ESS206

Fundamentals of Signal Processing and Data Analysis

Signal & System

韩 鹏 (Han Peng)
Email: hanp@sustc.edu.cn
Tel: 0755-88015515
Office: College of Science, Room E3136
Office hour: Friday 16:00-18:00

Lecture 1, February 15, 2023 2023 Signal Processing

Digital Signal Processors In Products
IP phone

Smart power meters

IP camera
Consumer audio
Video conferencing DSL modems
In-car entertainment

Mixing
board
Amp
Tablets
Communications
Pro-audio Multimedia
Topics of this lecture

• What is a signal ?
• What is signal processing ?
• Basic signals
• Representation of signals
• Representation of systems
 What is a signal ?
A signal is the carrier of information.
Electrocardiogram contains information related to sickness.
• medical diagnosis
Voice contains information about who is speaking and what is spoken
• Speech recognition
• Speaker recognition
Radar signal contains information about the target.
To avoid accidents for airport navigation
To defend the country from enemy's attack
Biometric images contain information about authentication.
Face images
Fingerprints
 Seismic wave
Electrocardiogram (ECG)
 Morse code

……
Samuel Morse

See Wii
You Csy
Tonight Xsrmklx
注：本页内容不构成任何投资建议。
 Standard normal distribution
1 − 12 x2
 ( x) = e
2

C = 2 r
360 180
Cn = n • 2r sin( ) = 2(n sin ）r
2n n
C = lim Cn
n →

180
 = lim(n sin )
n → n

S r  Pr
= = 1
S 4 P e = lim(1 + ) n
n → n
Monte Carlo
Discretization of “Space” and “Time”
A signal is the carrier of
information.

A signal is anything that serves to

indicate, warn, direct, command, or
the like, as a light, a gesture, an
act, etc.
 Matlab toolbox
• We can check functions contained in the “signal
processing toolbox” of Matlab using the following
command:

demo 'toolbox' ‘signal’

• Examples include:
–Signal generation, noise cleaning, peak detection, etc.
 What is signal processing ?
• The purpose of signal processing is to make it easier to extract useful
information and to make correct decisions. For this purpose, we have
– Signal enhancement
– Signal restoration
– Signal reconstruction
– Signal synthesis
– Signal estimation and prediction

• After signal processing, we can extract more important information,

and recognize, aware, and make decisions.
 Signal Pattern
Reasoning
Pr ocessing Recognition

Earthquake、
Run！
Signal Enhancement

• Signal amplification in
radio, TV, human ears
• Image enhancement in
image processing
Signal Restoration

• Restore a signal
corrupted by noise or
useless background
– Tuning in radio and TV
– Noise cleaning
– Pick up the voice of a
certain person in a
party
Restoration of face images
Signal Estimation/Prediction

• Estimate the future

values from observed
data
– Stock prediction
– Electrical Demand
prediction
– Image compression
– Adaptive signal
processing
 DSP Scenario
• Modern systems generally…
– get a continuous-time signal from a sensor
– a cont.-time system modifies the signal
– an “analog-to-digital converter” (ADC or A-to-D) sample the signal to
create a discrete-time signal … a “stream of numbers”
– A discrete-time system to do the processing
– and then (if desired) convert back to analog (not shown here)
x(t) x[n] y[n] y(t)

t n n t
Sensor
Analog DAC
ADC DSP
Electronics
Electrical
Electrical
Physical Electrical C-T Electrical Electrical D-T Electrical C-T
C-T C-T System C-T D-T D-T Signal
System
Signal Signal Signal Signal Signal 2/13
Discrete-Time (D-T) Signals
A discrete time signal is a sequence of numbers indexed by integers
Example: x[n] n = …, - 3, -2, -1, 0, 1, 2, 3, …

Brackets indicate discrete-time signal. A stem plot emphasizes that the

Use parentheses to indicate a C-T signal. signal does not exist
in-between integer n values

x[n]

n
Infinite Duration & Finite Duration D-T Signals
Most general case: the number of non-zero values in the signal is infinite.
“Infinite Duration” “Goes on forever”

“Doubly Infinite” 
x[n] = , 4,2,0,1,3,7,12,5.3,2.1,


“Singly Infinite” 
x[n] = 7,12,5.3,2.1,

 This means x[n] = 0 n  0


x[n] = , 4,2,0,1,3,7

 This means x[n] = 0 n 0

Special case: the number of non-zero values in the signal is finite.

“Finite Duration” “Starts and Stops”


x[n] = 7,12,5.3,2.1

 0 n  0
This means x[n] = 
0 n  3

This is “a finite-duration sequence of length 4”

Some Elementary DT Signals
Most real-world signals are quite complicated looking… but there are some
simple signals that are good for learning but also do show up sometimes in the
real world!

“Unit Sample Sequence” Also called “Unit Impulse”

or “Delta Function”
1, n = 0
 [n] =  or “Delta Sequence”
0, n  0

Two Important Properties

 [n] = 1
n=−

 x[n] [n − n ] = x[n ]
0 0
n=−
Sifting Property
“Unit Step Sequence”

0, n  0
u[n] = 
1, n  0

“Unit Ramp Sequence”

ur [n] or r[n]

0, n  0
ur[n] = 
n, n  0

ur [n] = nu[n]
 “Real Exponential Signal”

x[n] = a n n, a

Decays if |a| < 1 Explodes if |a| > 1

Monotonic
if a > 0

Oscillates
if a < 0
 “Complex Exponential Signal”
Includes Complex Sinusoids (r = 1)

x[n] = a n n, a

Im r
a = re j , r  0 
Re

x[n] = a n = (re j ) = r n e j n = r n (cos( n) + j sin( n) )

n

Radius r controls “decay rate”

• r < 1 (inside UC) decays Im{a}
• r > 1 (outside UC) “explodes”
Unit Circle
Angle θ controls “oscillation rate”
• θ = 0 (on positive Re) no oscillation Re{a}
• θ ≠ 0 (off positive Re) oscillation
Classification of DT Signals

Many times what we can or can not do with a signal depends on some
characteristic of that signal. Thus, it is helpful to classify DT signals into
some typical subsets to help keep track of all this…

Some Categories of DT Signals

• Infinite Duration vs Finite Duration
• We’ve seen that one already!
• Energy Signals vs. Power Signals
• We’ll discuss that next
• Periodic Signals vs. Aperiodic Signals
• Symmetric (even) Signals vs. Antisymmtric (odd)Signals
Energy Signals vs. Power Signals
Physically, defining power and energy for DT signals is not thatwell
defined… after all these are just a stream of numbers in a computer!
However, we can still do this from a mathematical sense that mimics the
definitions used for physical CT signals.

E =  x[n]
2
Energy of DT Signal:
n=−

Obviously, E > 0 for any non-zero signal!

But, since we are adding (possibly) infinitely many positive values it is
possible that for some signals the value of E is infinite.
This leads to our first classification:
A signal is said to be an Energy Signal if its energy is finite (i.e., 0 < E < ∞)
Clearly for this to happen we need |x[n]| to decay “fast enough” as n → 
Examples of Energy Signals Examples of Non-Energy Signals
1. Any finite-duration signal 1. Sinusoid
2. Decaying exponential: an , with a  1 2. Unit Step
For those signals that are NOT energy signals we can characterize their Power

Drawing from the physical idea that power is the rate of doing work (i.e., the
amount of energy consumed/delivered per unit time) we can define similar
ideas for DT signals.
N
Energy of DT Signal over finite interval of “length” 2N+1: EN =  x[n]
2

n=− N


EN 1
= =
2
Then… Power of a DT Signal is: P lim lim x[n]
N → 2N + 1 N → 2N + 1
n=− N

Energy per 2N+1 samples

Note that what you get for P depends on a “race” between EN and 2N+1:
• If EN grows but eventually “levels out” (then x[n] is an energy signal)
we get that P = 0. So only energy signals have P = 0.
• If EN grows and never levels out, but does not grow faster than 2N+1
then P is finite and nonzero. Not an energy signal, called power signal.
• If EN grows and never levels out, but does grow faster than 2N+1 then P
is infinite. Neither energy nor power signal.
 Classification of Signals on Energy & Power
N
EN
EN =  x[n] 2
E = lim EN
N→
P = lim
N → 2N +1
n=− N

Energy Signal EN → E < ∞ P=0

Power Signal EN → ∞ 0<P<∞

but not faster than 2N+1

Neither EN → ∞ P →∞
but faster than 2N+1
Symmetric (Even) vs. Anti-Symmetric (Odd)
A DT signal is symmetric (even symmetric) if x[−n] = x[n]
x(t)
Illustrated here
using CT because it
is easier to draw!
-to to t

Same value for all to

A DT signal is anti-symmetric (odd symmetric) if x[−n] = −x[n]

x(t)
Note what happens at
n = 0…
t x[−0] = −x[0]
-to to
 x[0] = −x[0]
That can only be true
Negative values for all to if x[0] = 0!!!
Why do we care about symmetry??? Because in Fourier analysis of signals
we are decomposing a signal into sines (odd) and cosines (even) and it turns
out that if the signal is even (odd) then we can build it using only cosines
(sines).
So if there is some form of symmetry, we can exploit it to make our
analysis job easier!!

Even we have a general signal we can sometimes make use of the

simplifications that come from symmetry:

Any signal x[n] can be broken into a sum of an even part and odd part like
this
x[n] = xe [n] + xo [n]
xe[n] =
1
x[n] + x[−n] xo[n] =
1
x[n] −x[−n]
2 2

Clearly, these two add to x[n]… and it is easy to verify they are even and odd.
 Simple Manipulation of DT Signals
Time Shift
A given signal x[n] is shifted in time by
replacing n by n – k, where k is an integer.

Don’t forget there is a

negative already in there!!

• If k is positive then the shift is to the

right… which is a “delay”

• If k is negative then the shift is to the

left… which is an “advance”
Time Scaling

A given signal x[n] is scaled in time by

replacing n by µn, where µ must be an integer.
y[n] = x[n]
An easy way to help visualize this is to let m = µn, let n run through the
integers and find the resulting values of m… those are the indices of the
samples you’ll have “keep”
You can see why this process is
also called “Down Sampling”!
n: 0 1 2 3 …
m: 0 1 2 3 4 5 6 7 8 9 …
µ= 3 x[m]

-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 m

y[n] Note

-2 -1 0 1 2 n
 Then…. We re-imagine the new signal on the same axis as the original:

x[n]

-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 n

y[n]

-2 -1 0 1 2 n

Addition, Multiplication & Amplitude Scaling

Adding two sequences, multiplying two sequences, or multiplying a sequence

by a number are defined in the obvious way.

34/25
DT Samples…. What CT Sinusoid did they come from????

t (sec)

They could have come from this blue one… f s F2 f 0

t (sec)
But…They could have come from this RED one!!!
Thus… if we
want to be able
to tell these two
apart we need
to sample
t (sec) faster!!
Sampling Sinusoids… DT Sinusoids
x(t) x[n] x[n] is just a stream
of numbers
t n
Sensor Optional

Analog Digital Elec.

ADC DAC
Electronics (Computer)

ADC Clock sets how

often samples are taken
How closely should the samples be spaced??
At first thought we might think we need to have the samples still “look like”
the original sinusoid… But that turns out to be excessive, as our theory will
show eventually show.
Looking at the samples x[n] above they don’t quite really look like a
sinusoid… yet they are taken at a rate suitable for most applications!
So… how do we determine how fast we need to sample???
 Homework
1. What do you want to learn most form this course,
and why?
2. What are Elementary DT Signals? .