Signals and Systems

ELESB16522
Lecture
0 (101) The Big Picture
Signals (sound / light / voltage) are information-bearing variations

Analog signals are real-world signals

They are defined at every point in time and may take an infinite number of
possible amplitudes

Word “away“
Courtesy Michael Noonan, Ph.D. Canisus College
Analog signals are not well-suited to processing by computer

They may be converted to digital signals

through sampling and quantization Word “ooo“
Digital signals are defined only at sampling points, and may take only a
finite number of discrete amplitudes

After processing, a digital signal is converted back into an analog signal

A system
analyzes, combines, modifies, records, and plays back signals.

Image compression system

less memory space → less time to transmit through Internet

Speech recognition system

understanding human speech automatically

For digital signals, with specialized hardware and software, the operations
can occur at lightning speeds.
Advantages of Digital Systems vs. Analog Systems

Analog systems:
circuits → hardware → properties vary within manufacturer’s tolerance &
with temperature

Digital systems:
 software → behaves in predictable, repeatable way
 much less affected by noise
 smaller and consume less power
 flexible → change a few lines in a program
 vs. build new circuits from scratch
1.1 SIGNALS AND SYSTEMS

 Speech synthesis
 Edge detection in images  Echo cancellation
 Digital signal and image filtering  Cochlear implants
 Seismic analysis  Antilock brakes
 Text recognition  Signal and image compression
 Speech recognition  Noise reduction
 Magnetic resonance image (MRI) scans  Companding
 Music synthesis  High definition television (HDTV)
 Bar code readers  Digital audio
 Sonar processing  Encryption
 Satellite image analysis  Motor control
 Digital mapping  Remote medical monitoring
 Cellular telephones  Smart appliances
 Digital cameras  Home security
 Detection of narcotics and explosives  High speed modems
 TOPICS
Analog Signals vs. Digital Signals

Basic Steps in A/D Conversion

Basic Steps in D/A Conversion

Signal and Its Spectrum

Basic Concepts of Filtering

Applications of DSP
Easy Reading

Many Code Examples

Lots of Example
Problems
(with Full Solutions)
 Best Web Support

(including LOOOTS of
examples for solving basic
problems!)

https://dspfirst.gatech.edu/

Additional BONUS

Many Code Examples

In Lithuanian

Chapters to Read:
2 – 6 & 12
 Review of Continuous-Time
Signals and Systems – Chapter 1

Analog Filters – Chapter 2

Fundamentals of Continuous-Time System Analysis:

Chapters 9 – 14, 18

Laplace Transforms | Frequency Response | Fourier Series

Fourier Transforms | Impulse Response | Convolution |
 GRADING SYSTEM

Mid-term 1 40%

Mid-term 2 40%

Homework 1 & 2 6%

Quiz 4%

Laboratory 10%
Studentai pagal modulio kortelėje nustatytus
lankomumo reikalavimus privalo lankyti iki
80 procentų laboratorinių darbų.

Išimtinais atvejais, pažangiam, atsiskaičiusiam visus tarpinius

atsiskaitymus studentui dėstytojo sprendimu gali būti sudaryta
galimybė atsiskaityti už praleistas pratybas.

Atsiskaitymai organizuojami per studijų grafike numatytą tarpinių

atsiskaitymų savaitę dėstytojo paskirtu laiku ir būdu, skiriant
papildomas užduotis ir atsiskaitant už jas.
Digital Filtering
Signal Frequency (Spectrum) Analysis
Digital Crossover Audio System
Interference Cancellation in ECG
Speech Coding and Compression
CD Recording System
Starts Explaining
The “Real Stuff”…
MP3 coder
 MP3 coder

MP3 coder is a frequency domain coder where segments of 576 time

domain samples are transformed into 576 frequency domain
samples.

The available bits are then allocated non-uniformly among various

frequency components depending on the just noticeable distortion (JND)
levels at different frequencies calculated from the psychoacoustic model.

It consists of the following subsystems. The filter bank is a uniformly

spaced filter bank consisting of 32 equal-bandwidth poly-phase filters of
length 512 taps each.

Poly-phase filters divide the bandwidth into N bands (where N is a power of 2) with
the signal in each sub-band down-sampled by a factor of N.

The filters are said to be critically sampled and have excellent anti-aliasing
property as they can achieve perfect reconstruction of the original signal.
Vibration Signature Analysis for Defected Gear Tooth – 1
Vibration Signature Analysis for Defected Gear Tooth – 2
Vibration Signature Analysis for Defected Gear Tooth – 3
Digital Image Enhancement
1.2 ANALOG-TO-DIGITAL AND DIGITAL-TO-ANALOG
CONVERSIONS

Sights / sounds are analog signals

Sensors:
microphones / CCD chips / thermistors / accelerometers / pressure sensors
A/D Conversion 1 – Sample and Hold

Analog signal is sampled (at regular intervals)

At each sampling point, signal value held steady until next sampling point.
A/D Conversion 2 – Sample and Hold

Sampling must be fast enough to capture most rapid changes in the signal.

If sampling is too slow, important signal characteristics can be lost →

aliasing
A/D Conversion 3 – Quantization & Digitization

After sample acquisition, analog values are quantized and digitized.

Converter selects the quantization level that approximates the sample-
and-hold value and then assigns a binary digital code.
A/D Conversion 4 – Quantization & Digitization

Digital values do not coincide with analog signal values at sampling points.

Perfect agreement is impossible due to the way digital numbers are stored
in PC.

Number of bits per memory location limits the accuracy possible when an
analog value is converted to digital.
A/D Conversion 5 – Important Characteristics

After A/D conversion, signal has two important characteristics:

1) Number of permitted signal values is limited by number of bits used by

PC.

2) Signal has values only at sampling instants.

D/A Conversion 1 – Digital Signal

Convert each digital code into an analog voltage level proportional to the
size of the digital number.
D/A Conversion 2 – Zero Order Hold Signal

This analog voltage level is held steady for the duration of a sampling
period.
D/A Conversion 3 – Recovered Analog Signal

To recover the original analog signal, we need to smooth the ZOH signal.

This introduces a time shift as a side effect.

All Steps in A/D and D/A Conversions
1.3 Digital Signals and Their Spectra

Signals themselves, whether analog or digital, are defined in the time

domain.

In this domain, signal behavior is recorded as time proceeds.

Does the signal increase in amplitude over time, or perhaps cycle up and
down?

Is its amplitude large or small?

Signals, however, may have important characteristics that are not obvious
in the time domain.

These characteristics must be identified in the frequency domain.

Frequency domain information about a signal is presented using a plot

called a spectrum.

The spectrum records the frequencies of all elements of the signal.

C note
CEG chord
Word “away“
Vowel “ooo“
Orca
256 Hz
Black and white image
Four frames from high-speed video sequence - 1/4
Four frames from high-speed video sequence - 2/4
Four frames from high-speed video sequence - 3/4
Four frames from high-speed video sequence - 4/4
1.4 Digital Filtering

Frequency domain is the best domain to see the effects of filtering a signal.

A filter passes the signal frequencies that lie in its pass band, and
attenuates those that lie in the stop band.

E.g., low pass filters (LPF) allow low frequency signals to pass and block
high frequency signals. LP filters tend to smooth signals, while high pass
filters (HPF) emphasize sharp signal changes.

Band pass filters (BPF) pass a particular band of frequencies; band stop
filters pass all frequencies except those in a particular band.
Low pass filter to extract bass voices
Low pass filter to extract soprano voices
Band pass filter to extract alto voices
Spectrum for CEG chord (250-2500 Hz) CEG chord
Low pass filtering a piano chord - 1/3

Filt CEG
Low pass filtering a piano chord - 2/3

Filt CE
Low pass filtering a piano chord - 3/3

Filt C
1.5 Speech, Music, Images and More

DSP has many exciting applications in a wide range of fields, and the
number of applications grows every year.

What the applications have in common is that they are defined primarily by
software.

Among the most accessible are those applications that operate on familiar
1D signals like speech or music, or 2D signals like images.
Large vocabularies of continuous speech
can be recognized using information
from digital speech signals.

Most often, speech recognition methods are based on a spectral analysis of

speech sounds.

Though this analysis can be time-consuming, fast DSP hardware makes real-
time recognition a reality.

The larger the population of users of a speech recognizer, the more variable
the speech and the more difficult the task.
Greater vocabulary also makes discrimination between words more
challenging.

Automatic voice recognition for radio control makes hands-free radio use
possible in jet cockpits, and similar technology allows hands-free use of
cellular phones in automobiles.

In complementary technology, advanced speech synthesis programs

generate naturally sounding speech for telephone information messages
(“The number 555-1234 has been changed; the new number is 555-4321”).
DSP also makes contributions in the arena of music and other sounds.

Old recordings of music and audio tracks can be cleaned up by removing

clicking and buzzing sounds in the background.

Also, the sounds of musical instruments can be closely copied for electronic
synthesizers.

Combinations of actual recordings and math models produce high quality

synthesis for many instruments, including piano, violin, and flute.
DSP can be applied to animal sounds as well.

Whale songs have been analyzed to study whale communications patterns.

Fantastic effects are possible when DSP is applied to imagery.

Images can be combined,

as when Forrest Gump was digitally
added to a film of President Nixon.
Background speckling can be removed by averaging several pictures of the
same scene.

Movement can be detected by subtracting one image from another of the

same scene, highlighting the places where the two images differ.

Morphing can allow one image to change smoothly into another, a technique
used by Star Trek’s changeling, Odo.
When several digital views of an object are available, they can be combined
mathematically to create a 3D view of the object, a technology that is
exploited in medical images like MRIs and CT scans.

These images are normally assessed and interpreted manually by doctors.

Visual inspection and robot vision systems, on the other hand, rely on
automatic identification of objects.

This is accomplished through detection of object edges coupled with

sophisticated pattern recognition schemes.