, Dec. 2010.
26:1–26:12. [Online]. Available: http://doi.acm.org/10.1145/1921168. [25] F. Rusek, D. Persson, B. K. Lau, E. G. Larsson, T. L. Marzetta,
1921203 O. Edfors, and F. Tufvesson, “Scaling up MIMO: opportunities and
[3] M. El-Sayed, A. Mukhopadhyay, C. Urrutia-Valdés, and Z. J. challenges with very large arrays,” IEEE Signal Processing Magazine,
Zhao, “Mobile data explosion: monetizing the opportunity through vol. 30, no. 1, pp. 40–60, Jan. 2013.
dynamic policies and QoS pipes,” Bell Labs Technical Journal, [26] E. Larsson, O. Edfors, F. Tufvesson, and T. Marzetta, “Massive
vol. 16, no. 2, pp. 79–99, Sep. 2011. [Online]. Available: MIMO for next generation wireless systems,” IEEE Communications
http://dx.doi.org/10.1002/bltj.20504 Magazine, vol. 52, no. 2, pp. 186–195, Feb. 2014.
[4] P. Ranganathan, “From microprocessors to nanostores: rethinking data- [27] F. Boccardi, R. W. Heath, A. Lozano, T. L. Marzetta, and P. Popovski,
centric systems,” IEEE Computer Magazine, vol. 44, no. 1, pp. 39–48, “Five disruptive technology directions for 5G,” IEEE Communications
Jan. 2011. Magazine, vol. 52, no. 2, pp. 74–80, Feb. 2014.
[5] B. Han, P. Hui, V. S. A. Kumar, M. V. Marathe, J. Shao, and A. Srini- [28] J. G. Andrews, S. Buzzi, W. Choi, S. V. Hanly, A. Lozano, A. C. K.
vasan, “Mobile data offloading through opportunistic communications Soong, and J. C. Zhang, “What will 5G be?” IEEE Journal on Selected
and social participation,” IEEE Transactions on Mobile Computing, Areas in Communications, vol. 32, no. 6, pp. 1065–1082, Jun. 2014.
vol. 11, no. 5, pp. 821–834, May 2012. [29] L. Lu, G. Li, A. Swindlehurst, A. Ashikhmin, and R. Zhang, “An
[6] “Cisco Visual Networking Index: Global Mobile overview of massive MIMO: Benefits and challenges,” IEEE Journal
Data Traffic Forecast Update, 2013-2018,” White of Selected Topics in Signal Processing, vol. 8, no. 5, Oct. 2014.
Paper, Cisco, Feb. 2014. [Online]. Available: [30] C. Han, T. Harrold, S. Armour, I. Krikidis, S. Videv, P. M. Grant,
http://www.cisco.com/c/en/us/solutions/collateral/service-provider/ H. Haas, J. S. Thompson, I. Ku, C.-X. Wang, T. A. Le, M. R. Nakhai,
visual-networking-index-vni/white paper c11-520862.pdf J. Zhang, and L. Hanzo, “Green radio: radio techniques to enable
[7] Cisco Visual Networking Index (VNI). Cisco. [Online]. energy-efficient wireless networks,” IEEE Communications Magazine,
Available: http://www.cisco.com/c/en/us/solutions/service-provider/ vol. 49, no. 6, pp. 46–54, Jun. 2011.
visual-networking-index-vni/index.html [31] H. Q. Ngo, E. G. Larsson, and T. L. Marzetta, “Energy and spectral
[8] G. Lawton, “Machine-to-machine technology gears up for growth,” efficiency of very large multiuser MIMO systems,” IEEE Transactions
IEEE Computer Magazine, vol. 37, no. 9, pp. 12–15, Sep. 2004. on Communications, vol. 61, no. 4, pp. 1436–1449, Apr. 2013.
[9] I. Cha, Y. Shah, A. U. Schmidt, A. Leicher, and M. V. Meyerstein, [32] M. Di Renzo, H. Haas, A. Ghrayeb, S. Sugiura, and L. Hanzo,
“Trust in M2M communication,” IEEE Vehicular Technology Maga- “Spatial modulation for generalized MIMO: Challenges, opportunities,
zine, vol. 4, no. 3, pp. 69–75, Sep. 2009. and implementation,” Proceedings of the IEEE, vol. 102, no. 1, pp.
[10] B. Emmerson, “M2M: the internet of 50 billion devices,” Huawei 56–103, Jan. 2014.
WinWin Magazine, no. 4, pp. 19–22, Jan. 2010. [Online]. Available: [33] T. S. Rappaport, S. Sun, R. Mayzus, H. Zhao, Y. Azar, K. Wang, G. N.
http://www-cnc.huawei.com/de/static/HW-072296.pdf Wong, J. K. Schulz, M. Samimi, and F. Gutierrez, “Millimeter wave
[11] M. Starsinic, “System architecture challenges in the home M2M net- mobile communications for 5G cellular: it will work!” IEEE Access,
work,” in Proc. IEEE 6th Annual Conference on Long Island Systems vol. 1, pp. 335–349, May 2013.
Applications and Technology (LISAT’10), Farmingdale, NY, USA, May [34] C. E. Shannon, “A mathematical theory of communication,” The Bell
2010, pp. 1–7. System Technical Journal, vol. 27, no. 3, pp. 379–423, Jul. 1948.
[12] Y. Chen and W. Wang, “Machine-to-machine communication in LTE- [35] S. V. Hanly and P. Whiting, “Information-theoretic capacity of multi-
A,” in Proc. IEEE 72nd Vehicular Technology Conference (VTC’10- receiver networks,” Telecommunication Systems, vol. 1, no. 1, pp. 1–42,
Fall), Ottawa, ON, Canada, Sep. 2010, pp. 1–4. 1993.
[13] Z. M. Fadlullah, M. M. Fouda, N. Kato, A. Takeuchi, N. Iwasaki, and [36] A. D. Wyner, “Shannon-theoretic approach to a Gaussian cellular
Y. Nozaki, “Toward intelligent machine-to-machine communications multiple-access channel,” IEEE Transactions on Information Theory,
in smart grid,” IEEE Communications Magazine, vol. 49, no. 4, pp. vol. 40, no. 6, pp. 1713–1727, Nov. 1994.
60–65, Apr. 2011. [37] S. Shamai (Shitz) and A. D. Wyner, “Information-theoretic considera-
[14] D. Niyato, L. Xiao, and P. Wang, “Machine-to-machine communi- tions for symmetric, cellular, multiple-access fading channels – Part I,”
cations for home energy management system in smart grid,” IEEE IEEE Transactions on Information Theory, vol. 43, no. 6, pp. 1877–
Communications Magazine, vol. 49, no. 4, pp. 53–59, Apr. 2011. 1894, Nov. 1997.
[15] S.-Y. Lien, K.-C. Chen, and Y. Lin, “Toward ubiquitous massive [38] ——, “Information-theoretic considerations for symmetric, cellular,
accesses in 3GPP machine-to-machine communications,” IEEE Com- multiple-access fading channels – Part II,” IEEE Transactions on
munications Magazine, vol. 49, no. 4, pp. 66–74, Apr. 2011. Information Theory, vol. 43, no. 6, pp. 1895–1911, Nov. 1997.
[16] R. Lu, X. Li, X. Liang, X. Shen, and X. Lin, “GRS: the green, relia- [39] O. Somekh and S. Shamai (Shitz), “Shannon-theoretic approach to a
bility, and security of emerging machine to machine communications,” Gaussian cellular multiple-access channel with fading,” IEEE Transac-
IEEE Communications Magazine, vol. 49, no. 4, pp. 28–35, Apr. 2011. tions on Information Theory, vol. 46, no. 4, pp. 1401–1425, Jul. 2000.

[40] S. Catreux, P. F. Driessen, and L. J. Greenstein, “Simulation results for [65] C. Studer, A. Burg, and H. Bolcskei, “Soft-output sphere decoding:
an interference-limited multiple-input multipleoutput cellular system,” algorithms and VLSI implementation,” IEEE Journal on Selected Areas
IEEE Communications Letters, vol. 4, no. 11, pp. 334–336, Nov. 2000. in Communications, vol. 26, no. 2, pp. 290–300, Feb. 2008.
[41] ——, “Attainable throughput of an interference-limited multiple-input [66] A. R. Murugan, H. El Gamal, M. O. Damen, and G. Caire, “A unified
multiple-output (MIMO) cellular system,,” IEEE Transactions on Com- framework for tree search decoding: rediscovering the sequential de-
munications, vol. 49, no. 8, pp. 1307–1311, Aug. 2001. coder,” IEEE Transactions on Information Theory, vol. 52, no. 3, pp.
[42] R. S. Blum, J. H. Winters, and N. R. Sollenberger, “On the capacity of 933–953, Mar. 2006.
cellular systems with MIMO,” IEEE Communications Letters, vol. 6, [67] R. Gowaikar and B. Hassibi, “Statistical pruning for near-maximum
no. 6, pp. 242–244, Jun. 2002. likelihood decoding,” IEEE Transactions on Signal Processing, vol. 55,
[43] H. Dai, A. F. Molisch, and H. V. Poor, “Downlink capacity of no. 6, pp. 2661–2675, Jun. 2007.
interference-limited MIMO system with joint detection,” IEEE Trans- [68] K. Lee and J. Chun, “ML symbol detection based on the shortest
actions on Wireless Communications, vol. 3, no. 2, pp. 442–453, Mar. path algorithm for MIMO systems,” IEEE Transactions on Signal
2004. Processing, vol. 55, no. 11, pp. 5477–5484, Nov. 2007.
[44] D. N. C. Tse and P. Viswanath, Fundamentals of Wireless Communi- [69] M. Stojnic, H. Vikalo, and B. Hassibi, “Speeding up the sphere decoder
cation. Cambridge, UK: Cambridge University Press, 2005. with H ∞ and SDP inspired lower bounds,” IEEE Transactions on
[45] P. van Emde Boas, “Another NP-complete partition problem and the Signal Processing, vol. 56, no. 2, pp. 712–726, Feb. 2008.
complexity of computing short vectors in a lattice,” Department of [70] T.-H. Kim and I.-C. Park, “High-throughput and area-efficient MIMO
Mathematics, University of Amsterdam, The Netherlands, Tech. Rep. symbol detection based on modified Dijkstra’s search,” IEEE Transac-
81-04, Apr. 1981. tions on Circuits and Systems I: Regular Papers, vol. 57, no. 7, pp.
[46] S. Verdú, “Computational complexity of optimum multiuser detection,” 1756–1766, Jul. 2010.
Algorithmica, vol. 4, no. 1-4, pp. 303–312, Jun. 1989. [Online]. [71] R. Y. Chang, W.-H. Chung, and S.-J. Lin, “A* algorithm inspired
Available: http://dx.doi.org/10.1007/BF01553893 memory-efficient detection for MIMO systems,” IEEE Wireless Com-
[47] D. Micciancio, “The hardness of the closest vector problem with munications Letters, vol. 1, no. 5, pp. 508–511, Oct. 2012.
preprocessing,” IEEE Transactions on Information Theory, vol. 47, [72] R. Y. Chang and W.-H. Chung, “Best-first tree search with probabilistic
no. 3, pp. 1212–1215, Mar. 2001. node ordering for MIMO detection: generalization and performance-
[48] A. Duel-Hallen, J. M. Holtzman, and Z. Zvonar, “Multiuser detection complexity tradeoff,” IEEE Transactions on Wireless Communications,
for CDMA systems,” IEEE Personal Communications Magazine, vol. 2, vol. 11, no. 2, pp. 780–789, Feb. 2012.
no. 2, pp. 46–58, Apr. 1995. [73] J. W. Choi, B. Shim, and A. C. Singer, “Efficient soft-input soft-output
[49] S. Moshavi, “Multi-user detection for DS-CDMA communications,” tree detection via an improved path metric,” IEEE Transactions on
IEEE Communications Magazine, vol. 34, no. 10, pp. 124–136, Oct. Information Theory, vol. 58, no. 3, pp. 1518–1533, Mar. 2012.
1996. [74] B. M. Hochwald and S. ten Brink, “Achieving near-capacity on a
[50] S. Verdú, Multiuser Detection. Cambridge, UK: Cambridge University multiple-antenna channel,” IEEE Transactions on Communications,
Press, 1998. vol. 51, no. 3, pp. 389–399, Mar. 2003.
[51] M. L. Honig, Ed., Advances in Multiuser Detection. John Wiley &
[75] J. Boutros, N. Gresset, L. Brunel, and M. Fossorier, “Soft-input soft-
Sons, Inc., 2009.
output lattice sphere decoder for linear channels,” in Proc. IEEE Global
[52] L. Bai and J. Choi, Low Complexity MIMO Detection. Springer, 2012. Telecommunications Conference (GLOBECOM’03), San Francisco,
[53] E. Viterbo and J. Boutros, “A universal lattice code decoder for fading USA, Dec. 2003, pp. 1583–1587.
channels,” IEEE Transactions on Information Theory, vol. 45, no. 7,
[76] H. Vikalo, B. Hassibi, and T. Kailath, “Iterative decoding for MIMO
pp. 1639–1642, Jul. 1999.
channels via modified sphere decoding,” IEEE Transactions on Wireless
[54] E. Agrell, T. Eriksson, A. Vardy, and K. Zeger, “Closest point search Communications, vol. 3, no. 6, pp. 2299–2311, Nov. 2004.
in lattices,” IEEE Transactions on Information Theory, vol. 48, no. 8,
[77] R. Wang and G. B. Giannakis, “Approaching MIMO channel capacity
pp. 2201–2214, Aug. 2002.
with soft detection based on hard sphere decoding,” IEEE Transactions
[55] M. O. Damen, A. Chkeif, and J.-C. Belfiore, “Lattice code decoder for
on Communications, vol. 54, no. 4, pp. 587–590, Apr. 2006.
space-time codes,” IEEE Communications Letters, vol. 4, no. 5, pp.
161–163, May 2000. [78] C. Studer and H. Bolcskei, “Soft-input sof-output single tree-search
sphere decoding,” IEEE Transactions on Information Theory, vol. 56,
[56] M. O. Damen, K. Abed-Meraim, and J.-C. Belfiore, “Generalised
no. 10, pp. 4827–4842, Oct. 2010.
sphere decoder for asymmetrical space-time communication architec-
ture,” Electronics Letters, vol. 36, no. 2, pp. 166–167, Jan. 2000. [79] M. Rachid and B. Daneshrad, “Iterative MIMO sphere decoding
[57] M. O. Damen, H. El Gamal, and G. Caire, “On maximum-likelihood throughput guarantees under realistic channel conditions,” IEEE Com-
detection and the search for the closest lattice point,” IEEE Trans- munications Letters, vol. 14, no. 4, pp. 342–344, Apr. 2010.
actions on Information Theory, vol. 49, no. 10, pp. 2389–2402, Oct. [80] L. G. Barbero and J. S. Thompson, “Fixing the complexity of the
2003. sphere decoder for MIMO detection,” IEEE Transactions on Wireless
[58] B. Hassibi and H. Vikalo, “On the sphere-decoding algorithm I. Communications, vol. 7, no. 6, pp. 2131–2142, Jun. 2008.
expected complexity,” IEEE Transactions on Signal Processing, vol. 53, [81] ——, “Extending a fixed-complexity sphere decoder to obtain likeli-
no. 8, pp. 2806–2818, Aug. 2005. hood information for Turbo-MIMO systems,” IEEE Transactions on
[59] H. Vikalo and B. Hassibi, “On the sphere-decoding algorithm II. gener- Vehicular Technology, vol. 57, no. 5, pp. 2804–2814, Sep. 2008.
alizations, second-order statistics, and applications to communications,” [82] J. Jaldén, L. G. Barbero, B. Ottersten, and J. S. Thompson, “The error
IEEE Transactions on Signal Processing, vol. 53, no. 8, pp. 2819–2834, probability of the fixed-complexity sphere decoder,” IEEE Transactions
Aug. 2005. on Signal Processing, vol. 57, no. 7, pp. 2711–2720, Jul. 2009.
[60] J. Jaldén and B. Ottersten, “On the complexity of sphere decoding [83] H. Yao and G. W. Wornell, “Lattice-reduction-aided detectors for
in digital communications,” IEEE Transactions on Signal Processing, MIMO communication systems,” in Proc. IEEE Global Telecommu-
vol. 53, no. 4, pp. 1474–1484, Apr. 2005. nications Conference (GLOBECOM’02), Nov. 2002, pp. 424–428.
[61] A. Burg, M. Borgmann, M. Wenk, M. Zellweger, W. Fichtner, and [84] C. Windpassinger and R. F. H. Fischer, “Low-complexity near-
H. Bolcskei, “VLSI implementation of MIMO detection using the maximum-likelihood detection and precoding for MIMO systems us-
sphere decoding algorithm,” IEEE Journal of Solid-State Circuits, ing lattice reduction,” in Proc. IEEE Information Theory Workshop
vol. 40, no. 7, pp. 1566–1577, Jul. 2005. (ITW’03), Paris, France, Mar. 2003, pp. 345–348.
[62] K.-W. Wong, C.-Y. Tsui, R. S. Cheng, and W.-H. Mow, “A VLSI [85] D. Wübben, R. Böhnke, V. Kühn, and K.-D. Kammeyer, “Near-
architecture of a K-best lattice decoding algorithm for MIMO chan- maximum-likelihood detection of MIMO systems using MMSE-based
nels,” in Proc. IEEE International Symposium on Circuits and Systems lattice reduction,” in Proc. IEEE International Conference on Commu-
(ISCAS’02), Scottsdale, AZ, USA, May 2002, pp. III–273–III–276. nications (ICC’04), Paris, France, Jun. 2004, pp. 798–802.
[63] Z. Guo and P. Nilsson, “Algorithm and implementation of the K-best [86] C. Windpassinger, R. F. H. Fischer, and J. B. Huber, “Lattice-reduction-
sphere decoding for MIMO detection,” IEEE Journal on Selected Areas aided broadcast precoding,” IEEE Transactions on Communications,
in Communications, vol. 24, no. 3, pp. 491–503, Mar. 2006. vol. 52, no. 12, pp. 2057–2060, Dec. 2004.
[64] S. Chen, T. Zhang, and Y. Xin, “Relaxed K-best MIMO signal detector [87] M. Taherzadeh, A. Mobasher, and A. K. Khandani, “LLL reduction
design and VLSI implementation,” IEEE Transactions on Very Large achieves the receive diversity in MIMO decoding,” IEEE Transactions
Scale Integration (VLSI) Systems, vol. 15, no. 3, pp. 328–337, 2007. on Information Theory, vol. 53, no. 12, pp. 4801–4805, Dec. 2007.

[88] C. Ling and N. Howgrave-Graham, “Effective LLL reduction for lattice [109] Y. Yin, Y. Huang, and J. Zhang, “Turbo equalization using probabilistic
decoding,” in Proc. IEEE International Symposium on Information data association,” in Proc. IEEE Global Telecommunications Confer-
Theory (ISIT’07), Nice, France, Jun. 2007, pp. 196–200. ence (GLOBECOM’04), Dallas, TX, USA, Dec. 2004, pp. 2535–2539.
[89] J. Jaldén, D. Seethaler, and G. Matz, “Worst- and average-case [110] Y. Huang and J. Zhang, “A generalized probabilistic data association
complexity of LLL lattice reduction in MIMO wireless systems,” in multiuser detector,” in Proc. IEEE International Symposium on Infor-
Proc. IEEE International Conference on Acoustics, Speech and Signal mation Theory (ISIT’04), Chicago, IL, USA, Jun. 2004, p. 529.
Processing (ICASSP’08), Las Vegas, NV, Mar. 2008, pp. 2685–2688. [111] D. Pham, K. R. Pattipati, P. K. Willett, and J. Luo, “A generalized
[90] D. Seethaler, G. Matz, and F. Hlawatsch, “Low-complexity MIMO data probabilistic data association detector for multiple antenna systems,”
detection using Seysen’s lattice reduction algorithm,” in Proc. IEEE IEEE Communications Letters, vol. 8, no. 4, pp. 205–207, Apr. 2004.
International Conference on Acoustics, Speech and Signal Processing [112] S. Liu and Z. Tian, “Near-optimum soft decision equalization for
(ICASSP’07), vol. 3, Apr. 2007, pp. III–53–III–56. frequency selective MIMO channels,” IEEE Transactions on Signal
[91] W. Zhang, F. Arnold, and X. Ma, “An analysis of Seysen’s lattice Processing, vol. 52, no. 3, pp. 721–733, Mar. 2004.
reduction algorithm,” Signal Processing, vol. 88, no. 10, pp. 2573– [113] ——, “A Kalman-PDA approach to soft-decision equalization for
2577, Oct. 2008. frequency-selective MIMO channels,” IEEE Transactions on Signal
[92] A. Burg, D. Seethaler, and G. Matz, “VLSI implementation of a lattice- Processing, vol. 53, no. 10, pp. 3819–3830, Oct. 2005.
reduction algorithm for multi-antenna broadcast precoding,” in Proc. [114] G. Latsoudas and N. D. Sidiropoulos, “A hybrid probabilistic data
IEEE International Symposium on Circuits and Systems (ISCAS’07), association-sphere decoding detector for multiple-input-multiple-output
New Orleans, LA, May 2007, pp. 673–676. systems,” IEEE Signal Processing Letters, vol. 12, no. 4, pp. 309–312,
[93] M. Shabany and P. G. Gulak, “The application of lattice-reduction to Apr. 2005.
the K-best algorithm for near-optimal MIMO detection,” in Proc. IEEE [115] Y. Jia, C. Andrieu, R. J. Piechocki, and M. Sandell, “Gaussian
International Symposium on Circuits and Systems (ISCAS’08), Seattle, approximation based mixture reduction for near optimum detection in
WA, May 2008, pp. 316–319. MIMO systems,” IEEE Communications Letters, vol. 9, no. 11, pp.
[94] B. Gestner, W. Zhang, X. Ma, and D. V. Anderson, “VLSI implementa- 997–999, Nov. 2005.
tion of a lattice reduction algorithm for low-complexity equalization,” [116] J. Fricke, M. Sandell, J. Mietzner, and P. A. Hoeher, “Impact of the
in Proc. IEEE International Conference on Circuits and Systems for Gaussian approximation on the performance of the probabilistic data
Communications (ICCSC’08), Shanghai, China, May 2008, pp. 643– association MIMO decoder,” EURASIP Journal on Wireless Commu-
647. nications and Networking, vol. 2005, no. 5, pp. 796–800, 2005.
[95] Y. H. Gan and W. H. Mow, “Complex lattice reduction algorithms for [117] J. Wang and S. Li, “MIMO turbo receiver with new probability
low-complexity MIMO detection,” in Proc. IEEE Global Telecommu- data association soft interference cancellation,” in Proc. International
nications Conference (GLOBECOM’05), St. Louis, MO, Dec. 2005, Conference on Communications, Circuits and Systems (ICCCAS’05)),
pp. 2953–2957. Hong Kong, China, May 2005, pp. 232–236.
[96] Y. H. Gan, C. Ling, and W. H. Mow, “Complex lattice reduction [118] P. H. Tan and L. K. Rasmussen, “Asymptotically optimal nonlinear
algorithm for low-complexity full-diversity MIMO detection,” IEEE MMSE multiuser detection based on multivariate gaussian approxi-
Transactions on Signal Processing, vol. 57, no. 7, pp. 2701–2710, Jul. mation,” IEEE Transactions on Communications, vol. 54, no. 8, pp.
2009. 1427–1428, Aug. 2006.
[119] Y. Cai, X. Xu, Y. Cheng, Y. Xu, and Z. Li, “A SISO iterative prob-
[97] X. Ma and W. Zhang, “Performance analysis for MIMO systems
abilistic data association detector for MIMO systems,” in Proc. 10th
with lattice-reduction aided linear equalization,” IEEE Transactions on
International Conference on Communication Technology (ICCT’06),
Communications, vol. 56, no. 2, pp. 309–318, Feb. 2008.
Guilin, China, Nov. 2006, pp. 1–4.
[98] ——, “Fundamental limits of linear equalizers: Diversity, capacity, and
[120] Y. Jia, C. M. Vithanage, C. Andrieu, and R. J. Piechocki, “Probabilistic
complexity,” IEEE Transactions on Information Theory, vol. 54, no. 8,
data association for symbol detection in MIMO systems,” Electronics
pp. 3442–3456, Aug. 2008.
Letters, vol. 42, no. 1, pp. 38–40, Jan. 2006.
[99] P. Silvola, K. Hooli, and M. Juntti, “Suboptimal soft-output MAP de-
[121] F. Cao, J. Li, and J. Yang, “On the relation between PDA and MMSE-
tector with lattice reduction,” IEEE Signal Processing Letters, vol. 13,
ISDIC,” IEEE Signal Processing Letters, vol. 14, no. 9, pp. 597–600,
no. 6, pp. 321–324, Jun. 2006.
Sep. 2007.
[100] X.-F. Qi and K. Holt, “A lattice-reduction-aided soft demapper for high- [122] S. Bavarian and J. K. Cavers, “Reduced complexity distributed base
rate coded MIMO-OFDM systems,” IEEE Signal Processing Letters, station processing in the uplink of cellular networks,” in Proc. IEEE
vol. 14, no. 5, pp. 305–308, May 2007. Global Telecommunications Conference (GLOBECOM’07), Washing-
[101] V. Ponnampalam, D. McNamara, A. Lillie, and M. Sandell, “On ton, DC, USA, Nov. 2007, pp. 4500–4504.
generating soft outputs for lattice-reduction-aided MIMO detection,” [123] ——, “Reduced-complexity belief propagation for system-wide MUD
in Proc. IEEE International Conference on Communications (ICC’07), in the uplink of cellular networks,” IEEE Journal on Selected Areas in
Glasgow, UK, Jun. 2007, pp. 4144–4149. Communications, vol. 26, no. 3, pp. 541–549, Apr. 2008.
[102] W. Zhang and X. Ma, “Low-complexity soft-output decoding with [124] Y. Jia, C. Andrieu, R. J. Piechocki, and M. Sandell, “Depth-first and
lattice-reduction-aided detectors,” IEEE Transactions on Communica- breadth-first search based multilevel sga algorithms for near optimal
tions, vol. 58, no. 9, pp. 2621–2629, Sep. 2010. symbol detection in mimo systems,” IEEE Transactions on Wireless
[103] Q. Zhou and X. Ma, “Element-based lattice reduction algorithms for Communications, vol. 7, no. 3, pp. 1052–1061, Mar. 2008.
large MIMO detection,” IEEE Journal on Selected Areas in Commu- [125] M. Grossmann and T. Matsumoto, “Nonlinear frequency domain
nications, vol. 31, no. 2, pp. 274–286, Feb. 2013. MMSE turbo equalization using probabilistic data association,” IEEE
[104] J. Luo, K. R. Pattipati, P. K. Willett, and F. Hasegawa, “Near optimal Communications Letters, vol. 12, no. 4, pp. 295–297, Apr. 2008.
multiuser detection in synchronous CDMA using probabilistic data [126] Y. J. D. Kim and J. Bajcsy, “An iterative receiver for non-coherent
association,” IEEE Communications Letters, vol. 5, no. 9, pp. 361– MIMO systems with differential encoding,” in Proc. 5th IEEE Con-
363, Sep. 2001. sumer Communications and Networking Conference (CCNC’08), Las
[105] J. Luo, “Improved multiuser detection in code-division multiple Vegas, NV, Jan. 2008, pp. 46–47.
access systems,” Ph.D. dissertation, Univ. of Connecticut, Storrs, May [127] S. Yang, T. Lv, X. Yun, X. Su, and J. Xia, “A probabilistic data associa-
2002. [Online]. Available: http://istec.colostate.edu/∼rockey/Papers/ tion based MIMO detector using joint detection of consecutive symbol
PhDThesis.pdf vectors,” in Proc. 11th IEEE Singapore International Conference on
[106] D. Pham, J. Luo, K. R. Pattipati, and P. K. Willett, “A PDA-Kalman Communication Systems (ICCS’08), Guangzhou, China, Nov. 2008, pp.
approach to multiuser detection in asynchronous CDMA,” IEEE Com- 436–440.
munications Letters, vol. 6, no. 11, pp. 475–477, Nov. 2002. [128] S. K. Mohammed, A. Chockalingam, and B. S. Rajan, “Low-
[107] J. Luo, K. R. Pattipati, and P. K. Willett, “A sliding window PDA for complexity near-MAP decoding of large non-orthogonal STBCs us-
asynchronous CDMA, and a proposal for deliberate asynchronicity,” ing PDA,” in IEEE International Symposium on Information Theory
IEEE Transactions on Communications, vol. 51, no. 12, pp. 1970– (ISIT’09), Seoul, Korea, Jul. 2009, pp. 1998–2002.
1974, 2003. [129] S. Bavarian and J. K. Cavers, “A new framework for soft decision
[108] P. H. Tan, L. K. Rasmussen, and J. Luo, “Iterative multiuser decoding equalization in frequency selective MIMO channels,” IEEE Transac-
based on probabilistic data association,” in Proc. IEEE International tions on Communications, vol. 57, no. 2, pp. 415 –422, Feb. 2009.
Symposium on Information Theory (ISIT’03), Yokohama, Japan, Jun. [130] S. Yang, T. Lv, R. G. Maunder, and L. Hanzo, “Unified bit-based
2003, pp. 301–301. probabilistic data association aided MIMO detection for high-order

QAM constellations,” IEEE Transactions on Vehicular Technology, [152] R. H. Roy and B. Ottersten, “Spatial division multiple access wireless
vol. 60, no. 3, pp. 981–991, Mar. 2011. communication systems,” U.S. Patent 5 515 378, May 7, 1996.
[131] ——, “Distributed probabilistic-data-association-based soft reception [153] D. Gerlach, “Adaptive transmitting antenna arrays at the base station
employing base station cooperation in MIMO-aided multiuser multicell in mobile radio networks,” Ph.D. dissertation, Department of Electrical
systems,” IEEE Transactions on Vehicular Technology, vol. 60, no. 7, Engineering, Stanford University, California, USA, 1995.
pp. 3532–3538, Sep. 2011. [154] B. Ottersten, “Array processing for wireless communications,” in Proc.
[132] S. Yang, L. Wang, T. Lv, and L. Hanzo, “Approximate Bayesian 8th IEEE Signal Processing Workshop on Statistical Signal and Array
probabilistic-data-association-aided iterative detection for MIMO sys- Processing, Corfu, Greece, Jun. 1996, pp. 466–473.
tems using arbitrary M -ary modulation,” IEEE Transactions on Vehic- [155] R. H. Roy, “Spatial division multiple access technology and its applica-
ular Technology, vol. 62, no. 3, pp. 1228–1240, Mar. 2013. tion to wireless communication systems,” in Proc. IEEE 47th Vehicular
[133] S. Yang, T. Lv, R. G. Maunder, and L. Hanzo, “From nominal to true a Technology Conference (VTC’97), Phoenix, AZ, USA, May 1997, pp.
posteriori probabilities: an exact Bayesian theorem based probabilistic 730–734.
data association approach for iterative MIMO detection and decoding,” [156] M. P. Lotter and P. van Rooyen, “Space division multiple access for
IEEE Transactions on Communications, vol. 61, no. 7, pp. 2782–2793, cellular CDMA,” in Proc. IEEE 5th International Symposium on Spread
Jul. 2013. Spectrum Techniques and Applications (ISSSTA’98), Sun City, South
[134] P. H. Tan and L. K. Rasmussen, “The application of semidefinite Africa, Sep. 1998, pp. 959–964.
programming for detection in CDMA,” IEEE Journal on Selected Areas [157] P. Vandenameele, L. Van der Perre, and M. Engels, Space Division
in Communications, vol. 19, no. 8, pp. 1442–1449, Aug. 2001. Multiple Access for Wireless Local Area Networks. Norwell, MA,
[135] W.-K. Ma, T. N. Davidson, K. M. Wong, Z.-Q. Luo, and P.-C. Ching, USA: Kluwer, 2001.
“Quasi-maximum-likelihood multiuser detection using semi-definite [158] A. Goldsmith, Wireless Communications, 1st ed. Cambridge Univer-
relaxation with application to synchronous CDMA,” IEEE Transactions sity Press, 2005.
on Signal Processing, vol. 50, no. 4, pp. 912–922, Apr. 2002. [159] A. V. Oppenheim, A. S. Willsky, and S. H. Nawab, Signals and Systems,
[136] W.-K. Ma, T. N. Davidson, K. M. Wong, and P.-C. Ching, “A block 2nd ed. Upper Saddle River, NJ, USA: Prentice-Hall, 1996.
alternating likelihood maximization approach to multiuser detection,” [160] B. Sklar, Digital Communications: Fundamentals and Applications,
IEEE Transactions on Signal Processing, vol. 52, no. 9, pp. 2600–2611, 2nd ed. Upper Saddle River, NJ, USA: Prentice-Hall, 2001.
Sep. 2004. [161] T. S. Rappaport, Wireless Communications: Principles and Practice,
[137] M. Kisialiou and Z.-Q. Luo, “Performance analysis of quasi-maximum- 2nd ed. Upper Saddle River, NJ: Prentice Hall, 2002.
likelihood detector based on semi-definite programming,” in Proc.
[162] J. G. Proakis, Digital Signal Processing: Principles, Algorithms and
IEEE International Conference on Acoustics, Speech, and Signal Pro-
Applications, 4th ed. Pearson Education, 2007.
cessing (ICASSP’05), vol. 3, Philadelphia, PA, Mar. 2005, pp. III/433–
[163] S. Haykin, Adaptive Filter Theory, 4th ed. Upper Saddle River, NJ:
III/436.
Prentice Hall, 2002.
[138] J. Jaldén and B. Ottersten, “The diversity order of the semidefinite
relaxation detector,” IEEE Transactions on Information Theory, vol. 54, [164] A. Jamalipour, T. Wada, and T. Yamazato, “A tutorial on multiple
no. 4, pp. 1406–1422, Apr. 2008. access technologies for beyond 3G mobile networks,” IEEE Commu-
nications Magazine, vol. 43, no. 2, pp. 110–117, Feb. 2005.
[139] Z.-Q. Luo, X. Luo, and M. Kisialiou, “An efficient quasi-maximum
likelihood decoder for PSK signals,” in Proc. IEEE International [165] S. Kandukuri and S. Boyd, “Optimal power control in interference-
Conference on Acoustics, Speech, and Signal Processing (ICASSP’03), limited fading wireless channels with outage-probability specifica-
vol. 6, Hong Kong, China, Apr. 2003, pp. VI/561–VI/564. tions,” IEEE Transactions on Wireless Communications, vol. 1, no. 1,
[140] W.-K. Ma, P.-C. Ching, and Z. Ding, “Semidefinite relaxation based pp. 46–55, Jan. 2002.
multiuser detection for M-ary PSK multiuser systems,” IEEE Trans- [166] I. Koffman and V. Roman, “Broadband wireless access solutions based
actions on Signal Processing, vol. 52, no. 10, pp. 2862–2872, Oct. on OFDM access in IEEE 802. 16,” IEEE Communications Magazine,
2004. vol. 40, no. 4, pp. 96–103, Apr. 2002.
[141] A. Wiesel, Y. C. Eldar, and S. Shamai (Shitz), “Semidefinite relaxation [167] M. Morelli, C.-C. J. Kuo, and M.-O. Pun, “Synchronization techniques
for detection of 16-QAM signaling in MIMO channels,” IEEE Signal for orthogonal frequency division multiple access (OFDMA): A tutorial
Processing Letters, vol. 12, no. 9, pp. 653–656, Sep. 2005. review,” Proceedings of the IEEE, vol. 95, no. 7, pp. 1394 –1427, Jul.
[142] Y. Yang, C. Zhao, P. Zhou, and W. Xu, “MIMO detection of 16-QAM 2007.
signaling based on semidefinite relaxation,” IEEE Signal Processing [168] M. Necker, “Interference coordination in cellular OFDMA networks,”
Letters, vol. 14, no. 11, pp. 797–800, Nov. 2007. IEEE Network, vol. 22, no. 6, pp. 12–19, Nov. 2008.
[143] N. D. Sidiropoulos and Z.-Q. Luo, “A semidefinite relaxation approach [169] N. Benvenuto, R. Dinis, D. Falconer, and S. Tomasin, “Single carrier
to MIMO detection for high-order QAM constellations,” IEEE Signal modulation with nonlinear frequency domain equalization: An idea
Processing Letters, vol. 13, no. 9, pp. 525–528, Sep. 2006. whose time has come — again,” Proceedings of the IEEE, vol. 98,
[144] A. Mobasher, M. Taherzadeh, R. Sotirov, and A. K. Khandani, “A near- no. 1, pp. 69–96, Jan. 2010.
maximum-likelihood decoding algorithm for MIMO systems based [170] S. Hara and R. Prasad, “Overview of multicarrier CDMA,” IEEE
on semi-definite programming,” IEEE Transactions on Information Communications Magazine, vol. 35, no. 12, pp. 126–133, Dec. 1997.
Theory, vol. 53, no. 11, pp. 3869–3886, Nov. 2007. [171] L.-L. Yang and L. Hanzo, “Performance of generalized multicarrier
[145] Z. Mao, X. Wang, and X. Wang, “Semidefinite programming relaxation DS-CDMA over Nakagami-m fading channels,” IEEE Transactions on
approach for multiuser detection of QAM signals,” IEEE Transactions Communications, vol. 50, no. 6, pp. 956–966, Jun. 2002.
on Wireless Communications, vol. 6, no. 12, pp. 4275–4279, Dec. 2007. [172] ——, “Multicarrier DS-CDMA: A multiple access scheme for ubiq-
[146] W.-K. Ma, C.-C. Su, J. Jaldén, T.-H. Chang, and C.-Y. Chi, “The uitous broadband wireless communications,” IEEE Communications
equivalence of semidefinite relaxation MIMO detectors for higher-order Magazine, vol. 41, no. 10, pp. 116–124, Oct. 2003.
QAM,” IEEE Journal of Selected Topics in Signal Processing, vol. 3, [173] F. Adachi, G. Garg, S. Takaoka, and K. Takeda, “Broadband CDMA
no. 6, pp. 1038–1052, Dec. 2009. techniques,” IEEE Wireless Communications Magazine, vol. 12, no. 2,
[147] S. Yang, T. Lv, and L. Hanzo, “Semidefinite programming relaxation pp. 8–18, Apr. 2005.
based virtually antipodal detection for MIMO systems using Gray- [174] L. Hanzo, L.-L. Yang, E.-L. Kuan, and K. Yen, Single- and Multi-
coded high-order QAM,” IEEE Transactions on Vehicular Technology, Carrier DS-CDMA: Multi-User Detection, Space-Time Spreading, Syn-
vol. 62, no. 4, pp. 1667–1677, 2013. chronisation, Networking and Standards. Chichester, UK: Wiley-IEEE
[148] E. G. Larsson, “MIMO detection methods: how they work,” IEEE Press, 2003.
Signal Processing Magazine, vol. 26, no. 3, pp. 91–95, May 2009. [175] L.-L. Yang, Multicarrier Communications. John Wiley & Sons, 2009.
[149] W. C. Y. Lee, “Overview of cellular CDMA,” IEEE Transactions on [176] K. Vardhan, S. K. Mohammed, A. Chockalingam, and B. S. Rajan,
Vehicular Technology, vol. 40, no. 2, pp. 291–302, May 1991. “A low-complexity detector for large MIMO systems and multicarrier
[150] P. Jung, P. W. Baier, and A. Steil, “Advantages of CDMA and spread CDMA systems,” IEEE Journal on Selected Areas in Communications,
spectrum techniques over FDMA and TDMA in cellular mobile radio vol. 26, no. 3, pp. 473–485, Apr. 2008.
applications,” IEEE Transactions on Vehicular Technology, vol. 42, [177] M. Kadir, S. Sugiura, S. Chen, and L. Hanzo, “Unified MIMO-
no. 3, pp. 357–364, Aug. 1993. multicarrier designs: A space-time shift keying approach,” IEEE Com-
[151] A. J. Viterbi, CDMA: Principles of Spread Spectrum Communication, munications Surveys and Tutorials, vol. 17, no. 2, pp. 550–579, Second
1st ed. Addison-Wesley, 1995. Quarter 2015.

[178] M. Tüchler and A. C. Singer, “Turbo equalization: An overview,” IEEE [199] H. Sato, “The capacity of the gaussian interference channel under
Transactions on Information Theory, vol. 57, no. 2, pp. 920–952, Feb. strong interference,” IEEE Transactions on Information Theory, vol. 27,
2011. no. 6, pp. 786–788, Nov. 1981.
[179] G. J. Foschini, “Layered space-time architecture for wireless commu- [200] M. H. M. Costa, “On the gaussian interference channel,” IEEE Trans-
nication in a fading environment when using multi-element antennas,” actions on Information Theory, vol. 31, no. 5, pp. 607–615, Sep. 1985.
Bell Labs Technical Journal, vol. 1, no. 2, pp. 41–59, 1996. [201] M. H. M. Costa and A. El Gamal, “The capacity region of the
[180] P. W. Wolniansky, G. J. Foschini, G. D. Golden, and R. A. Valenzuela, discrete memoryless interference channel with strong interference,”
“V-BLAST: an architecture for realizing very high data rates over IEEE Transactions on Information Theory, vol. 33, no. 5, pp. 710–
the rich-scattering wireless channel,” in Proc. URSI International 711, Sep. 1987.
Symposium on Signals, Systems, and Electronics (ISSSE’98), Pisa, Italy, [202] I. Sason, “On achievable rate regions for the gaussian interference
Sep. 1998, pp. 295–300. channel,” IEEE Transactions on Information Theory, vol. 50, no. 6,
[181] G. D. Golden, G. J. Foschini, R. A. Valenzuela, and P. W. Wolniansky, pp. 1345–1356, Jun. 2004.
“Detection algorithm and initial laboratory results using V-BLAST [203] G. Kramer, “Review of rate regions for interference channels,” in Proc.
space-time communication architecture,” Electronics Letters, vol. 35, International Zurich Seminar on Communications, Zurich, Switzerland,
no. 1, pp. 14–16, Jan. 1999. Feb. 2006, pp. 162–165.
[182] S. Sfar, R. D. Murch, and K. B. Letaief, “Layered space-time multiuser [204] S. A. Jafar and M. J. Fakhereddin, “Degrees of freedom for the
detection over wireless uplink systems,” IEEE Transactions on Wireless MIMO interference channel,” IEEE Transactions on Information The-
Communications, vol. 2, no. 4, pp. 653–668, Jul. 2003. ory, vol. 53, no. 7, pp. 2637–2642, Jul. 2007.
[183] S. Serbetli and A. Yener, “Transceiver optimization for multiuser [205] R. H. Etkin, D. N. C. Tse, and H. Wang, “Gaussian interference channel
MIMO systems,” IEEE Transactions on Signal Processing, vol. 52, capacity to within one bit,” IEEE Transactions on Information Theory,
no. 1, pp. 214–226, Jan. 2004. vol. 54, no. 12, pp. 5534–5562, Dec. 2008.
[184] F. Zhao, J. Liu, J. Liu, L. Guibas, and J. Reich, “Collaborative [206] S. Vishwanath, N. Jindal, and A. Goldsmith, “The ”Z” channel,” in
signal and information processing: an information-directed approach,” Proc. IEEE Global Telecommunications Conference (GLOBECOM’03),
Proceedings of the IEEE, vol. 91, no. 8, pp. 1199–1209, Aug. 2003. San Francisco, CA, USA, Dec. 2003, pp. 1726–1730.
[185] T. E. Hunter and A. Nosratinia, “Cooperation diversity through cod- [207] M. A. Maddah-Ali, A. S. Motahari, and A. K. Khandani, “Communi-
ing,” in Proc. IEEE International Symposium on Information Theory cation over MIMO X channels: interference alignment, decomposition,
(ISIT’02), Lausanne, Switzerland, Jul. 2002, p. 220. and performance analysis,” IEEE Transactions on Information Theory,
[186] J. N. Laneman and G. W. Wornell, “Distributed space-time-coded vol. 54, no. 8, pp. 3457–3470, Aug. 2008.
protocols for exploiting cooperative diversity in wireless networks,” [208] S. A. Jafar and S. Shamai (Shitz), “Degrees of freedom region of the
IEEE Transactions on Information Theory, vol. 49, no. 10, pp. 2415– MIMO X channel,” IEEE Transactions on Information Theory, vol. 54,
2425, Oct. 2003. no. 1, pp. 151–170, Jan. 2008.
[187] J. N. Laneman, D. N. C. Tse, and G. W. Wornell, “Cooperative diversity [209] V. R. Cadambe and S. A. Jafar, “Interference alignment and the degrees
in wireless networks: efficient protocols and outage behavior,” IEEE of freedom of wireless X networks,” IEEE Transactions on Information
Transactions on Information Theory, vol. 50, no. 12, pp. 3062–3080, Theory, vol. 55, no. 9, pp. 3893–3908, Sep. 2009.
Dec. 2004. [210] A. Grant, S. V. Hanly, J. S. Evans, and R. Muller, “Distributed decoding
[188] A. Nosratinia, T. E. Hunter, and A. Hedayat, “Cooperative communica- for Wyner cellular systems,” in Proc. 5th Australian Communications
tion in wireless networks,” IEEE Communications Magazine, vol. 42, Theory Workshop (AusCTW’04), Newcastle, Australia, Feb. 2004, pp.
no. 10, pp. 74–80, Oct. 2004. 77–81.
[189] M. Janani, A. Hedayat, T. E. Hunter, and A. Nosratinia, “Coded [211] H. Zhang, H. Dai, and Q. Zhou, “Base station cooperation for multiuser
cooperation in wireless communications: space-time transmission and MIMO: joint transmission and BS selection,” in Proc. 38th Annual
iterative decoding,” IEEE Transactions on Signal Processing, vol. 52, Conference on information sciences and systems (CISS’04), Princeton,
no. 2, pp. 362–371, Feb. 2004. NJ, USA, Mar. 2004, pp. 17–19.
[190] T. E. Hunter and A. Nosratinia, “Diversity through coded cooperation,” [212] H. Zhang, N. B. Mehta, A. F. Molisch, J. Zhang, and H. Dai, “Asyn-
IEEE Transactions on Wireless Communications, vol. 5, no. 2, pp. 283– chronous interference mitigation in cooperative base station systems,”
289, Feb. 2006. IEEE Transactions on Wireless Communications, vol. 7, no. 1, pp. 155–
[191] Q. H. Spencer, A. L. Swindlehurst, and M. Haardt, “Zero-forcing meth- 165, Jan. 2008.
ods for downlink spatial multiplexing in multiuser MIMO channels,” [213] Y. Hadisusanto, L. Thiele, and V. Jungnickel, “Distributed base sta-
IEEE Transactions on Signal Processing, vol. 52, no. 2, pp. 461–471, tion cooperation via block-diagonalization and dual-decomposition,” in
Feb. 2004. Proc. IEEE Global Telecommunications Conference (GLOBECOM’08),
[192] Q. H. Spencer, C. B. Peel, A. L. Swindlehurst, and M. Haardt, “An in- New Orleans, LO, Nov. 2008, pp. 1–5.
troduction to the multi-user MIMO downlink,” IEEE Communications [214] T. Mayer, H. Jenkac, and J. Hagenauer, “Turbo base-station cooperation
Magazine, vol. 42, no. 10, pp. 60–67, Oct. 2004. for intercell interference cancellation,” in Proc. IEEE International
[193] L.-U. Choi and R. D. Murch, “A transmit preprocessing technique Conference on Communications (ICC’06), Istanbul, Turkey, Jun. 2006,
for multiuser MIMO systems using a decomposition approach,” IEEE pp. 4977–4982.
Transactions on Wireless Communications, vol. 3, no. 1, pp. 20–24, [215] S. Khattak, W. Rave, and G. Fettweis, “Distributed iterative
Jan. 2004. multiuser detection through base station cooperation,” EURASIP
[194] H. Zhang and H. Dai, “Cochannel interference mitigation and J. Wireless Commun. Netw., vol. 2008, Article ID 390489,
cooperative processing in downlink multicell multiuser MIMO 15 pages, doi:10.1155/2008/390489, 2008. [Online]. Available:
networks,” EURASIP J. Wireless Commun. Netw., vol. 2004, no. 2, pp. http://downloads.hindawi.com/journals/wcn/2008/390489.pdf
222–235, doi:10.1155/S1687147204406148, 2004. [Online]. Available: [216] E. Aktas, J. S. Evans, and S. V. Hanly, “Distributed decoding in
http://downloads.hindawi.com/journals/wcn/2004/202654.pdf a cellular multiple-access channel,” IEEE Transactions on Wireless
[195] G. Caire, S. A. Ramprashad, H. C. Papadopoulos, C. Pepin, and C.- Communications, vol. 7, no. 1, pp. 241–250, Jan. 2008.
E. W. Sundberg, “Multiuser MIMO downlink with limited inter-cell [217] B. L. Ng, J. S. Evans, S. V. Hanly, and D. Aktas, “Distributed downlink
cooperation: approximate interference alignment in time, frequency and beamforming with cooperative base stations,” IEEE Transactions on
space,” in Proc. 46th Annual Allerton Conference on Communication, Information Theory, vol. 54, no. 12, pp. 5491–5499, Dec. 2008.
Control, and Computing (Allerton’08), Urbana-Champaign, IL, USA, [218] S. Yang, T. Lv, and L. Hanzo, “Base station cooperation in MIMO-
Sep. 2008, pp. 730–737. aided multi-user multi-cell systems employing distributed probabilistic
[196] G. Caire, N. Jindal, M. Kobayashi, and N. Ravindran, “Multiuser data association based soft reception,” in Proc. IEEE International
MIMO achievable rates with downlink training and channel state Conference on Communications (ICC’11), Kyoto, Japan, Jun. 2011,
feedback,” IEEE Transactions on Information Theory, vol. 56, no. 6, pp. 1–5.
pp. 2845–2866, Jun. 2010. [219] R. Zakhour and S. V. Hanly, “Base station cooperation on the downlink:
[197] A. Carleial, “A case where interference does not reduce capacity,” IEEE large system analysis,” IEEE Transactions on Information Theory,
Transactions on Information Theory, vol. 21, no. 5, pp. 569–570, Sep. vol. 58, no. 4, pp. 2079–2106, Apr. 2012.
1975. [220] M. A. Maddah-Ali, A. S. Motahari, and A. K. Khandani, “Signaling
[198] T. Han and K. Kobayashi, “A new achievable rate region for the inter- over MIMO multi-base systems: combination of multi-access and
ference channel,” IEEE Transactions on Information Theory, vol. 27, broadcast schemes,” in Proc. IEEE International Symposium on In-
no. 1, pp. 49–60, Jan. 1981. formation Theory (ISIT’06), Seattle, WA, Jul. 2006, pp. 2104–2108.

[221] ——, “Communication over X channel: signalling and multiplexing [244] M. K. Varanasi and B. Aazhang, “Multistage detection in asynchronous
gain,” Department of Electrical and Computer Engineering, University code-division multiple-access communications,” IEEE Transactions on
of Waterloo, Waterloo, ON, Canada, Tech. Rep. UW-ECE-2006-12, Communications, vol. 38, no. 4, pp. 509–519, Apr. 1990.
Jul. 2006. [245] ——, “Near-optimum detection in synchronous code-division multiple-
[222] ——, “Communication over MIMO X channel: signalling and perfor- access systems,” IEEE Transactions on Communications, vol. 39, no. 5,
mance analysis,” Department of Electrical and Computer Engineering, pp. 725–736, May 1991.
University of Waterloo, Waterloo, ON, Canada, Tech. Rep. UW-ECE- [246] ——, “Optimally near-far resistant multiuser detection in differentially
2006-27, Dec. 2006. coherent synchronous channels,” IEEE Transactions on Information
[223] S. A. Jafar and S. Shamai (Shitz), “Degrees of freedom of the MIMO Theory, vol. 37, no. 4, pp. 1006–1018, Jul. 1991.
X channel,” in Proc. IEEE Global Telecommunications Conference [247] Y. C. Yoon, R. Kohno, and H. Imai, “A spread-spectrum multiaccess
(GLOBECOM’07), Washington, DC, USA, Nov. 2007, pp. 1632–1636. system with cochannel interference cancellation for multipath fading
[224] V. R. Cadambe and S. A. Jafar, “Interference alignment and degrees channels,” IEEE Journal on Selected Areas in Communications, vol. 11,
of freedom of the K-user interference channel,” IEEE Transactions on no. 7, pp. 1067–1075, Sep. 1993.
Information Theory, vol. 54, no. 8, pp. 3425–3441, Aug. 2008. [248] D. Divsalar, M. K. Simon, and D. Raphaeli, “Improved parallel interfer-
[225] S. A. Jafar, “Interference alignment – a new look at signal dimensions ence cancellation for CDMA,” IEEE Transactions on Communications,
in a communication network,” Foundations and Trends in Communi- vol. 46, no. 2, pp. 258–268, Feb. 1998.
cations and Information Theory, vol. 7, no. 1, 2010. [249] R. M. Buehrer and B. D. Woerner, “Analysis of adaptive multistage
[226] D. A. Shnidman, “A generalized Nyquist criterion and an optimum interference cancellation for CDMA using an improved gaussian ap-
linear receiver for a pulse modulation system,” The Bell System proximation,” IEEE Transactions on Communications, vol. 44, no. 10,
Technical Journal, vol. 46, no. 9, pp. 2163–2177, Nov. 1967. pp. 1308–1321, Oct. 1996.
[227] A. Kaye and D. George, “Transmission of multiplexed PAM signals [250] T. Masamura, “Spread spectrum multiple access system with intrasys-
over multiple channel and diversity systems,” IEEE Transactions on tem interference cancellation,” IEICE Transactions, vol. E71, no. 3, pp.
Communication Technology, vol. COM-18, no. 5, pp. 520–526, Oct. 224–231, Mar. 1988.
1970. [251] A. J. Viterbi, “Very low rate convolution codes for maximum theoret-
[228] W. van Etten, “An optimum linear receiver for multiple channel digital ical performance of spread-spectrum multiple-access channels,” IEEE
transmission systems,” IEEE Transactions on Communications, vol. 23, Journal on Selected Areas in Communications, vol. 8, no. 4, pp. 641–
no. 8, pp. 828–834, Aug. 1975. 649, May 1990.
[229] ——, “Maximum likelihood receiver for multiple channel transmission [252] Z. Xie, C. K. Rushforth, and R. T. Short, “Multiuser signal detection
systems,” IEEE Transactions on Communications, vol. 24, no. 2, pp. using sequential decoding,” IEEE Transactions on Communications,
276–283, Feb. 1976. vol. 38, no. 5, pp. 578–583, May 1990.
[230] D. Horwood and R. Gagliardi, “Signal design for digital multiple access [253] Z. Xie, R. T. Short, and C. K. Rushforth, “A family of suboptimum
communications,” IEEE Transactions on Communications, vol. 23, detectors for coherent multiuser communications,” IEEE Journal on
no. 3, pp. 378–383, Mar. 1975. Selected Areas in Communications, vol. 8, no. 4, pp. 683–690, May
1990.
[231] K. S. Schneider, “Optimum detection of code division multiplexed
[254] Z. Xie, C. K. Rushforth, R. T. Short, and T. K. Moon, “Joint signal
signals,” IEEE Transactions on Aerospace and Electronic Systems, vol.
detection and parameter estimation in multiuser communications,”
AES-15, no. 1, pp. 181–185, Jan. 1979.
IEEE Transactions on Communications, vol. 41, no. 8, pp. 1208–1216,
[232] ——, “Crosstalk resistant receiver for M -ary multiplexed communica-
Aug. 1993.
tions,” IEEE Transactions on Aerospace and Electronic Systems, vol.
[255] A. Duel-Hallen, “Decorrelating decision-feedback multiuser detector
AES-16, no. 4, pp. 426–433, Jul. 1980.
for synchronous code-division multiple-access channel,” IEEE Trans-
[233] U. Timor, “Improved decoding scheme for frequency-hopped multilevel actions on Communications, vol. 41, no. 2, pp. 285–290, Feb. 1993.
FSK system,” The Bell System Technical Journal, vol. 59, no. 10, pp.
[256] ——, “A family of multiuser decision-feedback detectors for asyn-
1839–1855, Dec. 1980.
chronous code-division multiple-access channels,” IEEE Transactions
[234] ——, “Multistage decoding of frequency-hopped FSK system,” The on Communications, vol. 43, no. 2/3/4, pp. 421–434, Feb./Mar./Apr.
Bell System Technical Journal, vol. 60, no. 4, pp. 471–483, Apr. 1981. 1995.
[235] S. Verdú, “Optimum sequence detection of asynchronous multiple- [257] P. R. Patel and J. M. Holtzman, “Analysis of a simple successive inter-
access communications,” in Abstr. IEEE International Symposium on ference cancellation scheme in a DS/CDMA system,” IEEE Journal on
Information Theory (ISIT’83), St. Jovite, Canada, Sep. 1983, p. 80. Selected Areas in Communications, vol. 12, no. 5, pp. 796–807, Jun.
[236] ——, “Minimum probability of error for asynchronous multiple access 1994.
communication systems,” in Proc. IEEE Military Communications [258] M. K. Varanasi, “Decision feedback multiuser detection: a systematic
Conference (MILCOM’83), Washington, DC, Nov. 1983, pp. 213–219. approach,” IEEE Transactions on Information Theory, vol. 45, no. 1,
[237] ——, “Minimum probability of error for asynchronous Gaussian pp. 219–240, Jan. 1999.
multiple-access channels,” IEEE Transactions on Information Theory, [259] A. L. C. Hui and K. B. Letaief, “Successive interference cancellation
vol. IT-32, no. 1, pp. 85–96, Jan. 1986. for multiuser asynchronous DS/CDMA detectors in multipath fading
[238] ——, “Optimum multiuser asymptotic efficiency,” IEEE Transactions links,” IEEE Transactions on Communications, vol. 46, no. 3, pp. 384–
on Communications, vol. 34, no. 9, pp. 890–897, Sep. 1986. 391, Mar. 1998.
[239] R. Lupas and S. Verdú, “Linear multiuser detectors for synchronous [260] S. Verdú and S. Shamai (Shitz), “Spectral efficiency of CDMA with
code-division multiple-access channels,” IEEE Transactions on Infor- random spreading,” IEEE Transactions on Information Theory, vol. 45,
mation Theory, vol. 35, no. 1, pp. 123–136, Jan. 1989. no. 2, pp. 622–640, Mar. 1999.
[240] ——, “Near-far resistance of multiuser detectors in asynchronous [261] X. Wang and H. V. Poor, Wireless Communication Systems: Advanced
channels,” IEEE Transactions on Communications, vol. 38, no. 4, pp. Techniques for Signal Reception, 1st ed. Upper Saddle River, NJ,
496–508, Apr. 1990. USA: Prentice Hall, 2009.
[241] R. Kohno and M. Hatori, “Cancellation techniques of co-channel [262] A. Paulraj and T. Kailath, “Increasing capacity in wireless broad-
interference in asynchronous spread spectrum multiple access systems,” cast systems using distributed transmission/directional reception,” U.S.
Electronics and Communications in Japan (Part I: Communications), Patent 5 345 599, Sep. 6, 1994.
vol. 66, no. 5, pp. 20–29, May 1983. [Online]. Available: [263] E. Telatar, “Capacity of multi-antenna gaussian channels,” Technical
http://dx.doi.org/10.1002/ecja.4400660504 Report #BL0112170-950615-07TM, AT & T Bell Laboratories, 1995.
[242] R. Kohno, H. Imai, M. Hatori, and S. Pasupathy, “Combinations of [264] ——, “Capacity of multi-antenna gaussian channels,” European
an adaptive array antenna and a canceller of interference for direct- Transactions on Telecommunications, vol. 10, no. 6, pp. 585–595,
sequence spread-spectrum multiple-access system,” IEEE Journal on November-December 1999. [Online]. Available: http://dx.doi.org/10.
Selected Areas in Communications, vol. 8, no. 4, pp. 675–682, May 1002/ett.4460100604
1990. [265] G. J. Foschini and M. J. Gans, “On limits of wireless communications
[243] ——, “An adaptive canceller of cochannel interference for spread- in a fading environment when using multiple antennas,” Wireless
spectrum multiple-access communication networks in a power line,” Personal Communications, vol. 6, no. 3, pp. 311–335, Mar. 1998.
IEEE Journal on Selected Areas in Communications, vol. 8, no. 4, pp. [266] V. Tarokh, N. Seshadri, and A. R. Calderbank, “Space-time codes for
691–699, May 1990. high data rate wireless communication: performance criterion and code

construction,” IEEE Transactions on Information Theory, vol. 44, no. 2, [291] W. T. Webb and R. Steele, “Variable rate QAM for mobile radio,” IEEE
pp. 744–765, Mar. 1998. Transactions on Communications, vol. 43, no. 7, pp. 2223–2230, Jul.
[267] V. Tarokh, H. Jafarkhani, and A. R. Calderbank, “Space-time block 1995.
codes from orthogonal designs,” IEEE Transactions on Information [292] X. Dong, T. T. Tjhung, and F. Adachi, “Error probability analysis
Theory, vol. 45, no. 5, pp. 1456–1467, Jul. 1999. for 16 STAR-QAM in frequency-selective Rician fading with diversity
[268] A. F. Naguib, V. Tarokh, N. Seshadri, and A. R. Calderbank, “A space- reception,” IEEE Transactions on Vehicular Technology, vol. 47, no. 3,
time coding modem for high-data-rate wireless communications,” IEEE pp. 924–935, Aug. 1998.
Journal on Selected Areas in Communications, vol. 16, no. 8, pp. 1459– [293] X. Dong, N. C. Beaulieu, and P. H. Wittke, “Error probabilities of
1478, Oct. 1998. two-dimensional M-ary signaling in fading,” IEEE Transactions on
[269] V. Tarokh, H. Jafarkhani, and A. R. Calderbank, “Space-time block Communications, vol. 47, no. 3, pp. 352–355, Mar. 1999.
coding for wireless communications: performance results,” IEEE Jour- [294] J. Forney, G. D. and G. Ungerboeck, “Modulation and coding for linear
nal on Selected Areas in Communications, vol. 17, no. 3, pp. 451–460, Gaussian channels,” IEEE Transactions on Information Theory, vol. 44,
Mar. 1999. no. 6, pp. 2384–2415, Oct. 1998.
[270] V. Tarokh, A. F. Naguib, N. Seshadri, and A. R. Calderbank, “Space- [295] I. Nevat, G. W. Peters, and J. Yuan, “Detection of Gaussian constella-
time codes for high data rate wireless communication: performance tions in MIMO systems under imperfect CSI,” IEEE Transactions on
criteria in the presence of channel estimation errors, mobility, and Communications, vol. 58, no. 4, pp. 1151–1160, Apr. 2010.
multiple paths,” IEEE Transactions on Communications, vol. 47, no. 2, [296] D. E. Knuth, Seminumerical Algorithms, 3rd ed., ser. The Art of
pp. 199–207, Feb. 1999. Computer Programming. Reading, Massachusetts: Addison-Wesley,
[271] H. Jafarkhani, “A quasi-orthogonal space-time block code,” IEEE 1997.
Transactions on Communications, vol. 49, no. 1, pp. 1–4, Jan. 2001. [297] R. F. H. Fischer and C. Windpassinger, “Real versus complex-valued
[272] H. Jafarkhani and N. Seshadri, “Super-orthogonal space-time trellis equalisation in V-BLAST systems,” Electronics Letters, vol. 39, no. 5,
codes,” IEEE Transactions on Information Theory, vol. 49, no. 4, pp. pp. 470–471, Mar. 2003.
937–950, Apr. 2003. [298] M. Siti and M. P. Fitz, “A novel soft-output layered orthogonal lattice
[273] S. Alamouti, “A simple transmit diversity technique for wireless detector for multiple antenna communications,” in Proc. IEEE Inter-
communications,” IEEE Journal on Selected Areas in Communications, national Conference on Communications (ICC’06), vol. 4, Istanbul,
vol. 16, no. 8, pp. 1451–1458, Oct. 1998. Turkey, Jun. 2006, pp. 1686–1691.
[274] H. Jafarkhani, Space-Time Coding: Theory and Practice. Cambridge [299] L. Azzam and E. Ayanoglu, “Reduced complexity sphere decoding via
University Press, 2005. a reordered lattice representation,” IEEE Transactions on Communica-
[275] D. Divsalar and M. K. Simon, “Multiple-symbol differential detection tions, vol. 57, no. 9, pp. 2564–2569, Sep. 2009.
of MPSK,” IEEE Transactions on Communications, vol. 38, no. 3, pp. [300] T.-H. Liu and C.-N. Chiu, “On fast preprocessing schemes for
300–308, Mar. 1990. the real-valued spatially multiplexed MIMO detectors,” International
[276] V. Tarokh and H. Jafarkhani, “A differential detection scheme for trans- Journal of Communication Systems, 2012. [Online]. Available:
mit diversity,” IEEE Journal on Selected Areas in Communications, http://dx.doi.org/10.1002/dac.2365
vol. 18, no. 7, pp. 1169–1174, Jul. 2000. [301] F. D. Neeser and J. L. Massey, “Proper complex random processes with
[277] H. Jafarkhani and V. Tarokh, “Multiple transmit antenna differential applications to information theory,” IEEE Transactions on Information
detection from generalized orthogonal designs,” IEEE Transactions on Theory, vol. 39, no. 4, pp. 1293–1302, Jul. 1993.
Information Theory, vol. 47, no. 6, pp. 2626–2631, Sep. 2001. [302] T. Adali, P. Schreier, and L. Scharf, “Complex-valued signal process-
[278] B. L. Hughes, “Differential space-time modulation,” IEEE Transactions ing: The proper way to deal with impropriety,” IEEE Transactions on
on Information Theory, vol. 46, no. 7, pp. 2567–2578, Nov. 2000. Signal Processing, vol. 59, no. 11, pp. 5101–5125, Nov. 2011.
[279] D. Warrier and U. Madhow, “Spectrally efficient noncoherent commu- [303] D. Mandic and S. L. Goh, Complex Valued Nonlinear Adaptive Filters:
nication,” IEEE Transactions on Information Theory, vol. 48, no. 3, Noncircularity, Widely Linear and Neural Models, ser. Adaptive and
pp. 651–668, Mar. 2002. Learning Systems for Signal Processing, Communications and Control
[280] R. Schober and L. Lampe, “Noncoherent receivers for differen- Series. John Wiley & Sons, 2009.
tial space-time modulation,” IEEE Transactions on Communications, [304] D. W. Tufts, “Nyquist’s problem – the joint optimization of transmitter
vol. 50, no. 5, pp. 768–777, May 2002. and receiver in pulse amplitude modulation,” Proceedings of the IEEE,
[281] B. Hassibi and B. M. Hochwald, “How much training is needed in vol. 53, no. 3, pp. 248–259, Mar. 1965.
multiple-antenna wireless links?” IEEE Transactions on Information [305] Y. G. Li, J. H. Winters, and N. R. Sollenberger, “MIMO-OFDM
Theory, vol. 49, no. 4, pp. 951–963, Apr. 2003. for wireless communications: signal detection with enhanced channel
[282] L. Lampe, R. Schober, V. Pauli, and C. Windpassinger, “Multiple- estimation,” IEEE Transactions on Communications, vol. 50, no. 9, pp.
symbol differential sphere decoding,” IEEE Transactions on Commu- 1471–1477, Sep. 2002.
nications, vol. 53, no. 12, pp. 1981–1985, Dec. 2005. [306] B. Lu, G. Yue, and X. Wang, “Performance analysis and design opti-
[283] L. Hanzo, Y. Akhtman, L. Wang, and M. Jiang, MIMO-OFDM for LTE, mization of LDPC-coded MIMO OFDM systems,” IEEE Transactions
WiFi and WiMAX: Coherent versus Non-Coherent and Cooperative on Signal Processing, vol. 52, no. 2, pp. 348–361, Feb. 2004.
Turbo-Transceivers. Chichester, UK: Wiley, 2010. [307] A. Zanella, M. Chiani, and M. Z. Win, “MMSE reception and suc-
[284] G. L. Stuber, J. R. Barry, S. W. McLaughlin, Y. Li, M. A. Ingram, cessive interference cancellation for MIMO systems with high spectral
and T. G. Pratt, “Broadband MIMO-OFDM wireless communications,” efficiency,” IEEE Transactions on Wireless Communications, vol. 4,
Proceedings of the IEEE, vol. 92, no. 2, pp. 271–294, Feb. 2004. no. 3, pp. 1244–1253, May 2005.
[285] H. El Gamal, A. R. Hammons, Y. Liu, M. P. Fitz, and O. Y. [308] S. Chen, A. Livingstone, and L. Hanzo, “Minimum bit-error rate
Takeshita, “On the design of space-time and space-frequency codes design for space-time equalization-based multiuser detection,” IEEE
for MIMO frequency-selective fading channels,” IEEE Transactions Transactions on Communications, vol. 54, no. 5, pp. 824–832, May
on Information Theory, vol. 49, no. 9, pp. 2277–2292, Sep. 2003. 2006.
[286] T. Abe and T. Matsumoto, “Space-time turbo equalization in frequency- [309] D. P. Palomar, M. Bengtsson, and B. Ottersten, “Minimum BER linear
selective MIMO channels,” IEEE Transactions on Vehicular Technol- transceivers for MIMO channels via primal decomposition,” IEEE
ogy, vol. 52, no. 3, pp. 469–475, May 2003. Transactions on Signal Processing, vol. 53, no. 8, pp. 2866–2882, Aug.
[287] X. Zhu and R. D. Murch, “Layered space-frequency equalization in a 2005.
single-carrier MIMO system for frequency-selective channels,” IEEE [310] D. Wübben, R. Böhnke, V. Kühn, and K.-D. Kammeyer, “MMSE
Transactions on Wireless Communications, vol. 3, no. 3, pp. 701–708, extension of V-BLAST based on sorted QR decomposition,” in Proc.
May 2004. IEEE 58th Vehicular Technology Conference (VTC’03-Fall), Orlando,
[288] X. Ma, L. Yang, and G. B. Giannakis, “Optimal training for MIMO USA, Oct. 2003, pp. 508–512.
frequency-selective fading channels,” IEEE Transactions on Wireless [311] K.-W. Wong, C.-Y. Tsui, and R. S. Cheng, “A low complexity architec-
Communications, vol. 4, no. 2, pp. 453–466, Mar. 2005. ture of the V-BLAST system,” in Proc. IEEE Wireless Communications
[289] Z. Wang and G. B. Giannakis, “Wireless multicarrier communications– and Networking Confernce (WCNC’00), Chicago, IL, USA, Sep. 2000,
Where Fourier meets Shannon,” IEEE Signal Processing Magazine, pp. 310–314.
vol. 17, no. 3, pp. 29–48, May 2000. [312] R. Böhnke, D. Wübben, V. Kühn, and K.-D. Kammeyer, “Reduced
[290] W. T. Webb, L. Hanzo, and R. Steele, “Bandwidth efficient QAM complexity MMSE detection for BLAST architectures,” in Proc. IEEE
schemes for Rayleigh fading channels,” IEE Proceedings I (Communi- Global Telecommunications Conference (GLOBECOM’03), San Fran-
cations, Speech and Vision), vol. 138, no. 3, pp. 169–175, Jun. 1991. cisco, USA, Dec. 2003, pp. 2258–2262.

[313] B. Hassibi, “An efficient square-root algorithm for BLAST,” in Proc. [334] T. Datta, N. A. Kumar, A. Chockalingam, and B. S. Rajan, “A novel
IEEE International Conference on Acoustics, Speech, and Signal monte-carlo-sampling-based receiver for large-scale uplink multiuser
Processing (ICASSP’00), Istanbul, Turkey, Jun. 2000, pp. 737–740. MIMO systems,” IEEE Transactions on Vehicular Technology, vol. 62,
[314] J. Benesty, Y. Huang, and J. Chen, “A fast recursive algorithm for no. 7, pp. 3019–3038, Sep. 2013.
optimum sequential signal detection in a BLAST system,” IEEE [335] K. A. Singhal, T. Datta, and A. Chockalingam, “Lattice reduction aided
Transactions on Signal Processing, vol. 51, no. 7, pp. 1722–1730, Jul. detection in large-MIMO systems,” in Proc. IEEE 14th Workshop on
2003. Signal Processing Advances in Wireless Communications (SPAWC’13),
[315] T.-H. Liu, “Some results for the fast MMSE-SIC detection in spatially Darmstadt, Germany, Jun. 2013, pp. 594–598.
multiplexed MIMO systems,” IEEE Transactions on Wireless Commu- [336] P. Suthisopapan, K. Kasai, A. Meesomboon, and V. Imtawil, “Achieving
nications, vol. 8, no. 11, pp. 5443–5448, Nov. 2009. near capacity of non-binary LDPC coded large MIMO systems with a
[316] W. H. Chin, A. G. Constantinides, and D. B. Ward, “Parallel multistage novel ultra low-complexity soft-output detector,” IEEE Transactions on
detection for multiple antenna wireless systems,” Electronics Letters, Wireless Communications, vol. 12, no. 10, pp. 5185–5199, Oct. 2013.
vol. 38, no. 12, pp. 597–599, 2002. [337] T. Lakshmi Narasimhan and A. Chockalingam, “Channel hardening-
[317] Z. Luo, M. Zhao, S. Liu, and Y. Liu, “Generalized parallel interference exploiting message passing (CHEMP) receiver in large-scale MIMO
cancellation with near-optimal detection performance,” IEEE Transac- systems,” IEEE Journal of Selected Topics in Signal Processing, vol. 8,
tions on Signal Processing, vol. 56, no. 1, pp. 304–312, Jan. 2008. no. 5, pp. 847 – 860, Oct. 2014.
[318] C. Studer, S. Fateh, and D. Seethaler, “ASIC implementation of soft- [338] M. Cirkic and E. Larsson, “SUMIS: Near-optimal soft-in soft-out
input soft-output MIMO detection using MMSE parallel interference MIMO detection with low and fixed complexity,” IEEE Transactions
cancellation,” IEEE Journal of Solid-State Circuits, vol. 46, no. 7, pp. on Signal Processing, vol. 62, no. 12, pp. 3084 – 3097, Jun. 2014.
1754–1765, Jul. 2011.
[339] S. Wu, L. Kuang, Z. Ni, J. Lu, D. Huang, and Q. Guo, “Low-complexity
[319] E. Viterbo and E. Biglieri, “A universal decoding algorithm for lattice
iterative detection for large-scale multiuser MIMO-OFDM systems
codes,” in Proc. GRETSI 14-ème Colloque, Juan-les-Pins, France, Sep.
using approximate message passing,” IEEE Journal of Selected Topics
1993.
in Signal Processing, vol. 8, no. 5, pp. 902 – 915, Oct. 2014.
[320] Y. H. Wu, Y. T. Liu, H.-C. Chang, Y.-C. Liao, and H.-C. Chang, “Early-
[340] B. Yin, M. Wu, G. Wang, C. Dick, J. R. Cavallaro, and C. Studer,
pruned K-best sphere decoding algorithm based on radius constraints,”
“A 3.8 Gb/s large-scale MIMO detector for 3GPP LTE-Advanced,” in
in Proc. IEEE International Conference on Communications (ICC’08),
Proc. IEEE International Conference on Acoustics, Speech, and Signal
Beijing, China, May 2008, pp. 4496–4500.
Processing (ICASSP’14), Florence, Italy, May 2014, pp. 3879 – 3883.
[321] T. Fukatani, R. Matsumoto, and T. Uyematsu, “Two methods for
decreasing the computational complexity of the MIMO ML decoder,” [341] M. Wu, B. Yin, G. Wang, C. Dick, J. Cavallaro, and C. Studer,
IEICE Transactions on Fundamentals of Electronics, Communications “Large-scale MIMO detection for 3GPP LTE: Algorithm and FPGA
and Computer Sciences, vol. E87-A, no. 10, pp. 2571–2576, Oct. 2004. implementation,” IEEE Journal of Selected Topics in Signal Processing,
[322] A. Okawado, R. Matsumoto, and T. Uyematsu, “Near ML detection vol. 8, no. 5, pp. 916 – 929, Oct. 2014.
using Dijkstra’s algorithm with bounded list size over MIMO chan- [342] S. Sugiura, S. Chen, and L. Hanzo, “MIMO-aided near-capacity turbo
nels,” in Proc. IEEE International Symposium on Information Theory transceivers: Taxonomy and performance versus complexity,” IEEE
(ISIT’2008), Toronto, ON, Jul. 2008, pp. 2022–2025. Communications Surveys and Tutorials, vol. 14, no. 2, pp. 421–442,
[323] Z.-Q. Luo, W.-K. Ma, A. M.-C. So, Y. Ye, and S. Zhang, “Semidefinite Second Quarter 2012.
relaxation of quadratic optimization problems,” IEEE Signal Processing [343] J. G. Andrews, “Interference cancellation for cellular systems: a
Magazine, vol. 27, no. 3, pp. 20–34, May 2010. contemporary overview,” IEEE Wireless Communications Magazine,
[324] Z.-Q. Luo and W. Yu, “An introduction to convex optimization for vol. 12, no. 2, p. 2005, Apr. 19-29.
communications and signal processing,” IEEE Journal on Selected [344] A. J. Viterbi, “Error bounds for convolutional codes and an asymptoti-
Areas in Communications, vol. 24, no. 8, pp. 1426–1438, Aug. 2006. cally optimum decoding algorithm,” IEEE Transactions on Information
[325] Y.-C. Liang, S. Sun, and C. K. Ho, “Block-iterative generalized deci- Theory, vol. 13, no. 2, pp. 260–269, Apr. 1967.
sion feedback equalizers for large MIMO systems: algorithm design [345] J. Omura, “On the Viterbi decoding algorithm,” IEEE Transactions on
and asymptotic performance analysis,” IEEE Transactions on Signal Information Theory, vol. 15, no. 1, pp. 177–179, Jan. 1969.
Processing, vol. 54, no. 6, pp. 2035–2048, Jun. 2006. [346] G. D. Forney, “Maximum-likelihood sequence estimation of digital
[326] Y.-C. Liang, G. Pan, and Z. D. Bai, “Asymptotic performance of sequences in the presence of intersymbol interference,” IEEE Transac-
MMSE receivers for large systems using random matrix theory,” IEEE tions on Information Theory, vol. 18, no. 3, pp. 363–378, May 1972.
Transactions on Information Theory, vol. 53, no. 11, pp. 4173–4190, [347] ——, “The Viterbi algorithm,” Proceedings of the IEEE, vol. 61, no. 3,
Nov. 2007. pp. 268–278, Mar. 1973.
[327] Y.-C. Liang, E. Y. Cheu, L. Bai, and G. Pan, “On the relationship [348] A. J. Viterbi, “A personal history of the Viterbi algorithm,” IEEE Signal
between MMSE-SIC and BI-GDFE receivers for large multiple-input Processing Magazine, vol. 23, no. 4, pp. 120–142, Jul. 2006.
multiple-output channels,” IEEE Transactions on Signal Processing, [349] S. U. H. Qureshi, “Adaptive equalization,” Proceedings of the IEEE,
vol. 56, no. 8, pp. 3627–3637, Aug. 2008. vol. 73, no. 9, pp. 1349–1387, Sep. 1985.
[328] B. S. Rajan, S. K. Mohammed, A. Chockalingam, and N. Srinidhi,
[350] M. Honig, U. Madhow, and S. Verdú, “Blind adaptive multiuser
“Low-complexity near-ML decoding of large non-orthogonal STBCs
detection,” IEEE Transactions on Information Theory, vol. 41, no. 4,
using reactive tabu search,” in Proc. IEEE International Symposium on
pp. 944–960, Jul. 1995.
Information Theory (ISIT’09, Seoul, Korea, Jun. 2009, pp. 1993–1997.
[351] L. Tong, G. Xu, and T. Kailath, “Blind identification and equalization
[329] S. K. Mohammed, A. Zaki, A. Chockalingam, and B. S. Rajan, “High-
based on second-order statistics: a time domain approach,” IEEE
rate space-time coded large-MIMO systems: Low-complexity detection
Transactions on Information Theory, vol. 40, no. 2, pp. 340–349, Mar.
and channel estimation,” IEEE Journal of Selected Topics in Signal
1994.
Processing, vol. 3, no. 6, pp. 958–974, Dec 2009.
[330] A. Chockalingam, “Low-complexity algorithms for large-MIMO de- [352] X. Wang and H. V. Poor, “Blind equalization and multiuser detection
tection,” in Proc. 4th International Symposium on Communications, in dispersive cdma channels,” IEEE Transactions on Communications,
Control and Signal Processing (ISCCSP’10), Limassol, Cyprus, Mar. vol. 46, no. 1, pp. 91–103, Jan. 1998.
2010, pp. 1–6. [353] W. H. Tranter, D. P. Taylor, and R. E. Ziemer, Eds., The best of the best:
[331] T. Datta, N. Srinidhi, A. Chockalingam, and B. S. Rajan, “A hybrid fifty years of communications and networking research. Hoboken, NJ:
RTS-BP algorithm for improved detection of large-MIMO M-QAM John Wiley & Sons, 2007.
signals,” in Proc. National Conference on Communications (NCC’11), [354] S. Verdú, “Optimum multi-user signal detection,” Ph.D. dissertation,
Bangalore, India, Jan. 2011, pp. 1–5. Department of Electrical and Computer Engineering, University of
[332] P. Som, T. Datta, N. Srinidhi, A. Chockalingam, and B. S. Rajan, Illinois at Urbana-Champaign, Urbana, IL, Aug. 1984.
“Low-complexity detection in large-dimension MIMO-ISI channels [355] H. V. Poor and S. Verdú, “Single-user detectors for multiuser channels,”
using graphical models,” IEEE Journal of Selected Topics in Signal IEEE Transactions on Communications, vol. 36, no. 1, pp. 50–60, Jan.
Processing, vol. 5, no. 8, pp. 1497–1511, Dec. 2011. 1988.
[333] A. Kumar, S. Chandrasekaran, A. Chockalingam, and B. S. Rajan, [356] S. J. Grant and J. K. Cavers, “Performance enhancement through
“Near-optimal large-MIMO detection using randomized MCMC and joint detection of cochannel signals using diversity arrays,” IEEE
randomized search algorithms,” in Proc. IEEE International Confer- Transactions on Communications, vol. 46, no. 8, pp. 1038–1049, Aug.
ence on Communications (ICC’11), Jun. 2011, pp. 1–5. 1998.

[357] ——, “Further analytical results on the joint detection of cochannel [379] L. Zheng and D. N. C. Tse, “Diversity and multiplexing: A funda-
signals using diversity arrays,” IEEE Transactions on Communications, mental tradeoff in multiple-antenna channels,” IEEE Transactions on
vol. 48, no. 11, pp. 1788–1792, Nov. 2000. Information Theory, vol. 49, no. 5, pp. 1073–1096, May 2003.
[358] R. van Nee, V. van Zelst, and G. Awater, “Maximum likelihood [380] D. N. C. Tse, P. Viswanath, and L. Zheng, “Diversity-multiplexing
decoding in a space division multiplexing system,” in Proc. IEEE 51st tradeoff in multiple-access channels,” IEEE Transactions on Informa-
Vehicular Technology Conference (VTC’00-Spring), Tokyo, Japan, May tion Theory, vol. 50, no. 9, pp. 1859–1874, Sep. 2004.
2000, pp. 6–10. [381] A. Hedayat and A. Nosratinia, “Outage and diversity of linear receivers
[359] X. Zhu and R. D. Murch, “Performance analysis of maximum like- in flat-fading MIMO channels,” IEEE Transactions on Signal Process-
lihood detection in a MIMO antenna system,” IEEE Transactions on ing, vol. 55, no. 12, pp. 5868–5873, Dec. 2007.
Communications, vol. 50, no. 2, pp. 187–191, Feb. 2002. [382] E. A. Jorswieck and H. Boche, “Outage probability in multiple antenna
[360] M. Shin, D. S. Kwon, and C. Lee, “Performance analysis of maximum systems,” European Transactions on Telecommunications, vol. 18,
likelihood detection for MIMO systems,” in Proc. IEEE 63rd Vehicular no. 3, pp. 217–233, 2007.
Technology Conference (VTC’06-Spring), Melbourne, Austria, May [383] A. L. Moustakas, K. R. Kumar, and G. Caire, “Performance of MMSE
2006, pp. 2154–2158. MIMO receivers: A large N analysis for correlated channels,” in
[361] S. Verdú, “Demodulation in the presence of multiuser interference: Proc. IEEE 69th Vehicular Technology Conference (VTC’09 Spring),
progress and misconceptions,” in Intelligent Methods in Signal Barcelona, Spain, Apr. 2009, pp. 1–5.
Processing and Communications, D. Docampo, A. R. Figueiras-Vidal, [384] K. R. Kumar, G. Caire, and A. L. Moustakas, “Asymptotic performance
and F. Pérez-González, Eds. Boston: Birkhäuser, 1997, pp. 15–45. of linear receivers in MIMO fading channels,” IEEE Transactions on
[Online]. Available: http://dx.doi.org/10.1007/978-1-4612-2018-3 2 Information Theory, vol. 55, no. 10, pp. 4398–4418, Oct. 2009.
[362] P. Li, D. Paul, R. Narasimhan, and J. Cioffi, “On the distribution of [385] Y. Jiang, M. Varanasi, and J. Li, “Performance analysis of ZF and
SINR for the MMSE MIMO receiver and performance analysis,” IEEE MMSE equalizers for MIMO systems: An in-depth study of the high
Transactions on Information Theory, vol. 52, no. 1, pp. 271–286, Jan. SNR regime,” IEEE Transactions on Information Theory, vol. 57, no. 4,
2006. pp. 2008–2026, Apr. 2011.
[363] A. H. Mehana and A. Nosratinia, “Diversity of MMSE MIMO re- [386] N. B. Mandayam and B. Aazhang, “Generalized sensitivity analysis of
ceivers,” IEEE Transactions on Information Theory, vol. 58, no. 11, optical code division multiple access systems,” in Proc. 27th Annual
pp. 6788–6805, Nov. 2012. Conference on Information Sciences and Systems (CISS’93), Baltimore,
[364] K. Abend, T. J. Harley, B. D. Fritchman, and C. Gumacos, “On MD, Mar. 1993, pp. 302–307.
optimum receivers for channels having memory,” IEEE Transactions [387] ——, “Gradient estimation for stochastic optimization of optical code-
on Information Theory, vol. 14, no. 6, pp. 819–820, Nov. 1968. division multiple-access systems: Part I – generalized sensitivity anal-
[365] K. Abend and B. D. Fritchman, “Statistical detection for communica- ysis,” IEEE Journal on Selected Areas in Communications, vol. 15,
tion channels with intersymbol interference,” Proceedings of the IEEE, no. 4, pp. 731–741, May 1997.
vol. 58, no. 5, pp. 779–785, May 1970. [388] ——, “Gradient estimation for stochastic optimization of optical code-
[366] L. Bahl, J. Cocke, F. Jelinek, and J. Raviv, “Optimal decoding of division multiple-access systems: Part II – adaptive detection,” IEEE
linear codes for minimizing symbol error rate,” IEEE Transactions on Journal on Selected Areas in Communications, vol. 15, no. 4, pp. 742–
Information Theory, vol. 20, no. 2, pp. 284–287, Mar. 1974. 750, May 1997.
[367] S. Verdú and H. V. Poor, “Backward, forward and backward-forward [389] ——, “Gradient estimation for sensitivity analysis and adaptive mul-
dynamic programming models under commutativity conditions,” in tiuser interference rejection in code-division multiple-access systems,”
Proc. the 23rd IEEE Conference on Decision and Control (CDC’84), IEEE Transactions on Communications, vol. 45, no. 7, pp. 848–858,
Las Vegas, NV, Dec. 1984, pp. 1081–1086. Jul. 1997.
[368] ——, “Abstract dynamic programming models under commutativity [390] A. Burg, S. Haene, D. Perels, P. Luethi, N. Felber, and W. Fichtner,
conditions,” SIAM Journal on Control and Optimization, vol. 25, “Algorithm and VLSI architecture for linear MMSE detection in
no. 4, pp. 990–1006, Jul. 1987. [Online]. Available: http://epubs.siam. MIMO-OFDM systems,” in Proc. IEEE International Symposium on
org/doi/abs/10.1137/0325054 Circuits and Systems (ISCAS’06), Island of Kos, Greece, May 2006,
[369] M. R. Garey and D. S. Johnson, Computers and Intractability: A Guide pp. 4102–4105.
to the Theory of NP-Completeness. San Francisco, USA: W. H. [391] S. Yoshizawa and Y. Miyanaga, “VLSI implementation of a 4 x
Freeman and Co., 1979. 4 MIMO-OFDM transceiver with an 80-MHz channel bandwidth,”
[370] T. H. Cormen, C. E. Leiserson, R. L. Rivest, and C. Stein, Introduction in Proc. IEEE International Symposium on Circuits and Systems
to Algorithms, 3rd ed. Cambridge, Masachusetts: MIT Press, 2009. (ISCAS’09), Taipei, Republic of China, May 2009, pp. 1743–1746.
[371] D. C. Garrett, L. M. Davis, and G. K. Woodward, “19.2 Mbit/s 4 × 4 [392] I. N. Psaromiligkos, S. N. Batalama, and D. A. Pados, “On adaptive
BLAST/MIMO detector with soft ML outputs,” Electronics Letters, minimum probability of error linear filter receivers for DS-CDMA
vol. 39, no. 2, pp. 233–235, Jan 2003. channels,” IEEE Transactions on Communications, vol. 47, no. 7, pp.
[372] A. Burg, N. Felber, and W. Fichtner, “A 50 Mbps 4×4 maximum like- 1092–1102, Jul. 1999.
lihood decoder for multiple-input multiple-output systems with QPSK [393] X. Wang, W.-S. Lu, and A. Antoniou, “Constrained minimum-BER
modulation,” in Proc. IEEE International Conference on Electronics, multiuser detection,” in Proc. IEEE International Conference on Acous-
Circuits and Systems (ICECS’03), Sharjah, United Arab Emirates, Dec. tics, Speech, and Signal Processing (ICASSP’99), Phoenix, AZ, USA,
2003, pp. 332–335. Mar. 1999, pp. 2603–2606.
[373] J. Hoydis, S. ten Brink, and M. Debbah, “Massive mimo in the ul/dl of [394] ——, “Constrained minimum-BER multiuser detection,” IEEE Trans-
cellular networks: How many antennas do we need?” IEEE Journal on actions on Signal Processing, vol. 48, no. 10, pp. 2903–2909, Oct.
Selected Areas in Communications, vol. 31, no. 2, pp. 160–171, Feb. 2000.
2013. [395] C.-C. Yeh, R. R. Lopes, and J. R. Barry, “Approximate minimum bit-
[374] R. Lupas and S. Verdú, “Asymptotic efficiency of linear multiuser error rate multiuser detection,” in Proc. IEEE Global Telecommuni-
detectors,” in Proc. the 25th IEEE Conference on Decision and Control cations Conference (GLOBECOM’98), Sydney, NSW, Australia, Nov.
(CDC’86), Dec. 1986, pp. 2094–2100. 1998, pp. 3590–3595.
[375] ——, “Optimum near-far resistance of linear detectors for code- [396] C.-C. Yeh and J. R. Barry, “Adaptive minimum bit-error rate equal-
division multiple-access channels,” in Abstr. IEEE International ization for binary signaling,” IEEE Transactions on Communications,
Symposium on Information Theory (ISIT’88), Jun. 1988, vol. 48, no. 7, pp. 1226–1235, Jul. 2000.
p. 14. [Online]. Available: http://www.princeton.edu/∼verdu/reprints/ [397] S. Chen, A. K. Samingan, B. Mulgrew, and L. Hanzo, “Adaptive
OptimumNear-FarResistanceOfLinearDetectors.pdf minimum-BER linear multiuser detection for DS-CDMA signals in
[376] S. M. Kay, Fundamentals of Statistical Signal Processing, Volume I: multipath channels,” IEEE Transactions on Signal Processing, vol. 49,
Estimation Theory. Prentice Hall, 1993. no. 6, pp. 1240–1247, Jun. 2001.
[377] Z. Xie, R. T. Short, and C. K. Rushforth, “Suboptimum coherent [398] P. Bergmans and T. M. Cover, “Cooperative broadcasting,” IEEE
detection of direct-sequence multiple-access signals,” in Proc. IEEE Transactions on Information Theory, vol. 20, no. 3, pp. 317–324, May
Military Communications Conference (MILCOM’89), Boston, MA, 1974.
Oct. 1989, pp. 128–133. [399] T. M. Cover, “Some advances in broadcast channels,” in Advances
[378] H. V. Poor and S. Verdú, “Probability of error in MMSE multiuser in Communication Systems, A. J. Viterbi, Ed. New York: Academic
detection,” IEEE Transactions on Information Theory, vol. 43, no. 3, Press, Inc., 1975, vol. 4, pp. 229–260. [Online]. Available:
pp. 858–871, May 1997. http://www-isl.stanford.edu/∼cover/papers/paper34.pdf

[400] P. Dent, B. Gudmundson, and M. Ewerbring, “CDMA-IC: a novel code Control and Information Sciences, W. A. Porter and S. C. Kak, Eds.
division multiple access scheme based on interference cancellation,” Heidelberg: Springer-Verlag, 1989, vol. 129, pp. 27–38. [Online].
in Proc. 3rd IEEE International Symposium on Personal, Indoor and Available: http://dx.doi.org/10.1007/BFb0042716
Mobile Radio Communications (PIMRC’92), Boston, MA, Oct. 1992, [420] Z. Xie and C. K. Rushforth, “Multi-user signal detection using sequen-
pp. 98–102. tial decoding,” in Proc. IEEE Military Communications Conference
[401] S. Moshavi, E. G. Kanterakis, and D. L. Schilling, “Multistage linear (MILCOM’88), San Diego, CA, USA, Oct. 1988, pp. 983–988.
receivers for DS-CDMA systems,” International Journal of Wireless [421] Z. Xie, C. K. Rushforth, R. T. Short, and T. K. Moon, “A tree-search
Information Networks, vol. 3, no. 1, pp. 1–17, Jan. 1996. [Online]. algorithm for signal detection and parameter estimation in multi-user
Available: http://dx.doi.org/10.1007/BF02106658 communications,” in Proc. IEEE Military Communications Conference
[402] S. Kubota, S. Kato, and K. Feher, “Inter-channel interference cancel- (MILCOM’90), Monterey, CA, USA, Oct. 1990, pp. 796–800.
lation technique for CDMA mobile/personal communication systems,” [422] L. Wei, L. K. Rasmussen, and R. Wyrwas, “Near optimum tree-search
in Proc. 3rd IEEE International Symposium on Personal, Indoor and detection schemes for bit-synchronous multiuser CDMA systems over
Mobile Radio Communications (PIMRC’92), Boston, MA, Oct. 1992, gaussian and two-path rayleigh-fading channels,” IEEE Transactions
pp. 112–117. on Communications, vol. 45, no. 6, pp. 691–700, Jun. 1997.
[403] ——, “Inter-channel interference cancellation technique for CDMA [423] F. Jelinek and J. B. Anderson, Instrumentable tree encoding of infor-
mobile/personal communication base stations,” in Proc. 2nd IEEE mation sources. School of Electrical Engineering, Cornell University,
International Symposium on Spread Spectrum Techniques and Appli- Ithaca, NY, USA, Sep. 1969.
cations (ISSTA’92), Yokohama, Japan, Nov. 1992, pp. 91–94.
[424] ——, “Instrumentable tree encoding of information sources,” IEEE
[404] P. R. Patel and J. M. Holtzman, “Analysis of a DS/CDMA successive
Transactions on Information Theory, vol. 17, no. 1, pp. 118–119, Jan.
interference cancellation scheme using correlations,” in Proc. IEEE
1971.
Global Telecommunications Conference (GLOBECOM’93), Houston,
[425] W. L. Waltmann and R. J. Lambert, “T-algorithm for
TX, USA, Dec. 1993, pp. 76–80.
tridiagonalization,” Journal of the Society for Industrial and
[405] J. M. Holtzman, “DS/CDMA successive interference cancellation,”
Applied Mathematics, vol. 13, no. 4, pp. 1069–1078, Dec. 1965.
in Proc. IEEE 3rd International Symposium on Spread Spectrum
[Online]. Available: http://www.jstor.org/stable/2946426
Techniques and Applications (ISSSTA’94), Oulu, Finland, Jul. 1994,
pp. 69–78. [426] S. J. Simmons, “A reduced-computation trellis decoder with
[406] ——, “Successive interference cancellation for direct sequence code inherent parallelism,” Ph.D. dissertation, Department of Electrical
division multiple access,” in Proc. IEEE Military Communications and Computer Engineering, Queens University, Kingston, Ontario,
Conference (MILCOM’94), Fort Monmouth, NJ, Oct. 1994, pp. 997– Canada, Jun. 1986. [Online]. Available: http://istec.colostate.edu/
∼rockey/Papers/PhDThesis.pdf
1001.
[407] A. S. Gupta and A. C. Singer, “Successive interference cancellation [427] S. J. Simmons and P. Senyshyn, “Reduced-search trellis decoding of
using constellation structure,” IEEE Transactions on Signal Processing, coded modulations over ISI channels,” in Proc. IEEE Global Telecom-
vol. 55, no. 12, pp. 5716–5730, Dec. 2007. munications Conference (GLOBECOM’90), Dec. 1990, pp. 393–396.
[408] J. G. Andrews and T. H. Meng, “Optimum power control for successive [428] S. J. Simmons, “Breadth-first trellis decoding with adaptive effort,”
interference cancellation with imperfect channel estimation,” IEEE IEEE Transactions on Communications, vol. 38, no. 1, pp. 3–12, Jan.
Transactions on Wireless Communications, vol. 2, no. 2, pp. 375–383, 1990.
Mar. 2003. [429] M. V. Eyuboǧlu and S. U. Qureshi, “Reduced-state sequence estimation
[409] Y. C. Yoon, R. Kohno, and H. Imai, “A spread-spectrum multi-access with set partitioning and decision feedback,” IEEE Transactions on
system with a cascade of co-channel interference cancelers for multi- Communications, vol. 36, no. 1, pp. 13–20, Jan. 1988.
path fading channels,” in Proc. 2nd IEEE International Symposium on [430] R. Fano, “A heuristic discussion of probabilistic decoding,” IEEE
Spread Spectrum Techniques and Applications (ISSTA’92), Yokohama, Transactions on Information Theory, vol. 9, no. 2, pp. 64–74, Apr.
Japan, Nov. 1992, pp. 87–90. 1963.
[410] R. M. Buehrer, S. P. Nicoloso, and S. Gollamudi, “Linear versus [431] K. S. Zigangirov, “Some sequential decoding procedures,” Problemy
nonlinear interference cancellation,” Journal of Communication and Peredachi Informatsii, vol. 2, no. 4, pp. 13–25, Oct. 1966.
Networks, vol. 1, no. 2, pp. 118–133, Jun. 1999. [432] F. Jelinek, “Fast sequential decoding algorithm using a stack,” IBM
[411] D. Guo, L. K. Rasmussen, S. Sun, and T. J. Lim, “A matrix-algebraic Journal of Research and Development, vol. 13, no. 6, pp. 675–685,
approach to linear parallel interference cancellation in CDMA,” IEEE Nov. 1969.
Transactions on Communications, vol. 48, no. 1, pp. 152–161, Jan. [433] J. L. Massey, “Variable-length codes and the fano metric,” IEEE
2000. Transactions on Information Theory, vol. 18, no. 1, pp. 196–198, Jan.
[412] M. K. Varanasi and B. Aazhang, “An iterative detector for asyn- 1972.
chronous spread-spectrum multiple-access systems,” in Proc. IEEE [434] J. B. Anderson and S. Mohan, “Sequential coding algorithms: a survey
Global Telecommunications Conference (GLOBECOM’88), Nov. 1988, and cost analysis,” IEEE Transactions on Communications, vol. 32,
pp. 556–560. no. 2, pp. 169–176, Feb. 1984.
[413] ——, “Probability of error comparison of linear and iterative mul-
[435] J. B. Anderson, “Limited search trellis decoding of convolutional
tiuser detectors,” in Proc. International Conference on Advances in
codes,” IEEE Transactions on Information Theory, vol. 35, no. 5, pp.
Communications and Control Systems, Baton Rouge, LA, Oct. 1988,
944–955, Sep. 1989.
pp. 54–65.
[436] G. J. Pottie and D. P. Taylor, “A comparison of reduced complexity
[414] A. Duel-Hallen, “Linear and decision-feedback multiuser detectors,” in
decoding algorithms for trellis codes,” IEEE Journal on Selected Areas
Proc. IEEE International Symposium on Information Theory (ISIT’91),
in Communications, vol. 7, no. 9, pp. 1369–1380, Dec. 1989.
Budapest, Hungary, Jun. 1991, pp. 24–28.
[415] ——, “Performance of multiuser zero-forcing and MMSE decision- [437] M. Pohst, “On the computation of lattice vectors of minimal length,
feedback detectors for CDMA channels,” in Proc. IEEE Global successive minima and reduced bases with applications,” ACM SIGSAM
Telecommunications Conference (GLOBECOM’93), Houston, TX, Bulletin, vol. 15, no. 1, pp. 37–44, Feb. 1981.
USA, Nov. 1993, pp. 82–86. [438] U. Fincke and M. Pohst, “Improved methods for calculating vectors of
[416] Z. Guo and P. Nilsson, “A VLSI implementation of MIMO detection short length in a lattice, including a complexity analysis,” Mathematics
for future wireless communications,” in Proc. IEEE Proceedings on of Computation, vol. 44, no. 170, pp. 463–471, Apr. 1985.
Personal, Indoor and Mobile Radio Communications (PIMRC’03), [439] C. P. Schnorr and M. Euchner, “Lattice basis reduction: improved
Beijing, China, Sep. 2003, pp. 2852–2856. practical algorithms and solving subset sum problems,” Mathematical
[417] A. Duel-Hallen and C. Heegard, “Delayed decision-feedback sequence Programming, vol. 66, no. 1-3, pp. 181–199, Aug. 1994.
estimation,” IEEE Transactions on Communications, vol. 37, no. 5, pp. [440] C.-A. Shen, A. M. Eltawil, K. N. Salama, and S. Mondal, “A best-first
428–436, May 1989. soft/hard decision tree searching MIMO decoder for a 4 × 4 64-QAM
[418] A. Duel-Hallen, “Equalizers for multiple input/multiple output channels system,” IEEE Transactions on Very Large Scale Integration (VLSI)
and PAM systems with cyclostationary input sequences,” IEEE Journal Systems, vol. 20, no. 8, pp. 1537–1541, Aug. 2012.
on Selected Areas in Communications, vol. 10, no. 3, pp. 630–639, Apr. [441] B. Hassibi and H. Vikalo, “On the expected complexity of sphere
1992. decoding,” in Proc. 35th Annual Asilomar Conference on Signals,
[419] S. Verdú, “Recent progress in multiuser detection,” in Advances Systems and Computers (Asilomar’01), Pacific Grove, CA, USA, Nov.
in Communications and Signal Processing, ser. Lecture Notes in 2001, pp. 1051–1055.

[442] ——, “On the expected complexity of integer least-squares problems,” [463] I. V. L. Clarkson, “Approximation of linear forms by lattice points with
in Proc. IEEE IEEE International Conference on Acoustics, Speech, applications to signal processing,” Ph.D. dissertation, 1997.
and Signal Processing (ICASSP’02), vol. 2, Orlando, FL, USA, May [464] C. Ling, “Approximate lattice decoding: Primal versus dual basis
2002, pp. II–1497–II–1500. reduction,” in Proc. IEEE International Symposium on Information
[443] ——, “Maximum-likelihood decoding and integer least-squares: The Theory (ISIT’06), Jul. 2006, pp. 1–5.
expected complexity,” in DIMACS Series in Discrete Mathematics [465] H. Napias, “A generalization of the LLL-algorithm over euclidean rings
and Theoretical Computer Science: Multiantenna channels: capacity, or orders,” Journal de théorie des nombres de Bordeaux, vol. 8, pp.
coding and signal processing, G. J. Foschini and S. Verdú, Eds. New 387–396, 1996.
York: American Mathematical Society, 2003, vol. 62, pp. 161–192. [466] D. Wübben, R. Böhnke, V. Kühn, and K.-D. Kammeyer, “MMSE-based
[444] J. Jaldén and B. Ottersten, “An exponential lower bound on the lattice-reduction for near-ML detection of MIMO systems,” in Proc.
expected complexity of sphere decoding,” in Proc. IEEE International ITG Workshop on Smart Antennas (WSA’04), Mar. 2004, pp. 106–113.
Conference on Acoustics, Speech, and Signal Processing (ICASSP’04), [467] Y. Bar-Shalom and E. Tse, “Tracking in a cluttered environment with
vol. 4, Montreal, Canada, May 2004, pp. 393–396. probabilistic data association,” Automatica, vol. 11, no. 5, pp. 451 –
[445] B. M. Hochwald and S. ten Brink, “Iterative list sphere decoding to at- 460, 1975.
tain capacity on a multi-antenna link,” in Proc. of 39th Annual Allerton [468] T. E. Fortmann, Y. Bar-Shalom, and M. Scheffe, “Multi-target tracking
Conference On Communication, Control and Computing (Allerton’01), using joint probabilistic data association,” in 19th IEEE Conference on
Monticello, IL, USA, Oct. 2001, pp. 815–824. Decision and Control including the Symposium on Adaptive Processes,
[446] D. C. Garrett, L. M. Davis, S. ten Brink, B. Hochwald, and G. Knagge, Albuquerque, NM, USA, Dec. 1980, pp. 807–812.
“Silicon complexity for maximum likelihood MIMO detection using [469] ——, “Sonar tracking of multiple targets using joint probabilistic data
spherical decoding,” IEEE Journal of Solid-State Circuits, vol. 39, association,” IEEE Journal of Oceanic Engineering, vol. 8, no. 3, pp.
no. 9, pp. 1544–1552, Sep. 2004. 173–184, Jul. 1983.
[447] L. G. Barbero and J. S. Thompson, “A fixed-complexity MIMO [470] Y. Bar-Shalom and X. R. Li, Estimation and Tracking: Principles,
detector based on the complex sphere decoder,” in Proc. IEEE 7th Techniques and Software. Dedham, MA: Artech House, 1993.
Workshop on Signal Processing Advances in Wireless Communications [471] Y. Bar-Shalom, X. R. Li, and T. Kirubarajan, Estimation with Appli-
(SPAWC’06), Cannes, France, Jul. 2006, pp. 1–5. cations to Tracking and Navigation: Theory, Algorithms and Software.
[448] Z. Guo and P. Nilsson, “A VLSI architecture of the Schnorr-Euchner John Wiley & Sons, Inc., 2004.
decoder for MIMO systems,” in Proc. IEEE 6th Circuits and Systems [472] Y. Bar-Shalom, T. Kirubarajan, and X. Lin, “Probabilistic data asso-
Symposium on Emerging Technologies: Frontiers of Mobile and Wire- ciation techniques for target tracking with applications to sonar, radar
less Communication, Shanghai, China, May 2004, pp. 65–68. and EO sensors,” IEEE Aerospace and Electronics Systems Magazine,
[449] M. Wenk, M. Zellweger, A. Burg, N. Felber, and W. Fichtner, “K- vol. 20, no. 8, pp. 37–56, 2005.
best MIMO detection VLSI architectures achieving up to 424 Mbps,” [473] Y. Bar-Shalom, F. Daum, and J. Huang, “The probabilistic data
in Proc. IEEE International Symposium on Circuits and Systems association filter,” IEEE Control Systems Magazine, vol. 29, no. 6,
(ISCAS’06), May 2006, pp. 1151–1154. pp. 82–100, Dec. 2009.
[450] D. Patel, V. Smolyakov, M. Shabany, and P. G. Gulak, “VLSI [474] K.-C. Chang and Y. Bar-Shalom, “Joint probabilistic data association
implementation of a WiMAX/LTE compliant low-complexity high- for multitarget tracking with possibly unresolved measurements and
throughput soft-output K-best MIMO detector,” in Proc. IEEE Interna- maneuvers,” IEEE Transactions on Automatic Control, vol. 29, no. 7,
tional Symposium on Circuits and Systems (ISCAS’10), Paris, France, pp. 585–594, Jul. 1984.
May 2010, pp. 593–596. [475] K.-C. Chang, C.-Y. Chong, and Y. Bar-Shalom, “Joint probabilistic
data association in distributed sensor networks,” IEEE Transactions on
[451] E. W. Dijkstra, “A note on two problems in connexion with graphs,”
Automatic Control, vol. 31, no. 10, pp. 889–897, Oct. 1986.
Numerische Mathematik, vol. 1, no. 1, pp. 269–271, 1959. [Online].
[476] J. A. Roecker and G. L. Phillis, “Suboptimal joint probabilistic data
Available: http://dx.doi.org/10.1007/BF01386390
association,” IEEE Transactions on Aerospace and Electronic Systems,
[452] W. Xu, Y. Wang, Z. Zhou, and J. Wang, “A computationally efficient
vol. 29, no. 2, pp. 510–517, Apr. 1993.
exact ML sphere decoder,” in Proc. IEEE Global Telecommunications
[477] D. Musicki, R. J. Evans, and S. Stankovic, “Integrated probabilistic
Conference (GLOBECOM’04), Dallas, TX, USA, Dec. 2004, pp. 2594–
data association,” IEEE Transactions on Automatic Control, vol. 39,
2598.
no. 6, pp. 1237–1241, Jun. 1994.
[453] W. H. Mow, “Maximum likelihood sequence estimation from the lattice
[478] D. J. Kershaw and R. J. Evans, “Waveform selective probabilistic data
viewpoint,” IEEE Transactions on Information Theory, vol. 40, no. 5,
association,” IEEE Transactions on Aerospace and Electronic Systems,
pp. 1591–1600, Sep. 1994.
vol. 33, no. 4, pp. 1180–1188, Oct. 1997.
[454] D. Wübben, D. Seethaler, J. Jaldén, and G. Matz, “Lattice reduction,” [479] H. A. P. Blom and E. A. Bloem, “Probabilistic data association avoiding
IEEE Signal Processing Magazine, vol. 28, no. 3, pp. 70–91, May track coalescence,” IEEE Transactions on Automatic Control, vol. 45,
2011. no. 2, pp. 247–259, Feb. 2000.
[455] H. Daudé, P. Flajolet, and B. Vallée, “An analysis of the gaussian [480] T. Kirubarajan and Y. Bar-Shalom, “Probabilistic data association
algorithm for lattice reduction,” in Proc. of the 1st International techniques for target tracking in clutter,” Proceedings of the IEEE,
Symposium on Algorithmic Number Theory, Ithaca, NY, 1994, pp. 144– vol. 92, no. 3, pp. 536–557, Mar. 2004.
158. [481] D. Musicki and R. J. Evans, “Joint integrated probabilistic data
[456] A. Korkine and G. Zolotareff, “Sur les formes quadratiques,” association: JIPDA,” IEEE Transactions on Aerospace and Electronic
Mathematische Annalen, vol. 6, no. 3, pp. 366–389, 1873. [Online]. Systems, vol. 40, no. 3, pp. 1093–1099, Jul. 2004.
Available: http://dx.doi.org/10.1007/BF01442795 [482] I. J. Cox, “A review of statistical data association techniques
[457] H. Yao, “Efficient signal, code, and receiver designs for MIMO for motion correspondence,” International Journal of Computer
communication systems,” Ph.D. dissertation, 2003. Vision, vol. 10, no. 1, pp. 53–66, Feb. 1993. [Online]. Available:
[458] W. Z. Zhang, “Wireless receiver designs: From information theory to http://dx.doi.org/10.1007/BF01440847
VLSI implementation,” Ph.D. dissertation, 2009. [483] D. Schulz, W. Burgard, D. Fox, and A. B. Cremers, “Tracking multiple
[459] C.-P. Schnorr and M. Euchner, “Lattice basis reduction: Improved moving targets with a mobile robot using particle filters and statistical
practical algorithms and solving subset sum problems,” Mathematical data association,” in Proc. IEEE International Conference on Robotics
programming, vol. 66, no. 1-3, pp. 181–199, 1994. and Automation (ICRA’01), Seoul, Korea, May 2001, pp. 1665–1670.
[460] A. K. Lenstra, J. Lenstra, H. W., and L. Lovász, “Factoring [484] ——, “People tracking with mobile robots using sample-based joint
polynomials with rational coefficients,” Mathematische Annalen, probabilistic data association filters,” International Journal of Robotics
vol. 261, no. 4, pp. 515–534, 1982. [Online]. Available: http: Research, vol. 22, no. 2, pp. 99–116, Feb. 2003.
//dx.doi.org/10.1007/BF01457454 [485] C. Rasmussen and G. D. Hager, “Probabilistic data association methods
[461] M. Seysen, “Simultaneous reduction of a lattice basis and its reciprocal for tracking complex visual objects,” IEEE Transactions on Pattern
basis,” Combinatorica, vol. 13, no. 3, pp. 363–376, 1993. [Online]. Analysis and Machine Intelligence, vol. 23, no. 6, pp. 560–576, Jun.
Available: http://dx.doi.org/10.1007/BF01202355 2001.
[462] D. Seethaler and G. Matz, “Efficient vector perturbation in multi- [486] Z. J. Wang, Z. Han, and K. J. R. Liu, “MIMO-OFDM channel
antenna multi-user systems based on approximate integer relations,” estimation via probabilistic data association based TOAs,” in Proc.
in Proc. 14th European Signal Processing Conference (EUSIPCO’06), IEEE Global Telecommunications Conference (GLOBECOM’03), San
Florence, Italy, Sep. 2006. Francisco, CA, USA, Dec. 2003, pp. 626–630.

[487] ——, “A MIMO-OFDM channel estimation approach using time of [509] D. Chizhik, G. J. Foschini, and R. A. Valenzuela, “Capacities of multi-
arrivals,” IEEE Transactions on Wireless Communications, vol. 4, no. 3, element transmit and receive antennas: correlations and keyholes,”
pp. 1207–1213, May 2005. Electronics Letters, vol. 36, no. 13, pp. 1099–1100, Jun. 2000.
[488] S. Yang, T. Lv, and L. Hanzo, “Unified bit-based probabilistic data [510] D. Chizhik, G. J. Foschini, M. J. Gans, and R. A. Valenzuela,
association aided MIMO detection for high-order QAM,” in Proc. IEEE “Keyholes, correlations, and capacities of multielement transmit and
Wireless Communications and Networking Conference (WCNC’11), receive antennas,” IEEE Transactions on Wireless Communications,
Cancun, Mexico, Mar. 2011, pp. 1629–1634. vol. 1, no. 2, pp. 361–368, Apr. 2002.
[489] S. Yang and L. Hanzo, “Iterative detection and decoding using approx- [511] D. Gesbert, H. Bolcskei, D. Gore, and A. Paulraj, “MIMO wireless
imate Bayesian theorem based PDA method over MIMO Nakagami-m channels: capacity and performance prediction,” in Proc. IEEE Global
fading channels,” in Proc. IEEE Global Communications Conference Telecommunications Conference (GLOBECOM’00), San Francisco,
(GLOBECOM’12), Anaheim, CA, USA, Dec. 2012, pp. 3588–3593. CA, US, Nov. 2000, pp. 1083–1088.
[490] ——, “Exact Bayes’ theorem based probabilistic data association [512] P. Almers, F. Tufvesson, and A. F. Molisch, “Measurement of keyhole
for iterative MIMO detection and decoding,” in Proc. IEEE Global effect in a wireless multiple-input multiple-output MIMO channel,”
Communications Conference (GLOBECOM’13), Atlanta, GA, USA, IEEE Communications Letters, vol. 7, no. 8, pp. 373–375, Aug. 2003.
Dec. 2013. [513] ——, “Keyhole effect in MIMO wireless channels: measurements and
[491] C. Berrou, A. Glavieux, and P. Thitimajshima, “Near Shannon limit theory,” IEEE Transactions on Wireless Communications, vol. 5, no. 12,
error-correcting coding and decoding: Turbo-codes (1),” in Proc. IEEE pp. 3596–3604, Dec. 2006.
International Conference on Communications (ICC’93), vol. 2, Geneva, [514] M. K. Varanasi, “Group detection for synchronous Gaussian code-
Switzerland, May 1993, pp. 1064–1070. division multiple-access channels,” IEEE Transactions on Information
[492] C. Berrou and A. Glavieux, “Near optimum error correcting coding Theory, vol. 41, no. 4, pp. 1083–1096, Jul. 1995.
and decoding: turbo-codes,” IEEE Transactions on Communications, [515] C. Schlegel, S. Roy, P. D. Alexander, and Z.-J. Xiang, “Multiuser pro-
vol. 44, no. 10, pp. 1261–1271, Oct. 1996. jection receivers,” IEEE Journal on Selected Areas in Communications,,
[493] R. Gallager, “Low-density parity-check codes,” IRE Transactions on vol. 14, no. 8, pp. 1610–1618, Oct. 1996.
Information Theory, vol. 8, no. 1, pp. 21–28, Jan. 1962. [516] E. A. Fain and M. K. Varanasi, “Diversity order gain for narrow-band
[494] D. J. C. MacKay and R. M. Neal, “Near Shannon limit performance multiuser communications with pre-combining group detection,” IEEE
of low density parity check codes,” Electronics Letters, vol. 32, no. 18, Transactions on Communications,, vol. 48, no. 4, pp. 533–536, Apr.
pp. 1645–1646, 1996. 2000.
[495] L. Vandenberghe and S. Boyd, “Semidefinite programming,” SIAM [517] A. Kapur and M. K. Varanasi, “Multiuser detection for overloaded
Review, vol. 38, no. 1, pp. 49–95, Mar. 1996. CDMA systems,” IEEE Transactions on Information Theory, vol. 49,
[496] S. Boyd and L. Vandenberghe, Convex Optimization. New York, USA: no. 7, pp. 1728–1742, Jul. 2003.
Cambridge University Press, 2004. [518] B. Zarikoff, J. K. Cavers, and S. Bavarian, “An iterative groupwise
multiuser detector for overloaded MIMO applications,” IEEE Trans-
[497] C. Helmberg, F. Rendl, R. J. Vanderbei, and H. Wolkowicz, “An
actions on Wireless Communications, vol. 6, no. 2, pp. 443–447, Feb.
interior-point method for semidefinite programming,” SIAM Journal
2007.
on Optimization, vol. 6, pp. 342–361, 1996.
[519] M. Krause, D. P. Taylor, and P. A. Martin, “List-based group-wise
[498] P. H. Tan, L. K. Rasmussen, and T. M. Aulin, “The application of
symbol detection for multiple signal communications,” IEEE Trans-
semidefinite programming for detection in CDMA,” in Proc. IEEE In-
actions on Wireless Communications,, vol. 10, no. 5, pp. 1636–1644,
ternational Symposium on Information Theory (ISIT’01), Washington,
May 2011.
DC, Jun. 2001, p. 9.
[520] P. Dayal and M. K. Varanasi, “A fast generalized sphere decoder
[499] W.-K. Ma, T. N. Davidson, K. M. Wong, Z.-Q. Luo, and P.-C. for optimum decoding of under-determined MIMO systems,” in Proc.
Ching, “Efficient quasi-maximum-likelihood multiuser detection by 41st Annual Allerton Conference On Communication, Control and
semi-definite relaxation,” in Proc. IEEE International Conference on Computing (Allerton’03), Montecello, IL, US, Oct. 2003, pp. 1216–
Communications (ICC’01), Helsinki, Finland, Jun. 2001, pp. 6–10. 1225.
[500] X. Wang, W.-S. Lu, and A. Antoniou, “A near-optimal multiuser [521] T. Cui and C. Tellambura, “An efficient generalized sphere decoder for
detector for CDMA channels using semidefinite programming relax- rank-deficient MIMO systems,” IEEE Communications Letters, vol. 9,
ation,” in Proc. IEEE International Symposium on Circuits and Systems no. 5, pp. 423–425, May 2005.
(ISCAS’01), Sydney, NSW, Australia, May 2001, pp. 298–301. [522] Z. Yang, C. Liu, and J. He, “A new approach for fast generalized sphere
[501] ——, “A near-optimal multiuser detector for DS-CDMA systems using decoding in MIMO systems,” IEEE Signal Processing Letters, vol. 12,
semidefinite programming relaxation,” IEEE Transactions on Signal no. 1, pp. 41–44, Jan. 2005.
Processing, vol. 51, no. 9, pp. 2446–2450, Sep. 2003. [523] A. Wolfgang, J. Akhtman, S. Chen, and L. Hanzo, “Iterative MIMO
[502] B. Steingrimsson, Z.-Q. Luo, and K. M. Wong, “Soft quasi-maximum- detection for rank-deficient systems,” IEEE Signal Processing Letters,
likelihood detection for multiple-antenna wireless channels,” IEEE vol. 13, no. 11, pp. 699–702, Nov. 2006.
Transactions on Signal Processing, vol. 51, no. 11, pp. 2710–2719, [524] K.-K. Wong, A. Paulraj, and R. D. Murch, “Efficient high-performance
Nov. 2003. decoding for overloaded MIMO antenna systems,” IEEE Transactions
[503] P. F. Driessen and G. J. Foschini, “On the capacity formula for multiple on Wireless Communications, vol. 6, no. 5, pp. 1833–1843, May 2007.
input-multiple output wireless channels: a geometric interpretation,” [525] L. Wang, L. Xu, S. Chen, and L. Hanzo, “Generic iterative search-
IEEE Transactions on Communications, vol. 47, no. 2, pp. 173–176, centre-shifting K-best sphere detection for rank-deficient SDM-OFDM
Feb. 1999. systems,” Electronics Letters, vol. 44, no. 8, pp. 552–553, 2008.
[504] D.-S. Shiu, G. J. Foschini, M. J. Gans, and J. M. Kahn, “Fading [526] Z. Tian, G. Leus, and V. Lottici, “Detection of sparse signals under
correlation and its effect on the capacity of multielement antenna finite-alphabet constraints,” in Proc. IEEE International Conference
systems,” IEEE Transactions on Communications, vol. 48, no. 3, pp. on Acoustics, Speech and Signal Processing (ICASSP’2009), Taipei,
502–513, Mar. 2000. Republic of China, Apr. 2009, pp. 2349–2352.
[505] D. Chizhik, F. Rashid-Farrokhi, J. Ling, and A. Lozano, “Effect of [527] I. Kanaras, A. Chorti, M. R. D. Rodrigues, and I. Darwazeh, “A
antenna separation on the capacity of BLAST in correlated channels,” fast constrained sphere decoder for ill conditioned communication
IEEE Communications Letters, vol. 4, no. 11, pp. 337–339, Nov. 2000. systems,” IEEE Communications Letters, vol. 14, no. 11, pp. 999–1001,
[506] C.-N. Chuah, D. N. C. Tse, J. M. Kahn, and R. A. Valenzuela, Nov. 2010.
“Capacity scaling in MIMO wireless systems under correlated fading,” [528] T. Cui, T. Ho, and C. Tellambura, “Polynomial moment relaxation
IEEE Transactions on Information Theory, vol. 48, no. 3, pp. 637–650, for MIMO detection,” in Proc. IEEE International Conference on
Mar. 2002. Communications (ICC’06), Istanbul, Turkey, Jun. 2006, pp. 3129–3134.
[507] A. F. Molisch, M. Steinbauer, M. Toeltsch, E. Bonek, and R. S. Thoma, [529] M. J. Juntti, T. Schlosser, and J. O. Lilleberg, “Genetic algorithms for
“Capacity of MIMO systems based on measured wireless channels,” multiuser detection in synchronous CDMA,” in Proc. IEEE Interna-
IEEE Journal on Selected Areas in Communications, vol. 20, no. 3, tional Symposium on Information Theory (ISIT’97), Ulm, Germany,
pp. 561–569, Apr. 2002. Jul. 1997, p. 492.
[508] H. Shin and J. H. Lee, “Capacity of multiple-antenna fading channels: [530] C. Ergün and K. Hacioglu, “Application of a genetic algorithm to
spatial fading correlation, double scattering, and keyhole,” IEEE Trans- multi-stage detection in CDMA systems,” in Proc. 9th Mediterranean
actions on Information Theory,, vol. 49, no. 10, pp. 2636–2647, Oct. Electrotechnical Conference (MELECON’98), Tel-Aviv, Israel, May
2003. 1998, pp. 846–850.

[531] ——, “Multiuser detection using a genetic algorithm in CDMA com- IEEE International Conference on Communications (ICC’93), vol. 3,
munications systems,” IEEE Transactions on Communications, vol. 48, Geneva, Switzerland, May 1993, pp. 1740–1745.
no. 8, pp. 1374–1383, Aug. 2000. [552] J. Hagenauer, “The Turbo principle: tutorial introduction and state of
[532] X. Wang, W.-S. Lu, and A. Antoniou, “A genetic-algorithm-based the art,” in Proc. 1st International Symposium on Turbo Codes and
multiuser detector for multiple-access communications,” in Proc. IEEE Related Topics, Brest, France, Sep. 1997, pp. 1–11.
International Symposium on Circuits and Systems (ISCAS’98), Mon- [553] J. Hagenauer, E. Offer, and L. Papke, “Iterative decoding of binary
terey, CA, USA, Jun. 1998, pp. 534–537. block and convolutional codes,” IEEE Transactions on Information
[533] K. Yen and L. Hanzo, “Genetic algorithm assisted joint multiuser sym- Theory, vol. 42, no. 2, pp. 429–445, Mar. 1996.
bol detection and fading channel estimation for synchronous CDMA [554] M. Moher, “Turbo-based multiuser detection,” in Proc. IEEE Interna-
systems,” IEEE Journal on Selected Areas in Communications, vol. 19, tional Symposium on Information Theory (ISIT’97), Ulm, Germany,
no. 6, pp. 985–998, Jun. 2001. Jun. 1997, p. 195.
[534] ——, “Antenna-diversity-assisted genetic-algorithm-based multiuser [555] ——, “An iterative multiuser decoder for near-capacity communica-
detection schemes for synchronous CDMA systems,” IEEE Transac- tions,” IEEE Transactions on Communications, vol. 46, no. 7, pp. 870–
tions on Communications, vol. 51, no. 3, pp. 366–370, Mar. 2003. 880, Jul. 1998.
[535] ——, “Genetic-algorithm-assisted multiuser detection in asynchronous [556] M. C. Reed, P. D. Alexander, J. A. Asenstorfer, and C. B. Schlegel,
CDMA communications,” IEEE Transactions on Vehicular Technology, “Near single user performance using iterative multi-user detection for
vol. 53, no. 5, pp. 1413–1422, Sep. 2004. CDMA with turbo-code decoders,” in Proc. 8th IEEE Symposium
[536] G. W. K. Colman and T. J. Willink, “Overloaded array processing using on Personal, Indoor and Mobile Radio Communications (PIMRC’97),
genetic algorithms with soft-biased initialization,” IEEE Transactions vol. 2, Helsinki, Finland, Sep. 1997, pp. 740–744.
on Vehicular Technology, vol. 57, no. 4, pp. 2123–2131, Jul. 2008. [557] M. C. Reed, C. B. Schlegel, P. D. Alexander, and J. A. Asenstorfer,
[537] M. Dorigo, M. Birattari, and T. Stutzle, “Ant colony optimization,” “Iterative multiuser detection for CDMA with FEC: Near-single-user
IEEE Computational Intelligence Magazine, vol. 1, no. 4, pp. 28–39, performance,” IEEE Transactions on Communications, vol. 46, no. 12,
Nov 2006. pp. 1693–1699, 1998.
[538] J. Sun, W. Xu, and B. Feng, “A global search strategy of quantum- [558] F. Tarköy, “Iterative multi-user decoding for asynchronous users,” in
behaved particle swarm optimization,” in Proc. IEEE Conference on Proc. IEEE International Symposium on Information Theory (ISIT’97),
Cybernetics and Intelligent Systems (CIS’04), vol. 1, Dec. 2004, pp. Ulm, Germany, Jun. 1997, p. 30.
111–116. [559] L. Nelson and H. V. Poor, “Iterative multiuser receivers for CDMA
[539] S. L. Hijazi and B. Natarajan, “Novel low-complexity DS-CDMA channels: an EM-based approach,” IEEE Transactions on Communica-
multiuser detector based on ant colony optimization,” in Proc. IEEE tions, vol. 44, no. 12, pp. 1700–1710, Dec. 1996.
60th Vehicular Technology Conference (VTC’04-Fall), vol. 3, Sep. [560] X. Wang and H. V. Poor, “Iterative (turbo) soft interference cancellation
2004, pp. 1939–1943. and decoding for coded CDMA,” IEEE Transactions on Communica-
[540] C. Xu, B. Hu, L.-L. Yang, and L. Hanzo, “Ant-colony-based multiuser tions, vol. 47, no. 7, pp. 1046–1061, Jul. 1999.
detection for multifunctional-antenna-array-assisted MC DS-CDMA
[561] H. El Gamal and E. Geraniotis, “Iterative multiuser detection for coded
systems,” IEEE Transactions on Vehicular Technology, vol. 57, no. 1,
CDMA signals in awgn and fading channels,” IEEE Journal on Selected
pp. 658–663, Jan. 2008.
Areas in Communications, vol. 18, no. 1, pp. 30–41, Jan. 2000.
[541] C. Xu, R. G. Maunder, L.-L. Yang, and L. Hanzo, “Near-optimum
[562] H. Lee, B. Lee, and I. Lee, “Iterative detection and decoding with
multiuser detectors using soft-output ant-colony-optimization for the
an improved V-BLAST for MIMO-OFDM systems,” IEEE Journal on
DS-CDMA uplink,” IEEE Signal Processing Letters, vol. 16, no. 2,
Selected Areas in Communications, vol. 24, no. 3, pp. 504–513, Mar.
pp. 137–140, Feb. 2009.
2006.
[542] J.-K. Lain and J.-Y. Chen, “Near-MLD MIMO detection based on
[563] Y. De Jong and T. Willink, “Iterative tree search detection for MIMO
a modified ant colony optimization,” IEEE Communications Letters,
wireless systems,” IEEE Transactions on Communications, vol. 53,
vol. 14, no. 8, pp. 722–724, August 2010.
no. 6, pp. 930–935, Jun. 2005.
[543] K. T. Tasneem, P. A. Martin, and D. P. Taylor, “Iterative soft detection
of cochannel signals using ant colony optimization,” in Proc. 23rd [564] M. Nekuii, M. Kisialiou, T. Davidson, and Z.-Q. Luo, “Efficient soft-
IEEE International Symposium on Personal, Indoor and Mobile Radio output demodulation of MIMO QPSK via semidefinite relaxation,”
Communications (PIMRC’12), Sydney, Australia, Sep. 2012, pp. 1617– IEEE Journal of Selected Topics in Signal Processing, vol. 5, no. 8,
1621. pp. 1426–1437, Dec. 2011.
[544] P. A. Haris, E. Gopinathan, and C. K. Ali, “Artificial bee colony and [565] V. Pauli, L. Lampe, and R. Schober, “Turbo DPSK using soft multiple-
tabu search enhanced TTCM assisted MMSE multi-user detectors for symbol differential sphere decoding,” IEEE Transactions on Informa-
rank deficient SDMA-OFDM system,” Wireless Personal Communica- tion Theory, vol. 52, no. 4, pp. 1385–1398, Apr. 2006.
tions, vol. 65, no. 2, pp. 425–442, 2012. [566] S. ten Brink, “Convergence behavior of iteratively decoded parallel
[545] S. Yang, M. Wang, and L. Jiao, “A quantum particle swarm opti- concatenated codes,” IEEE Transactions on Communications, vol. 49,
mization,” in Proc. Congress on Evolutionary Computation (CEC’04), no. 10, pp. 1727–1737, Oct. 2001.
vol. 1, Jun. 2004, pp. 320–324. [567] J. Hagenauer, “The EXIT chart–Introduction to extrinsic information
[546] Y. Zhao and J. Zheng, “Particle swarm optimization algorithm in signal transfer in iterative processing,” in Proc. 12th European Signal Pro-
detection and blind extraction,” in Proc. 7th International Symposium cessing Conference (EUSIPCO’04), Vienna, Austria, Sep. 2004, pp.
on Parallel Architectures, Algorithms and Networks (ISPAN’04), Hong 1541–1548.
Kong, China, May 2004, pp. 37–41. [568] S. L. Ariyavisitakul, “Turbo space-time processing to improve wireless
[547] S. Chen, W. Yao, H. R. Palally, and L. Hanzo, “Particle swarm channel capacity,” IEEE Transactions on Communications, vol. 48,
optimisation aided MIMO transceiver designs,” in Computational Intel- no. 8, pp. 1347–1359, Aug. 2000.
ligence in Expensive Optimization Problems, ser. Adaptation Learning [569] S. Baro, J. Hagenauer, and M. Witzke, “Iterative detection of MIMO
and Optimization, Y. Tenne and C.-K. Goh, Eds. Springer Berlin transmission using a list-sequential (LISS) detector,” in Proc. IEEE
Heidelberg, 2010, vol. 2, pp. 487–511. International Conference on Communications (ICC’03), Anchorage,
[548] T. Abrão, L. D. de Oliveira, F. Ciriaco, B. A. Angélico, P. E. Jeszensky, Alaska, US, May 2003, pp. 2653–2657.
and F. J. Casadevall Palacio, “S/MIMO MC-CDMA heuristic multiuser [570] S. Haykin, M. Sellathurai, Y. De Jong, and T. Willink, “Turbo-
detectors based on single-objective optimization,” Wireless Personal MIMO for wireless communications,” IEEE Communications Maga-
Communications, vol. 53, no. 4, pp. 529–553, 2010. zine, vol. 42, no. 10, pp. 48–53, Oct. 2004.
[549] J. Xia, T. Lv, X. Yun, X. Su, and S. Yang, “Simulated annealing based [571] J. Hagenauer and C. Kuhn, “The list-sequential (LISS) algorithm and
multiuser detection for synchronous SDMA system,” in Proc. 11th its application,” IEEE Transactions on Communications, vol. 55, no. 5,
IEEE Singapore International Conference on Communication Systems pp. 918–928, May 2007.
(ICCS’08), Nov. 2008, pp. 441–445. [572] C. B. Peel, B. M. Hochwald, and A. L. Swindlehurst, “A vector-
[550] T. Datta, N. Srinidhi, A. Chockalingam, and B. S. Rajan, “Low- perturbation technique for near-capacity multiantenna multiuser com-
complexity near-optimal signal detection in underdetermined large- munication – part I: channel inversion and regularization,” IEEE
MIMO systems,” in Proc. National Conference on Communications Transactions on Communications, vol. 53, no. 1, pp. 195–202, Jan.
(NCC’12), Kharagpur, India, Feb. 2012, pp. 1–5. 2005.
[551] J. Lodge, R. Young, P. A. Hoeher, and J. Hagenauer, “Separable MAP [573] B. M. Hochwald, C. B. Peel, and A. L. Swindlehurst, “A vector-
’filters’ for the decoding of product and concatenated codes,” in Proc. perturbation technique for near-capacity multiantenna multiuser com-

munication - part II: perturbation,” IEEE Transactions on Communica- [596] J. Jaldén, L. G. Barbero, B. Ottersten, and J. S. Thompson, “Full
tions, vol. 53, no. 3, pp. 537–544, Mar. 2005. diversity detection in MIMO systems with a fixed-complexity sphere
[574] O. Alamri, J. Wang, S. X. Ng, L.-L. Yang, and L. Hanzo, “Near- decoder,” in Proc. IEEE International Conference on Acoustics, Speech,
capacity three-stage turbo detection of irregular convolutional coded and Signal Processing (ICASSP’07), vol. 3, Honolulu, HI, Apr. 2007,
joint sphere-packing modulation and space-time coding,” IEEE Trans- pp. II–49–III–52.
actions on Communications, vol. 57, no. 5, pp. 1486–1495, May 2009. [597] C. Ling, “Towards characterizing the performance of approximate
[575] L. Hanzo, M. El-Hajjar, and O. Alamri, “Near-capacity wireless lattice decoding in MIMO communications,” in Proc. 4th International
transceivers and cooperative communications in the MIMO era: Evo- Symposium on Turbo Codes Related Topics/6th International ITG-
lution of standards, waveform design, and future perspectives,” Pro- Conference on Source and Channel Coding, Munich, Germany, Apr.
ceedings of the IEEE, vol. 99, no. 8, pp. 1343–1385, Aug. 2011. 2006, pp. 1–6.
[576] P. Suthisopapan, K. Kasai, V. Imtawil, and A. Meesomboon, “Ap- [598] M. Taherzadeh and A. K. Khandani, “On the limitations of the naive
proaching capacity of large MIMO systems by non-binary LDPC codes lattice decoding,” IEEE Transactions on Information Theory, vol. 56,
and MMSE detection,” in Proc. IEEE International Symposium on no. 10, pp. 4820–4826, Oct. 2010.
Information Theory (ISIT’12), Cambridge, MA, Jul. 2012, pp. 1712– [599] C. Ling, “On the proximity factors of lattice reduction-aided decoding,”
1716. IEEE Transactions on Signal Processing, vol. 59, no. 6, pp. 2795–2808,
[577] L. Hanzo, O. Alamri, M. El-Hajjar, and N. Wu, Near-Capacity Multi- Jun. 2011.
Functional MIMO Systems: Sphere-Packing, Iterative Detection and [600] J. Jaldén and P. Elia, “DMT optimality of LR-aided linear decoders for
Cooperation. John Wiley & Sons, 2009. a general class of channels, lattice designs, and system models,” IEEE
[578] L. Hanzo, T. H. Liew, B. L. Yeap, R. Y. S. Tee, and S. X. Ng, Transactions on Information Theory, vol. 56, no. 10, pp. 4765–4780,
Turbo Coding, Turbo Equalisation and Space-Time Coding: EXIT- Oct. 2010.
Chart-Aided Near-Capacity Designs for Wireless Channels. John [601] A. K. Singh, P. Elia, and J. Jaldén, “Achieving a vanishing SNR
Wiley & Sons, 2011. gap to exact lattice decoding at a subexponential complexity,” IEEE
[579] L. Hanzo, R. G. Maunder, J. Wang, and L.-L. Yang, Near-capacity Transactions on Information Theory, vol. 58, no. 6, pp. 3692–3707,
variable-length coding: regular and EXIT-chart-aided irregular de- June 2012.
signs. John Wiley & Sons, 2011. [602] A. Yellepeddi, K. J. Kim, C. Duan, and P. Orlik, “On probabilistic
[580] H. Wymeersch, Iterative receiver design. Cambridge University Press, data association for achieving near-exponential diversity over fading
2007. channels,” in Proc. IEEE International Conference on Communications
[581] H. Sampath, P. Stoica, and A. Paulraj, “Generalized linear precoder (ICC’13), Budapest, Hungary, Jun. 2013, pp. 5409–5414.
and decoder design for MIMO channels using the weighted MMSE [603] J. Jaldén, C. Martin, and B. Ottersten, “Semidefinite programming
criterion,” IEEE Transactions on Communications, vol. 49, no. 12, pp. for detection in linear systems - optimality conditions and space-
2198–2206, Dec. 2001. time decoding,” in Acoustics, Speech, and Signal Processing, 2003.
[582] D. P. Palomar, J. M. Cioffi, and M.-A. Lagunas, “Joint Tx-Rx beam- Proceedings. (ICASSP ’03). 2003 IEEE International Conference on,
forming design for multicarrier MIMO channels: a unified framework vol. 4, Apr. 2003, pp. IV–9 – IV–12.
for convex optimization,” IEEE Transactions on Signal Processing, [604] A. M.-C. So, “On the performance of semidefinite relaxation mimo
vol. 51, no. 9, pp. 2381–2401, Sep. 2003. detectors for qam constellations,” in Proc. IEEE International Confer-
[583] D. J. Love and R. W. Heath, “Limited feedback unitary precoding ence on Acoustics, Speech and Signal Processing (ICASSP’09), Taipei,
for spatial multiplexing systems,” IEEE Transactions on Information Republic of China, Apr. 2009, pp. 2449–2452.
Theory, vol. 51, no. 8, pp. 2967–2976, Aug. 2005. [605] M. Kisialiou, X. Luo, and Z.-Q. Luo, “Efficient implementation of
[584] A. Wiesel, Y. C. Eldar, and S. Shamai, “Linear precoding via conic quasi- maximum-likelihood detection based on semidefinite relax-
optimization for fixed MIMO receivers,” IEEE Transactions on Signal ation,” IEEE Transactions on Signal Processing, vol. 57, no. 12, pp.
Processing, vol. 54, no. 1, pp. 161–176, Jan. 2006. 4811–4822, Dec. 2009.
[585] M. Sadek, A. Tarighat, and A. H. Sayed, “A leakage-based precoding [606]
scheme for downlink multi-user MIMO channels,” IEEE Transactions [607] T. L. Marzetta, “Noncooperative cellular wireless with unlimited
on Wireless Communications, vol. 6, no. 5, pp. 1711–1721, May 2007. numbers of base station antennas,” IEEE Transactions on Wireless
[586] M. Vu and A. Paulraj, “Mimo wireless linear precoding,” IEEE Signal Communications, vol. 9, no. 11, pp. 3590–3600, Nov. 2010.
Processing Magazine, vol. 24, no. 5, pp. 86–105, Sep. 2007. [608] A. Hu, T. Lv, H. Gao, Z. Zhang, and S. Yang, “An ESPRIT-based
[587] J. Zhang, R. Chen, J. G. Andrews, A. Ghosh, and R. W. Heath, “Net- approach for 2-D localization of incoherently distributed sources in
worked MIMO with clustered linear precoding,” IEEE Transactions on massive MIMO systems,” IEEE Journal of Selected Topics in Signal
Wireless Communications, vol. 8, no. 4, pp. 1910–1921, Apr. 2009. Processing, vol. 8, no. 5, pp. 996–1011, Oct. 2014.
[588] J. H. Winters, J. Salz, and R. D. Gitlin, “The impact of antenna [609] J. Zhang, B. Zhang, S. Chen, X. Mu, M. El-Hajjar, and L. Hanzo,
diversity on the capacity of wireless communication systems,” IEEE “Pilot contamination elimination for large-scale multiple-antenna aided
Transactions on Communications, vol. 42, no. 2/3/4, pp. 1740–1751, OFDM systems,” IEEE Journal of Selected Topics in Signal Processing,
Feb./Mar./Apr. 1994. vol. 8, no. 5, pp. 759 – 772, Oct. 2014.
[589] N. Prasad and M. K. Varanasi, “Analysis of decision feedback detection [610] H. Elgala, R. Mesleh, and H. Haas, “Indoor optical wireless commu-
for MIMO rayleigh-fading channels and the optimization of power and nication: potential and state-of-the-art,” IEEE Communications Maga-
rate allocations,” IEEE Transactions on Information Theory, vol. 50, zine, vol. 49, no. 9, pp. 56–62, Sep. 2011.
no. 6, pp. 1009–1025, Jun. 2004. [611] L. Hanzo, H. Haas, S. Imre, D. O’Brien, M. Rupp, and L. Gyongyosi,
[590] S. Loyka and F. Gagnon, “V-BLAST without optimal ordering: An- “Wireless myths, realities, and futures: from 3G/4G to optical and
alytical performance evaluation for Rayleigh fading channels,” IEEE quantum wireless,” Proceedings of the IEEE, vol. 100, no. Special
Transactions on Communications, vol. 54, no. 6, pp. 1109–1120, Jun. Centennial Issue, pp. 1853–1888, May 2012.
2006. [612] J. Hoydis, K. Hosseini, S. ten Brink, and M. Debbah, “Making smart
[591] Y. Jiang, X. Zheng, and J. Li, “Asymptotic performance analysis of use of excess antennas: massive MIMO, small cells, and TDD,” Bell
V-BLAST,” in Proc. IEEE Global Telecommunications Conference Labs Technical Journal, vol. 18, no. 2, pp. 5–21, Sep. 2013.
(GLOBECOM’05), St. Louis, US, Dec. 2005, pp. 3882–3886. [613] K. Zheng, S. Ou, and X. Yin, “Massive MIMO channel models: A
[592] S. Loyka and F. Gagnon, “Performance analysis of the V-BLAST survey,” International Journal of Antennas and Propagation, vol. 2014,
algorithm: An analytical approach,” IEEE Transactions on Wireless pp. 1 – 10, Jun. 2014.
Communications, vol. 3, no. 4, pp. 1326–1337, Jul. 2004. [614] A. M. Tulino and S. Verdú, “Random matrix theory and wireless
[593] D. Seethaler and H. Bolcskei, “Performance and complexity analysis communications,” Foundations and Trends R in Communications and
of infinity-norm sphere-decoding,” IEEE Transactions on Information Information Theory, vol. 1, no. 1, pp. 1–182, 2004. [Online].
Theory, vol. 56, no. 3, pp. 1085–1105, March 2010. Available: http://dx.doi.org/10.1561/0100000001
[594] D. Seethaler, J. J. Jaldén, C. Studer, and H. Bolcskei, “On the [615] M. L. Mehta, Random Matrices, 3rd ed., ser. Pure and Applied
complexity distribution of sphere decoding,” IEEE Transactions on Mathematics. Elsevier Science, 2004.
Information Theory, vol. 57, no. 9, pp. 5754–5768, Sep. 2011. [616] V. A. Marčenko and L. A. Pastur, “Distribution of eigenvalues for some
[595] J. Jaldén and P. Elia, “Sphere decoding complexity exponent for sets of random matrices,” Math USSR Shornik, vol. 1, no. 4, pp. 457–
decoding full-rate codes over the quasi-static MIMO channel,” IEEE 483, 1967.
Transactions on Information Theory, vol. 58, no. 9, pp. 5785–5803, [617] A. Chockalingam and B. S. Rajan, Large MIMO Systems, ser. Large
Sep. 2012. MIMO Systems. Cambridge University Press, 2014.

[618] T. L. Marzetta, “How much training is required for multiuser MIMO?” MIMO cognitive networks,” IEEE Transactions on Signal Processing,
in Proc. 40th Asilomar Conference on Signals, Systems and Computers vol. 58, no. 7, pp. 3728–3741, Jul. 2010.
(ACSSC’06), Pacific Grove, CA, Oct. 2006, pp. 359–363. [641] S.-J. Kim and G. B. Giannakis, “Optimal resource allocation for MIMO
[619] Y. Sun, “Eliminating-highest-error and fastest-metric-descent criteria ad hoc cognitive radio networks,” IEEE Transactions on Information
and iterative algorithms for bit-synchronous CDMA multiuser detec- Theory, vol. 57, no. 5, pp. 3117–3131, May 2011.
tion,” in Proc. IEEE International Conference on Communications [642] Y. J. A. Zhang and A. M.-C. So, “Optimal spectrum sharing in MIMO
(ICC’98), vol. 3, Atlanta, GA, Jun. 1998, pp. 1576–1580. cognitive radio networks via semidefinite programming,” IEEE Journal
[620] ——, “A family of linear complexity likelihood ascent search detectors on Selected Areas in Communications, vol. 29, no. 2, pp. 362–373, Feb.
for CDMA multiuser detection,” in Proc. IEEE 6th International 2011.
Symposium on Spread Spectrum Techniques and Applications, vol. 2, [643] J. Wang, G. Scutari, and D. P. Palomar, “Robust MIMO cognitive radio
Parsippany, NJ, Sep. 2000, pp. 713–717. via game theory,” IEEE Transactions on Signal Processing, vol. 59,
[621] ——, “A family of likelihood ascent search multiuser detectors: ap- no. 3, pp. 1183–1201, Mar. 2011.
proaching optimum performance via random multicodes with linear [644] C. S. Vaze and M. K. Varanasi, “The degree-of-freedom regions of
complexity,” IEEE Transactions on Communications, vol. 57, no. 8, MIMO broadcast, interference, and cognitive radio channels with no
pp. 2215–2220, Aug. 2009. CSIT,” IEEE Transactions on Information Theory, vol. 58, no. 8, pp.
[622] ——, “Hopfield neural network based algorithms for image restora- 5354–5374, Aug. 2012.
tion and reconstruction – part I: Algorithms and simulations,” IEEE [645] S. Sun, M. Kadoch, L. Gong, and B. Rong, “Integrating network
Transactions on Signal Processing, vol. 48, no. 7, pp. 2105–2118, Jul. function virtualization with SDR and SDN for 4G/5G networks,” IEEE
2000. Network, vol. 29, no. 3, pp. 54–59, May 2015.
[623] ——, “Hopfield neural network based algorithms for image restoration [646] E. Fishler, A. Haimovich, R. S. Blum, D. Chizhik, L. J. Cimini, and
and reconstruction – part II: Performance analysis,” IEEE Transactions R. A. Valenzuela, “MIMO radar: an idea whose time has come,” in
on Signal Processing, vol. 48, no. 7, pp. 2119–2131, Jul. 2000. Proceedings of the IEEE Radar Conference, Philadelphia, Pennsylva-
[624] B.-P. Paris, G. Orsak, M. K. Varanasi, and B. Aazhang, “Neural net nia, USA, Apr. 2004, pp. 71–78.
receivers in multiple access-communications,” in IEEE Conference on [647] E. Fishler, A. Haimovich, R. S. Blum, L. J. Cimini, D. Chizhik, and
Neural Information Processing Systems (NIPS’88), Denver, CO., Nov. R. A. Valenzuela, “Spatial diversity in radars-models and detection
1988, pp. 272–280. performance,” IEEE Transactions on Signal Processing, vol. 54, no. 3,
[625] B. Aazhang, B.-P. Paris, and G. C. Orsak, “Neural networks for pp. 823–838, Mar. 2006.
multiuser detection in code-division multiple-access communications,” [648] J. Li and P. Stoica, “MIMO radar with colocated antennas,” IEEE
IEEE Transactions on Communications, vol. 40, no. 7, pp. 1212–1222, Signal Processing Magazine, vol. 24, no. 5, pp. 106–114, Sep. 2007.
Jul. 1992. [649] A. M. Haimovich, R. S. Blum, and L. J. Cimini, “MIMO radar with
[626] U. Mitra and H. V. Poor, “Adaptive receiver algorithms for near- widely separated antennas,” IEEE Signal Processing Magazine, vol. 25,
far resistant CDMA,” in Proc. 3rd IEEE International Symposium no. 1, pp. 116–129, Jan. 2008.
on Personal, Indoor and Mobile Radio Communications (PIMRC’92), [650] J. Li and P. Stoica, Eds., MIMO radar signal processing. John Wiley
Boston, MA, Oct. 1992, pp. 639–644. & Sons, 2009.
[627] ——, “Neural network techniques for adaptive multiuser demodula- [651] A. De Maio and M. Lops, “Design principles of MIMO radar detec-
tion,” IEEE Journal on Selected Areas in Communications, vol. 12, tors,” IEEE Transactions on Aerospace and Electronic Systems, vol. 43,
no. 9, pp. 1460–1470, Dec. 1994. no. 3, pp. 886–898, Jul. 2007.
[652] J. Li, P. Stoica, and X. Zheng, “Signal synthesis and receiver design
[628] ——, “Adaptive receiver algorithms for near-far resistant CDMA,”
for MIMO radar imaging,” IEEE Transactions on Signal Processing,
IEEE Transactions on Communications, vol. 43, no. 2/3/4, pp. 1713–
vol. 56, no. 8, pp. 3959–3968, Aug. 2008.
1724, Feb./Mar./Apr. 1995.
[653] A. R. Shah, R. C. J. Hsu, A. Tarighat, A. H. Sayed, and B. Jalali, “Co-
[629] G. Kechriotis and E. S. Manolakos, “Hopfield neural network imple-
herent optical MIMO (COMIMO),” Journal of Lightwave Technology,
mentation of the optimal CDMA multiuser detector,” IEEE Transac-
vol. 23, no. 8, pp. 2410–2419, Aug. 2005.
tions on Neural Networks, vol. 7, no. 1, pp. 131–141, Jan. 1996.
[654] A. Tarighat, R. C. J. Hsu, A. Shah, A. H. Sayed, and B. Jalali,
[630] P. Li and R. D. Murch, “Multiple output selection-LAS algorithm in
“Fundamentals and challenges of optical multiple-input multiple-output
large MIMO systems,” IEEE Communications Letters, vol. 14, no. 5,
multimode fiber links,” IEEE Communications Magazine, vol. 45, no. 5,
pp. 399–401, May 2010.
pp. 57–63, May 2007.
[631] J. Mitola III and G. Q. Maguire Jr., “Cognitive radio: Making software [655] S. Randel, R. Ryf, A. Sierra, P. J. Winzer, A. H. Gnauck, C. A. Bolle,
radios more personal,” IEEE Personal Communications, vol. 6, no. 4, R.-J. Essiambre, D. W. Peckham, A. McCurdy, R. Lingle et al., “6×
pp. 13–18, Aug. 1999. 56-gb/s mode-division multiplexed transmission over 33-km few-mode
[632] J. Mitola III, “Cognitive radio: An integrated agent architecture for soft- fiber enabled by 6× 6 MIMO equalization,” Optics Express, vol. 19,
ware defined radio,” Ph.D. dissertation, Royal Institute of Technology no. 17, pp. 16 697–16 707, Aug. 2011.
(KTH), Sweden, May 2000. [656] R. Ryf, S. Randel, A. H. Gnauck, C. Bolle, R.-J. Essiambre, P. J.
[633] S. Haykin, “Cognitive radio: Brain-empowered wireless communica- Winzer, D. W. Peckham, A. McCurdy, and R. Lingle, “Space-division
tions,” IEEE Journal on Selected Areas in Communications, vol. 23, multiplexing over 10 km of three-mode fiber using coherent 6 × 6
no. 2, pp. 201–220, Feb. 2005. MIMO processing,” in Proceedings of Optical Fiber Communication
[634] R. Zhang and Y.-C. Liang, “Exploiting multi-antennas for opportunis- Conference / National Fiber Optic Engineers Conference, Los Angeles,
tic spectrum sharing in cognitive radio networks,” IEEE Journal of California, USA, Mar. 2011, p. PDPB10.
Selected Topics in Signal Processing, vol. 2, no. 1, pp. 88–102, Feb. [657] R. Ryf, S. Randel, A. H. Gnauck, C. Bolle, A. Sierra, S. Mumtaz,
2008. M. Esmaeelpour, E. C. Burrows, R.-J. Essiambre, P. J. Winzer et al.,
[635] S. Sridharan and S. Vishwanath, “On the capacity of a class of “Mode-division multiplexing over 96 km of few-mode fiber using
MIMO cognitive radios,” IEEE Journal of Selected Topics in Signal coherent 6 × 6 MIMO processing,” Journal of Lightwave Technology,
Processing, vol. 2, no. 1, pp. 103–117, Feb. 2008. vol. 30, no. 4, pp. 521–531, Feb. 2012.
[636] G. Scutari, D. P. Palomar, and S. Barbarossa, “Cognitive MIMO radio,” [658] R. Essiambre and R. W. Tkach, “Capacity trends and limits of optical
IEEE Signal Processing Magazine, vol. 25, no. 6, pp. 46–59, Nov. communication networks,” Proceedings of the IEEE, vol. 100, no. 5,
2008. pp. 1035–1055, May 2012.
[637] K. Hamdi, W. Zhang, and K. Letaief, “Opportunistic spectrum sharing [659] S. Arik, J. M. Kahn, and K.-P. Ho, “MIMO signal processing for
in cognitive MIMO wireless networks,” IEEE Transactions on Wireless mode-division multiplexing: An overview of channel models and signal
Communications, vol. 8, no. 8, pp. 4098–4109, Aug. 2009. processing architectures,” IEEE Signal Processing Magazine, vol. 31,
[638] G. Zheng, K.-K. Wong, and B. Ottersten, “Robust cognitive beamform- no. 2, pp. 25 – 34, Mar. 2014.
ing with bounded channel uncertainties,” IEEE Transactions on Signal
Processing, vol. 57, no. 12, pp. 4871–4881, Dec. 2009.
[639] G. Scutari and D. P. Palomar, “MIMO cognitive radio: A game
theoretical approach,” IEEE Transactions on Signal Processing, vol. 58,
no. 2, pp. 761–780, Feb. 2010.
[640] S. M. Perlaza, N. Fawaz, S. Lasaulce, and M. Debbah, “From spectrum
pooling to space pooling: Opportunistic interference alignment in

Shaoshi Yang (SIEEE’09-MIEEE’13) received his

B.Eng. degree in Information Engineering from Bei-
jing University of Posts and Telecommunications
(BUPT), Beijing, China in Jul. 2006, his first Ph.D.
degree in Electronics and Electrical Engineering
from University of Southampton, U.K. in Dec. 2013,
and his second Ph.D. degree in Signal and Informa-
tion Processing from BUPT in Mar. 2014. He is now
working as the IU-ATC Senior Research Fellow in
University of Southampton, U.K. From November
2008 to February 2009, he was an Intern Research
Fellow with the Communications Technology Lab (CTL), Intel Labs, Beijing,
China, where he focused on Channel Quality Indicator Channel (CQICH)
design for mobile WiMAX (802.16m) standard. His research interests include
MIMO signal processing, green radio, heterogeneous networks, cross-layer
interference management, convex optimization and its applications. He has
published in excess of 30 research papers on IEEE journals and conferences.
Shaoshi has received a number of academic and research awards, includ-
ing the PMC-Sierra Telecommunications Technology Scholarship at BUPT,
the Electronics and Computer Science (ECS) Scholarship of University of
Southampton, the Best PhD Thesis Award of BUPT, and the Dean’s Award
for Early Career Research Excellence of University of Southampton. He is
a member of IEEE/IET, and a junior member of Isaac Newton Institute for
Mathematical Sciences, Cambridge University, U.K. He also serves as a TPC
member of several major IEEE conferences, including IEEE ICC, PIMRC,
ICCVE, HPCC, and as a Guest Associate Editor of IEEE Journal on Selected
Areas in Communications. (https://sites.google.com/site/shaoshiyang/)

Lajos Hanzo (FREng, FIEEE’04, FIET, FEURASIP,

DSc) received his degree in electronics in 1976 and
his doctorate in 1983. In 2009 he was awarded the
honorary doctorate “Doctor Honoris Causa” by the
Technical University of Budapest and also by the
University of Edinburgh in 2015. During his 39-year
career in telecommunications he has held various
research and academic posts in Hungary, Germany
and the UK. Since 1986 he has been with the School
of Electronics and Computer Science, University
of Southampton, UK, where he holds the chair in
telecommunications. He has successfully supervised 100+ PhD students, co-
authored 20 John Wiley/IEEE Press books on mobile radio communications
totalling in excess of 10 000 pages, published 1400+ research entries at
IEEE Xplore, acted both as TPC and General Chair of IEEE conferences,
presented keynote lectures and has been awarded a number of distinctions.
Currently he is directing a 60-strong academic research team, working on a
range of research projects in the field of wireless multimedia communications
sponsored by industry, the Engineering and Physical Sciences Research
Council (EPSRC) UK, the European Research Council’s Advanced Fellow
Grant and the Royal Society’s Wolfson Research Merit Award. He is an
enthusiastic supporter of industrial and academic liaison and he offers a range
of industrial courses. He is also a Governor of the IEEE VTS. During 2008 -
2012 he was the Editor-in-Chief of the IEEE Press and a Chaired Professor
also at Tsinghua University, Beijing. Lajos has 22 000+ citations. For further
information on research in progress and associated publications please refer
to http://www-mobile.ecs.soton.ac.uk

HND Networking and Cloud Computing (NCC)
100% (2)
HND Networking and Cloud Computing (NCC)
177 pages
Arbor APS STT - Unit 11 - Protection Setting Details - 25jan2018
No ratings yet
Arbor APS STT - Unit 11 - Protection Setting Details - 25jan2018
88 pages
WBT MCQ Bank
No ratings yet
WBT MCQ Bank
108 pages
Introduction To Juniper Security (IJSEC)
No ratings yet
Introduction To Juniper Security (IJSEC)
3 pages
Large MIMO Systems PDF
100% (1)
Large MIMO Systems PDF
336 pages
Simulation of Digital Communication Systems Using Matlab
From Everand
Simulation of Digital Communication Systems Using Matlab
Mathuranathan Viswanathan
3.5/5 (22)
Massive MIMO Detection Techniques: A Survey
No ratings yet
Massive MIMO Detection Techniques: A Survey
25 pages
Detailed Study On Low Complexity Detection Techniques For Large MIMO System: A Review
No ratings yet
Detailed Study On Low Complexity Detection Techniques For Large MIMO System: A Review
7 pages
Novel MIMO Detection Algorithm For High-Order Constellations in The Complex Domain
No ratings yet
Novel MIMO Detection Algorithm For High-Order Constellations in The Complex Domain
14 pages
Detection of Data Symbol in A Massive MIMO Systems For 5G Wireless Communication
No ratings yet
Detection of Data Symbol in A Massive MIMO Systems For 5G Wireless Communication
4 pages
VLSI Implementation of MIMO Detection Using The Sphere Decoding Algorithm
No ratings yet
VLSI Implementation of MIMO Detection Using The Sphere Decoding Algorithm
12 pages
Efficient Soft MIMO Detection Algorithms Based On Differential Metrics
No ratings yet
Efficient Soft MIMO Detection Algorithms Based On Differential Metrics
5 pages
International Journal of Computational Engineering Research (IJCER)
No ratings yet
International Journal of Computational Engineering Research (IJCER)
6 pages
Compressive Sensing Based Multi-User Detector For Large-Scale SM-MIMO Uplink
No ratings yet
Compressive Sensing Based Multi-User Detector For Large-Scale SM-MIMO Uplink
6 pages
Incredible VLSI Design For MIMO System Using SEC-QPSK Detection
No ratings yet
Incredible VLSI Design For MIMO System Using SEC-QPSK Detection
12 pages
A Survey of VLSI Implementations of Tree Search Al
No ratings yet
A Survey of VLSI Implementations of Tree Search Al
32 pages
Signal Detection in MIMO Communications System With Non-Gaussian Noises Based On Deep Learning and Maximum Correntropy Criterion
No ratings yet
Signal Detection in MIMO Communications System With Non-Gaussian Noises Based On Deep Learning and Maximum Correntropy Criterion
16 pages
A Performance Study of MIMO Detectors: Christoph Windpassinger, Lutz Lampe, Robert F. H. Fischer, Thorsten Hehn
No ratings yet
A Performance Study of MIMO Detectors: Christoph Windpassinger, Lutz Lampe, Robert F. H. Fischer, Thorsten Hehn
16 pages
Performance Comparison of Massive MIMO and Conventional MIMO Using Channel Parameters
No ratings yet
Performance Comparison of Massive MIMO and Conventional MIMO Using Channel Parameters
5 pages
Development of A DFT-Precoding Scheme For Spatially Multiplexed 4G Wireless Communication
No ratings yet
Development of A DFT-Precoding Scheme For Spatially Multiplexed 4G Wireless Communication
4 pages
Diversity and Spatial Multiplexing of MIMO Amplitude Detection Receivers
No ratings yet
Diversity and Spatial Multiplexing of MIMO Amplitude Detection Receivers
5 pages
MIMO Channel Estimation Using The LS and MMSE Algorithm
No ratings yet
MIMO Channel Estimation Using The LS and MMSE Algorithm
11 pages
MIMO Channel Estimation Using The LS and MMSE Algorithm: Mohammed Ali Mohammed MOQBEL, Wangdong, Al-Marhabi Zaid Ali
No ratings yet
MIMO Channel Estimation Using The LS and MMSE Algorithm: Mohammed Ali Mohammed MOQBEL, Wangdong, Al-Marhabi Zaid Ali
10 pages
InTech-Joint Ls Estimation and ML Detection For Flat Fading Mimo Channels PDF
No ratings yet
InTech-Joint Ls Estimation and ML Detection For Flat Fading Mimo Channels PDF
19 pages
Design and Simulation of 5G Massive MIMO Kernel Algorithm On SIMD Vector Processor
No ratings yet
Design and Simulation of 5G Massive MIMO Kernel Algorithm On SIMD Vector Processor
5 pages
Design of MIMO K-Best Detection Algorithm and Its FPGA Implementation
No ratings yet
Design of MIMO K-Best Detection Algorithm and Its FPGA Implementation
7 pages
Compressive Sensing Based Multi-User Detector For The Large-Scale SM-MIMO Uplink
No ratings yet
Compressive Sensing Based Multi-User Detector For The Large-Scale SM-MIMO Uplink
7 pages
Proceedings of Spie: An Improved Detection Algorithm For Massive MIMO System
No ratings yet
Proceedings of Spie: An Improved Detection Algorithm For Massive MIMO System
9 pages
Paper 14 PDF
No ratings yet
Paper 14 PDF
6 pages
System Architecture and Implementation of MIMO PDF
No ratings yet
System Architecture and Implementation of MIMO PDF
10 pages
Performance Evaluation of Max-D Precoding in Impulsive Noise For Train-To-Wayside Communications in Subway Tunnels
No ratings yet
Performance Evaluation of Max-D Precoding in Impulsive Noise For Train-To-Wayside Communications in Subway Tunnels
12 pages
Iterative QR Decomposition Architecture Using The Modified GramSchmidt Algorithm For MIMO Systems
No ratings yet
Iterative QR Decomposition Architecture Using The Modified GramSchmidt Algorithm For MIMO Systems
8 pages
57789-Điều văn bản-162562-1-10-20210525
No ratings yet
57789-Điều văn bản-162562-1-10-20210525
16 pages
Cascaded Channel Estimation For Large Intelligent Metasurface Assisted Massive MIMO
No ratings yet
Cascaded Channel Estimation For Large Intelligent Metasurface Assisted Massive MIMO
5 pages
Least Square Regressor Selection Based Detection For Uplink 5G Massive MIMO Systems
No ratings yet
Least Square Regressor Selection Based Detection For Uplink 5G Massive MIMO Systems
6 pages
pro.Document
No ratings yet
pro.Document
26 pages
Multi Antenna Systems
No ratings yet
Multi Antenna Systems
23 pages
Successive Interference Cancelling MIMO Signal Detection Method Using Dynamic Multi-Trace-Likelihood Ordering Detection
No ratings yet
Successive Interference Cancelling MIMO Signal Detection Method Using Dynamic Multi-Trace-Likelihood Ordering Detection
7 pages
Massive MIMO in 5G How Beamforming, Codebooks and Feedback Enable Larger Array
No ratings yet
Massive MIMO in 5G How Beamforming, Codebooks and Feedback Enable Larger Array
7 pages
Electronics: Massive MIMO Wireless Networks: An Overview
No ratings yet
Electronics: Massive MIMO Wireless Networks: An Overview
29 pages
08648511
No ratings yet
08648511
16 pages
SPH Decodin
No ratings yet
SPH Decodin
44 pages
A Survey On Deep-Learning Based Techniques For Modeling and Estimation of MassiveMIMO Channels
No ratings yet
A Survey On Deep-Learning Based Techniques For Modeling and Estimation of MassiveMIMO Channels
12 pages
Closed-Form_Bit_Error_Probability_of_ZF_Detection_for_OFDM-M-MIMO_Systems_Using_Effective_Noise_PDF
No ratings yet
Closed-Form_Bit_Error_Probability_of_ZF_Detection_for_OFDM-M-MIMO_Systems_Using_Effective_Noise_PDF
14 pages
towards_6G
No ratings yet
towards_6G
36 pages
2018 Smart Sorting in Massive MIMO Detection
No ratings yet
2018 Smart Sorting in Massive MIMO Detection
6 pages
MIMO Performance and Condition Number in
No ratings yet
MIMO Performance and Condition Number in
14 pages
Adaptive Neural Signal Detection
No ratings yet
Adaptive Neural Signal Detection
14 pages
Optimization of MIMO Detectors: Unleashing The Multiplexing Gain
No ratings yet
Optimization of MIMO Detectors: Unleashing The Multiplexing Gain
5 pages
A New Approach For Finite-Length Framework of Multi-Input Multi-Output Mmse
No ratings yet
A New Approach For Finite-Length Framework of Multi-Input Multi-Output Mmse
6 pages
Sphere Decoder For Massive MIMO Systems
No ratings yet
Sphere Decoder For Massive MIMO Systems
6 pages
Massive MIMO PDF
No ratings yet
Massive MIMO PDF
5 pages
Efficient DSP Circuit Architectures For Massive MIMO 1807.05882
No ratings yet
Efficient DSP Circuit Architectures For Massive MIMO 1807.05882
21 pages
Toward Extra Large-Scale MIMO New Channel Properties and Low-Cost Designs
No ratings yet
Toward Extra Large-Scale MIMO New Channel Properties and Low-Cost Designs
26 pages
Comparison of Different Diversity Techniques in MIMO Antennas
No ratings yet
Comparison of Different Diversity Techniques in MIMO Antennas
4 pages
Successive Interference Cancellation Via Rank-Reduced Maximum A Posteriori Detection
No ratings yet
Successive Interference Cancellation Via Rank-Reduced Maximum A Posteriori Detection
10 pages
MIMO Algorithm
No ratings yet
MIMO Algorithm
3 pages
Proposal Thesis Massive Mimo
No ratings yet
Proposal Thesis Massive Mimo
14 pages
Downlink Precoding With Mixed Statistical and Imperfect Instantaneous CSI For Massive MIMO Systems
No ratings yet
Downlink Precoding With Mixed Statistical and Imperfect Instantaneous CSI For Massive MIMO Systems
14 pages
Massive Mimo
100% (2)
Massive Mimo
21 pages
Reduced-Complexity LLL Algorithm For Lattice-Reduction-Aided MIMO Detection
No ratings yet
Reduced-Complexity LLL Algorithm For Lattice-Reduction-Aided MIMO Detection
5 pages
MIMO Channel
No ratings yet
MIMO Channel
5 pages
1.8 MIMO Channel PDF
No ratings yet
1.8 MIMO Channel PDF
5 pages
6G Wireless Communications and Mobile Networking
From Everand
6G Wireless Communications and Mobile Networking
PublishDrive
No ratings yet
signal_processing_in_5g_systems
No ratings yet
signal_processing_in_5g_systems
14 pages
coordinated_multipoint_transmission_in_5g
No ratings yet
coordinated_multipoint_transmission_in_5g
29 pages
08680765
No ratings yet
08680765
13 pages
08607077
No ratings yet
08607077
4 pages
08315053
No ratings yet
08315053
13 pages
Wang 2020
No ratings yet
Wang 2020
6 pages
Huang 2019
No ratings yet
Huang 2019
9 pages
He 2020
No ratings yet
He 2020
14 pages
Adaptive Transmission in Cellular Networks: Fixed-Rate Codes With Power Control Vs Physical Layer Rateless Codes
No ratings yet
Adaptive Transmission in Cellular Networks: Fixed-Rate Codes With Power Control Vs Physical Layer Rateless Codes
14 pages
Basnayaka 2015
No ratings yet
Basnayaka 2015
6 pages
0Rgxodwlrq&Odvvlilfdwlrqri0,022) '0 6Ljqdove/,Qghshqghqw&Rpsrqhqw $Qdo/Vlvdqg6Xssruw9Hfwru0Dfklqhv
No ratings yet
0Rgxodwlrq&Odvvlilfdwlrqri0,022) '0 6Ljqdove/,Qghshqghqw&Rpsrqhqw $Qdo/Vlvdqg6Xssruw9Hfwru0Dfklqhv
5 pages
Near-Optimal Pilot Signal Design For FDD Massive MIMO System: An Energy-Efficient Perspective
No ratings yet
Near-Optimal Pilot Signal Design For FDD Massive MIMO System: An Energy-Efficient Perspective
14 pages
Circuit-Aware Design of Energy-Efficient Massive MIMO Systems
No ratings yet
Circuit-Aware Design of Energy-Efficient Massive MIMO Systems
5 pages
Sensors 22 06500 v2
No ratings yet
Sensors 22 06500 v2
11 pages
10 1016@j DSP 2020 102656
No ratings yet
10 1016@j DSP 2020 102656
10 pages
Ma 2020
No ratings yet
Ma 2020
14 pages
Data-Driven-Based Beam Selection For Hybrid Beamforming in Ultra-Dense Networks
No ratings yet
Data-Driven-Based Beam Selection For Hybrid Beamforming in Ultra-Dense Networks
10 pages
Aruba Switching Fundamentals For Mobility, Rev. 16.41: Course Description
No ratings yet
Aruba Switching Fundamentals For Mobility, Rev. 16.41: Course Description
3 pages
Firewall - pfBlockerNG - Alerts PDF
No ratings yet
Firewall - pfBlockerNG - Alerts PDF
1 page
RS InstructorPPT Chapter4 Final
No ratings yet
RS InstructorPPT Chapter4 Final
56 pages
Cisco Ucs 9108 100g Intelligent Fabric Module Spec Sheet
No ratings yet
Cisco Ucs 9108 100g Intelligent Fabric Module Spec Sheet
15 pages
DS-QAZ1325G1T (C) Datasheet 20241023
No ratings yet
DS-QAZ1325G1T (C) Datasheet 20241023
5 pages
Name-Shubhankar Kesari Id No - 19BAJMC024 Submitted To-Rishabh Chaudhary
No ratings yet
Name-Shubhankar Kesari Id No - 19BAJMC024 Submitted To-Rishabh Chaudhary
5 pages
TDC
No ratings yet
TDC
243 pages
More Embedded Systems: FX of Phenoelit
No ratings yet
More Embedded Systems: FX of Phenoelit
45 pages
DPH 120se
0% (1)
DPH 120se
2 pages
Openstack Neutron and SDN
No ratings yet
Openstack Neutron and SDN
26 pages
GSM Protocol Stack
100% (1)
GSM Protocol Stack
3 pages
Net2 Reviewer-Midterm
No ratings yet
Net2 Reviewer-Midterm
3 pages
Xdoc - Tips Vmware Realtime Issues Question Amp Answers
No ratings yet
Xdoc - Tips Vmware Realtime Issues Question Amp Answers
20 pages
CS313 Computer Communication and Networking (Spring 2018) Assignment # 1
No ratings yet
CS313 Computer Communication and Networking (Spring 2018) Assignment # 1
5 pages
3) Assignment (CN)
No ratings yet
3) Assignment (CN)
5 pages
Multicast VLAN Commands
No ratings yet
Multicast VLAN Commands
8 pages
CN Unit 5 Notes
No ratings yet
CN Unit 5 Notes
19 pages
Gi-Fi Document
No ratings yet
Gi-Fi Document
5 pages
Cisco IPSec VPN Configuration - IPSec For VPN
No ratings yet
Cisco IPSec VPN Configuration - IPSec For VPN
19 pages
3G Fundamentals: Regional Telecom Training Centre, Trivandrum: 695 040
No ratings yet
3G Fundamentals: Regional Telecom Training Centre, Trivandrum: 695 040
29 pages
FortiNAC Quiz - Attempt Review
0% (1)
FortiNAC Quiz - Attempt Review
3 pages
LTE Events
No ratings yet
LTE Events
25 pages
5.3 TCP IP Protocol Architecture
No ratings yet
5.3 TCP IP Protocol Architecture
22 pages
Release Notes For The Catalyst 4900 Series Switch, Cisco IOS Release 12.2 (54) SG
No ratings yet
Release Notes For The Catalyst 4900 Series Switch, Cisco IOS Release 12.2 (54) SG
278 pages
Aim, Apparatus Required, Procedure Circuit Diagram / Tabulation / Program Execution & Output Viva-Voce Record Total
No ratings yet
Aim, Apparatus Required, Procedure Circuit Diagram / Tabulation / Program Execution & Output Viva-Voce Record Total
2 pages
NETGEAR Wireless-N Access Point WN802T v2 Reference Manual
No ratings yet
NETGEAR Wireless-N Access Point WN802T v2 Reference Manual
96 pages

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.