V .

ROHINI
22NM1A04F8
Unit-4
 SIGNAL COMPARISON: CORRELATION OF
FUNCTIONS
Concept of correlation:
The signals may be compared on the basis of similarity of waveforms.
If x1 (t) and x2(t) are two waveforms, then the waveform x1(t) contains an
amount C12x2(t) of that particular waveform x2(t) in the interval (t1, t2), where

Correlation of two functions or signals or waveforms is a measure of similarity between

those signals.
The correlation is of two types: cross correlation and autocorrelation.
The autocorrelation and cross correlation are defined separately for energy (or aperiodic)
signals and power (or periodic) signals.
 Correlation

Cross Correlation Auto Correlation

Power signal Energy signal Power signal

Energy signal
(Periodic Signals) (Aperiodic Signals) (Periodic Signals)
(Aperiodic Signals)
 Cross Correlation
• The cross correlation between two different waveforms or signals is a
measure of similarity or match or relatedness or coherence between one
signal and the time delayed version of another signal.

• Cross correlation between two signals indicates how much one signal is
related to the time delayed version of another signal.
 Cross correlation of Energy Signals
: x1(t) and x2(t) complex signals of finite energy

: x1(t) and x2(t) real signals .(4.1)

R12(τ) : finite value over the range of τ then x1(t) and x2(t) signals have some similarity
Cross correlation of .(4.2)
R12(τ) : Zero   orthogonal signals  orthogonal signals
is zero.

τ : delay parameter or scanning parameter or searching parameter

Two signals for which the cross correlation is zero for all values of T are called uncorrelated or
incoherent signals.
Properties of cross correlation function for energy signals

1. Cross correlation functions exhibit conjugate symmetry

..(4.3)
2. Two signals are said to be orthogonal over the entire time
interval.
..(4.4)

3. Correlation Theorem: The cross correlation of two energy

signals corresponds to the multiplication of the Fourier
transform of one signal by the complex conjugate of Fourier
transform of second signal.
..(4.5)
 Cross correlation of power (periodic) signals

• Cross correlation of two periodic signals x1(t) and x2(t)

or
..(4.6)
 Properties of cross correlation function for power
(periodic) signals
1. The Fourier transform of the cross correlation of two signals
is equal to the multiplication of Fourier transform of one
signal and complex conjugate of Fourier transform of other
signal. ..(4.7)
2. Signals are said to be orthogonal over the entire time interval
if
..(4.8)

3. cross correlation is not commutative,

..(4.9)
4. Cross correlation exhibits conjugate symmetry,
..(4.10)
 Autocorrelation
• Measure of match or similarity or relatedness or coherence
between a signal and its time delayed version. This means
that the autocorrelation function is a special form of cross
correlation function.

• It is defined as the correlation of a signal with itself.

Autocorrelation for energy signals

..(4.11)

..(4.12)
Properties of autocorrelation function
of energy signals
1. The autocorrelation function exhibits conjugate symmetry,
..(4.13)

Real part of R(τ) is an even function of τ

Imaginary part of R(τ) is an odd function of τ.
2. The value of autocorrelation function of an energy signal at
origin (i.e. at τ = 0) is equal to the total energy of that signal,
..(4.14)
3. If τ is increased in either direction, the autocorrelation R(τ)
reduces. As τ reduces autocorrelation, R(τ) increases and it is
maximum at t= 0, i.e. at the origin.
..(4.15)
 4. Autocorrelation theorem: Fourier transform of autocorrelation function
R(τ) yields energy density function of signal x(t) i.,e the autocorrelation
function R(τ) and energy spectral density function ψ(ω) of energy signal form
a Fourier tránsform pair.

..(4.16)

The above equation is a relation between Autocorrelation Function R(τ) and

Energy Spectral Density (ESD)
• Note: For a given function x(t), there is a unique
autocorrelation function but the reverse is not true.

• A given autocorrelation function may correspond to an

infinite variety of waveforms.
Autocorrelation function for power (periodic) signals

..(4.17)
Properties of autocorrelation function for power
signals
1. The autocorrelation function exhibits conjugate symmetry,
..(4.18)
..(4.19)

..(4.20)

..(4.21)

..(4.22)
 ENERGY DENSITY SPECTRUM

• Spectral density : It is the distribution of energy or power of a signal per

unit bandwidth as a function of frequency.

• Energy signals : Signals with finite energy and zero average power, i.e.
0<E< ꚙ and P = 0 are called energy signal.
Eg: aperiodic signals like pulse.

• Normalized energy: Simply energy of a signal x(t) is defined as the energy

dissipated by a voltage signal applied across 1-Ω resistor (or by a current
signal flowing through 1- Ω resistor). Mathematically,
Parseval's theorem for Energy signals : (Rayleigh's energy theorem)
Energy of a signal x(t) can be evaluated directly from its frequency spectrum X(ω)
without the knowledge of its time domain version, i.e. x(t).

..(4.23)
 Energy Spectral Density (ESD or ED)
• Gives the distribution of energy of a signal in the frequency
domain.

• For an energy signal, the total area under the spectral density
curve plotted as a function of frequency is equal to the total
energy of the signal. It is also called energy density spectrum
(ESD or ED).

• It is designated by ψ(ω) and given by

..(4.24)
..(4.25)
Properties of ESD

..(4.26)

From (4.25)

From (4.16)
 POWER DENSITY SPECTRUM
• Power signals: Signals with finite average power and infinite
energy, i.e. 0 < P<ꚙ and E = 0 are called power signals.
Eg: periodic signals.

• Average power: Average power dissipated by a voltage x(t)

applied across 1-Ω resistor (or by a current signal flowing
through 1-Ω resistor). Mathematically,
Parseval's power theorem: The power of a Signal in terms of its
Fourier series coefficients, i.e, in terms of the harmonic
components present in the signal.
..(4.27)
 Power spectral density (PSD)
• The distribution of average power of the signal in the
frequency domain is called power spectral density or power
density or power density spectrum (PSD or PD).

..(4.28)
 Properties of PSD

1. The area under the PSD function is equal to the average

power of that signal,

..(4.29)

2. The input and output PSDs of an LTI system are related as:

From (4.28)
 Properties of PSD
3. The autocorrelation function R(τ) and PSD S(ω) form a Fourier
transform pair,
From (4.21)

The above equation is a relation between Autocorrelation Function R(τ) and Power
Spectral Density (PSD).

Proof:
Comparision of ESD and PSD
 RELATION BETWEEN CONVOLUTION AND CORRELATION

• In Cross correlation, Multiply x1(t) with function x2(t) delayed

by τ sec. The area under the product curve is the cross
correlation between x1(t) and x2(t) at τ .

• The convolution of x1(t) and x2(t) at t= τ is obtained by folding

x2(t) backward about the vertical axis at the origin and taking
the area under the product curve of x1(t) and the folded
function x2(-t) displaced by τ.

• It, therefore, follows that the cross correlation of x1(t) and

x2(t) is the same as the convolution of x1(t) and x2(-t).
Eg: Find the cross correlation between triangular and gate functions shown

Solution:
Correlation of a triangular function
and gate function leads to triangular
function with double time period
 DETECTION OF PERIODIC SIGNALS IN THE PRESENCE OF NOISE
BY CORRELATION

• Detection of periodic signals masked by random noise is of great

importance.
• Applications:
– Detection of Radar and Sonar signals
– Detection of Periodic Components in brain waves
– Detection of cyclical components in ocean wave analysis
– Areas of Geophysics and including Metrology
• Correlation techniques provide a powerful too the solution of the above
problems.
• Both autocorrelation and cross correlation can be use the detection of a
periodic signal masked by noise.
• If s(t): periodic signal, n(t): noise signal, then the cross correlation function of s(t)
and n(t) is:
..(4.22)
 Detection by Autocorrelation:
s(t): periodic signal, n(t): Noise signal
Received signal is also periodic, y(t)=s(t)+n(t)
Auto correlation of y(t) is Ryy(τ), s(t) is Rss(τ), and n(t) is Rnn(τ)

..(4.23)

For large values of ' τ ', Ryy(τ) is essentially equal to Rss(τ). Therefore, Ryy(τ)
will exhibit a periodic nature at sufficiently large values of ' τ '.
Detection by Cross correlation:
Detection by cross correlation is much more effective than that by the
autocorrelation.
If the frequency is not known before hand  use autocorrelation technique.
If the frequency is known before hand  use cross correlation technique.

..(4.24)
 EXTRACTION OF A SIGNAL FROM NOISE BY FILTERING

• A signal masked by noise can be detected either by

correlation techniques or by filtering.
• The two techniques are equivalent.
• The correlation technique is a means of extraction of a given
signal in the time domain.
• Filtering achieves exactly the same results in the frequency
domain.
• Correlation in the time domain corresponds to filtering in the
frequency domain.
 Relationship between correlation and filtering:

Operation of cross correlation of x(t) and y(t) in the time domain is equivalent
to multiplication of the spectra X(ω) and Y(- ω) in the frequency domain.

The cross correlation function of signals x(t) and y(t) is the response of a system with
transfer function Y(-ω) [or the impulse response y(-t)] when the driving function is x(t).
Detection by filtering:

s(t): desired periodic signal component

n(t): random noise component
Received signal y(t) is:

The periodic component s(t) present in y(t) can be detected by

cross correlating y(t) with another periodic signal l(t) of the
same period as that of s(t).

Consider a system which has a unit impulse response l(-t) or

which has a transfer function L(-ω) where