ECTE301: Digital Signal Processing

Lecture 1
Introduction to Digital Signal Processing
Part A’s contents
Textbook
Lecture Topic Practice
Chapter
Lab 1
L1 Introduction to DSP 1, 2.1, 2.2
Tutorial 1
Lab 2
L2 Convolution and LTI Systems 2.3, 2.4
Tutorial 2
The z-transform: Tutorial 3
L3 3.1, 3.2, 3.3
Forward Transform and Properties Lab 3
The z-transform:
L4 3.4, 3.5 Tutorial 4
Inverse Transform and Applications
2.4, 2.5,
Random signals and Tutorial 4
L5 2.6, 12.1
Cross-correlation Sequences Report

L6 Revision of Weeks 1 to 6 - -

Digital Signal Processing S. L. Phung & A. Bouzerdoum 2/67

Lecture 1’s overview

 Introduction to digital signal processing (DSP).

 Advantages and disadvantages of DSP.

 DSP operations and applications.

 Sampling process from continuous to discrete-time signals.

 Types of signals, signal energy, signal power.

 Textbook reading: Chapters 1, 2.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 3/67

Lecture 1’s sequence

1.1 Overview of digital signal processing

1.2 Sampling continuous-time signals

1.3 Discrete-time signals

Digital Signal Processing S. L. Phung & A. Bouzerdoum 4/67

1.1 Overview of DSP
 A signal is any variable that carries information, which can be recorded,
conveyed, displayed, or manipulated.

 Examples of signals include:

● speech,
● sound,
● music,
● images
● biomedical signals,
● radar signals,
● sonar signals.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 5/67

Example signals: Speech
0.5

0.4

0.3

0.2
Amplitude

0.1

-0.1

-0.2

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Time (s)
A speech signal generated by Microsoft Speech SDK: ‘Hello, World’
Digital Signal Processing S. L. Phung & A. Bouzerdoum 6/67
Example signals: Speech

Xinhua’s English AI Anchor (11/2018) - video

Digital Signal Processing S. L. Phung & A. Bouzerdoum 7/67
Example signals: EEG

 Electroencephalography (EEG) is the recording of the

brain’s electrical activity along the scalp.
 It is used for designing brain-computer interfaces,
detecting seizures, monitoring the effects of sedatives.

One second of EEG signal

amplitude

time (s)

Digital Signal Processing S. L. Phung & A. Bouzerdoum 8/67

Example signals: Depth image
x

meter
f(x, y)
y

SwissRanger
SR4000
depth camera

Each pixel value is the distance from a scene point to the camera.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 9/67

Example signals: Thermal image
oC

Recorded by NEC InfRec R300 thermal camera

UOW Centre for Signal and Information Processing
Digital Signal Processing S. L. Phung & A. Bouzerdoum 10/67
Example signals: Thermal image

Detecting flu Analysing house heat

Sources: www.tempsensornews.com, www.saniglow.com

Digital Signal Processing S. L. Phung & A. Bouzerdoum 11/67

Digital Signal Processing

 DSP is concerned with

● the digital representation of signals,
● the synthesis, analysis, and modification of signals, and
● the extraction of information from signals.

 DSP aims to:

● remove interference or noise from signals,
● obtain and analyse the spectrum of a signal,
● transform the signal into a more suitable format.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 12/67

Advantages of DSP
DSP offers many advantages over analog signal processing:

 Flexibility: Different DSP algorithms can be implemented by software

running on a computer or digital processor.

 Reliability: DSP leads to very stable processing. It requires only

additions, multiplications, and delays, which are less affected by
external factors, e.g. temperature.

 Adaptability: DSP operations can be modified in real-time, by changing

the program.

 Economy: DSP has lower cost due to VLSI technology.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 13/67

Advantages of DSP

adder
× +
delay
delay

multiplier
×

An analog low-pass A digital low-pass filter consisting of

RC filter delays, adders, multipliers

Digital Signal Processing S. L. Phung & A. Bouzerdoum 14/67

Disadvantages of DSP

However, DSP has some disadvantages.

 Speed: DSP algorithms can be slow, especially for signals with high
frequencies or high bandwidths.

 Finite word length: Each number is represented using a finite number of

bits.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 15/67

Key DSP operations

 Convolution

 Correlation

 Digital filtering

 Modulation

 Discrete transforms (DFT, FFT, DCT, …)

Digital Signal Processing S. L. Phung & A. Bouzerdoum 16/67

DSP Systems

 This diagram illustrates a typical DSP system.

 ADC: analog to digital converter.

 DAC: digital to analog converter.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 17/67

DSP Systems

Digital Signal Processing S. L. Phung & A. Bouzerdoum 18/67

DSP Systems

Virtex-6 DSP Development Kit

UOW Centre for Information and Signal Processing Lab
Digital Signal Processing S. L. Phung & A. Bouzerdoum 19/67
DSP Applications

 DSP plays a major role in many applications.

 Image Processing
● image and video compression
Image processing applications
● machine vision
Car tracking
● pattern recognition Gender recognition
● facsimile 3-D range imaging
● satellite weather map Atomic force microscopy imaging
Robot soccer
● animation

Digital Signal Processing S. L. Phung & A. Bouzerdoum 20/67

DSP Applications
 Telecommunications
● adaptive equalization and echo cancellation
● spread spectrum
● video conferencing
● data communication

 Biomedical Engineering
● MRI, X-ray, ultrasound scanners
● EEG brain mapping
● ECG analysis

Digital Signal Processing S. L. Phung & A. Bouzerdoum 21/67

DSP Applications
 Instrumentation and Control
● spectrum analysis and noise reduction
● position and rate control
 Speech and Audio
● speech synthesis and recognition
● text to speech conversion
● digital audio
● equalization
 Military
● secure communication
● radar and sonar processing
● missile guidance
Digital Signal Processing S. L. Phung & A. Bouzerdoum 22/67
DSP apps: Lie detector by MRI brain scan
Response to a question Response to another question

Magnetic resonance imaging (MRI) Prefrontal cortex becomes highly active

scan shows no unusual activity in (shown in hot colours).
prefrontal cortex. Source: IEEE Spectrum.
Digital Signal Processing S. L. Phung & A. Bouzerdoum 23/67
DSP apps: Ground Penetrating Radar

Using GPR to assess rail-track ballast (UOW CSIP research)

Digital Signal Processing S. L. Phung & A. Bouzerdoum 30/67
DSP apps: Ground Penetrating Radar

Block diagram of a ground-penetrating radar

Digital Signal Processing S. L. Phung & A. Bouzerdoum 31/67
DSP apps: Ground Penetrating Radar
with clay
clean
with coal

GRP signals from 3 types of underground material.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 32/67
DSP apps: Image Compression (JPEG & DCT)

Uncompressed color image: 5000 × 2500 pixels

File size = 37,500,054 Bytes, Bit rate = 24 bits per pixel (bpp)

Digital Signal Processing S. L. Phung & A. Bouzerdoum 34/67

DSP Apps: Doppler radar
Human motions Radar Doppler signatures

Digital Signal Processing S. L. Phung & A. Bouzerdoum 38/67

Lecture 1’s sequence

1.1 Overview of digital signal processing

1.2 Sampling continuous-time signals

1.3 Discrete-time signals

Digital Signal Processing S. L. Phung & A. Bouzerdoum 39/67

1.2 Sampling of continuous-time signals
 Consider a continuous-time signal xa(t).

 A discrete-time signal x(n) is obtained by periodic sampling.

 If T be the sampling period, or the interval between successive samples:

 The sampling rate (or sampling frequency) is:

 The relationship between variables t and n:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 40/67

Sampling of continuous-time signals

Sampling
Sampling period

 ( ) is obtained by replacing with in the formula of ( ).

 ( ) is obtained by replacing with / in the formula of ( ).

Digital Signal Processing S. L. Phung & A. Bouzerdoum 41/67

Sampling a sinusoidal signal
 Consider a signal with an analog frequency of F (Hz):

 Suppose that the signal is sampled at a rate of Fs:

 The resulting discrete-time signal is:

 f is the normalized frequency:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 42/67

Sampling a sinusoidal signal

 Angular analog frequency Ω is defined as

 Angular normalized frequency ω is defined as

 The relationship between angular analog and normalized frequencies:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 43/67

Sampling a sinusoidal signal
 Example 1.1: Consider an analog signal

 Analog frequency: F = 20 Hz.

 Suppose that the sampling rate is Fs = 100 Hz.

 The digital signal is

 The normalized frequency

Digital Signal Processing S. L. Phung & A. Bouzerdoum 44/67

The sampling theorem
A continuous-time signal x(t) with frequencies of no more than Fmax
can be reconstructed exactly from its samples x(n) = x(nTs) if
the sampling rate Fs is at least 2 × .

Significance:
 If the maximum signal frequency is max, the sampling rate should be
at least the Nyquist rate:
= 2 × max.

 If the sampling rate is , the maximum frequency in the signal must

not exceed the folding frequency to avoid aliasing:

Folding frequency = .
Digital Signal Processing S. L. Phung & A. Bouzerdoum 45/67
Frequency aliasing
 Example 1.2: Consider the two analog signals

 The analog frequencies are: F1 = 20 Hz, F2 = 80 Hz.

 Sampling these two signals at a rate Fs = 100 Hz gives

 In this case, x1(n) and x2(n) are two identical discrete-time signals.

 Frequency aliasing: Two different analog frequencies appear as the

same normalized frequency.
Digital Signal Processing S. L. Phung & A. Bouzerdoum 46/67
Frequency aliasing
Sampling rate: = 100 Hz
Folding frequency: 50 Hz
Sampled signal

1( ): 1 = 20 Hz 2( ): 2 = 80 Hz

Frequency aliasing occurs when: .

Digital Signal Processing S. L. Phung & A. Bouzerdoum 47/67
Frequency aliasing

and

 Because cos(. ) has a period of 2π, any normalized frequency f can be

mapped to −0.5 ≤ ≤ 0.5 by subtracting an integer.

 Discrete signal = cos 1.6 has a normalized frequency: = 0.8.

 = 0.8 is equivalent to an alias normalized frequency of:

= − 1 = 0.8 – 1 = −0.2 (or 0.2).

Digital Signal Processing S. L. Phung & A. Bouzerdoum 48/67

Frequency aliasing

 In analog domain, let 1 be any frequency in the fundamental range:

− ≤ 1 ≤
2 2

 Then, all following frequencies are an alias of 1:

± 1,

where is an arbitrary integer.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 49/67

Lecture 1’s sequence

1.1 Overview of digital signal processing

1.2 Sampling continuous-time signals

1.3 Discrete-time signals

Digital Signal Processing S. L. Phung & A. Bouzerdoum 50/67

Discrete-time signals
 A signal is a function of one or more independent variables.

 A discrete-time signal is represented as a sequence of numbers:

 x(n) denotes the n-th sample of the sequence.

1
0.5
0.25 0.25
0
-2 -1 0 1 2 n
Non-listed elements are assumed to be zero.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 51/67

Elementary discrete-time signals

Next, we’ll study some basic discrete-time signals.

S1 S2 S3
Unit sample Unit step Unit ramp

S4 S5 S6
Sinusoidal Exponential Complex
sequence exponential

Digital Signal Processing S. L. Phung & A. Bouzerdoum 52/67

S1) Unit sample
 The unit-sample sequence δ(n) is defined as

-2 -1 0 1 2 n

 The unit-sample sequence is sometimes called the impulse sequence.

 Property: An arbitrary sequence x(n) can be expressed as

Digital Signal Processing S. L. Phung & A. Bouzerdoum 53/67

S2) Unit step
 The unit-step sequence u(n) is defined as

-2 -1 0 1 2 n

 Relationship between u(n) and unit-sample sequence δ(n):

Digital Signal Processing S. L. Phung & A. Bouzerdoum 54/67

S3) Unit ramp

 The unit-ramp sequence r(n) is defined as

r(n)
3
2
1

-2 -1 0 1 2 3 … n

Digital Signal Processing S. L. Phung & A. Bouzerdoum 55/67

S4) Sinusoidal sequence
 A discrete-time sinusoidal signal has the form of

● A is the amplitude.
● ω is the normalized angular frequency.
● f is the normalized frequency.
● Φ is the phase.

 ω and f are related as

 A discrete-time sinusoid is periodic if its frequency f is a rational number:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 56/67

Sinusoidal sequence

Digital Signal Processing S. L. Phung & A. Bouzerdoum 57/67

S5) Exponential sequence
 The discrete-time exponential signal is defined as

where α is a fixed number.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 58/67

S6) Complex exponential sequence
 If α is a complex number, we can express it as

 Then, signal x(n) can be written as

 Different components of the signal:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 59/67

Types of discrete-time signals

Next, we’ll study some categories

of discrete-time signals.

T1 T2 T3
Energy Power Periodic
signals signals signals

T4 T5 T6
Even Odd Conjugate
signals signals even/odd signals

Digital Signal Processing S. L. Phung & A. Bouzerdoum 60/67

T1) Energy signal
 The energy E of a discrete-time signal is defined as

 Signal x(n) is an energy signal if its energy is a finite number.

 Example 1.3: Find the energy of the sequence

Answer:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 61/67

T2) Power signals
 The average power P of a discrete-time signal is defined as

 Signal x(n) is a power signal if its power is a finite number.

 Example 1.4: Find the energy of the unit-step signal u(n).

Answer:

The unit-step signal is a power signal.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 62/67
T3) Periodic signals

 A signal x(n) is periodic with a period of N (where N ≠ 0) if

 A bounded periodic signal x(n) is a power signal:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 63/67

T4) Even signals
 A real-valued signal x(n) is called even (symmetric) if

 Three examples of even sequences:

a)

b)

c)

-2 -1 0 1 2 … n

Digital Signal Processing S. L. Phung & A. Bouzerdoum 64/67

T5) Odd signals
 A signal x(n) is called odd (or anti-symmetric) if

 Three examples of odd sequences:

a)

b)

c)

-2 -1 0 1 2 … n

Digital Signal Processing S. L. Phung & A. Bouzerdoum 65/67

T6) Conjugate even or odd

 Let * denote complex conjugate. For example,

 A complex-valued signal x(n) is called conjugate

even (or Hermitian symmetric) if

 A complex-valued signal x(n) is called conjugate

odd (or Hermitian anti-symmetric) if
Charles Hermite (1822-1901)
French mathematician
Digital Signal Processing S. L. Phung & A. Bouzerdoum 66/67
Lecture 1’s summary
Key aspects of this lecture:

 Overview of DSP and applications.

 The sampling process: the relationship between xa(t) and x(n).

 Definitions of normalized frequency f, sampling frequency Fs, folding

frequency, Nyquist rate, and Nyquist frequency.

 Common discrete-time signals: unit-sample, unit-step, unit-ramp,

sinusoidal, and exponential.

 Periodic, even, and odd signals, signal energy, signal power.

 Practice: Tutorial 1, Labs 1 and 2.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 67/67