Digital Signal Processing

Markus Kuhn

Computer Laboratory, University of Cambridge

https://www.cl.cam.ac.uk/teaching/1920/{DSP,DSP2,L314}/

Michaelmas 2019
CST Part II(75%)Unit/Part II(50%)/Part III/MPhil ACS

dsp-slides.pdf 2019-11-12 15:52 e67247f 1 / 231

Signals
I flow of information
I measured quantity that varies with time (or position)
I electrical signal received from a transducer
(microphone, thermometer, accelerometer, antenna, etc.)
I electrical signal that controls a process

Continuous-time signals: voltage, current, temperature, speed, . . .

Discrete-time signals: daily minimum/maximum temperature,
lap intervals in races, sampled continuous signals, . . .

Electronics (unlike optics) can only deal easily with time-dependent

signals. Spatial signals, such as images, are typically first converted into
a time signal with a scanning process (TV, fax, etc.).

2 / 231
Signal processing

Signals may have to be transformed in order to

I amplify or filter out embedded information
I detect patterns
I prepare the signal to survive a transmission channel
I prevent interference with other signals sharing a medium
I undo distortions contributed by a transmission channel
I compensate for sensor deficiencies
I find information encoded in a different domain
To do so, we also need
I methods to measure, characterise, model and simulate transmission
channels
I mathematical tools that split common channels and transformations
into easily manipulated building blocks

3 / 231
Analog electronics

Passive networks (resistors, capacitors,

R
inductances, crystals, SAW filters),
non-linear elements (diodes, . . . ),
(roughly) linear operational amplifiers Uin L C Uout

Advantages: Uin

Uout
I passive networks are highly linear
over a very large dynamic range √
ω (= 2πf)
0 1/ LC
and large bandwidths
Uin
I analog signal-processing circuits
require little or no power Uout

t
I analog circuits cause little
additional interference
Z t
Uin − Uout 1 dUout
= Uout dτ +C
R L −∞ dt
4 / 231
Digital signal processing

Analog/digital and digital/analog converter, CPU, DSP, ASIC, FPGA.

Advantages:
I noise is easy to control after initial quantization
I highly linear (within limited dynamic range)
I complex algorithms fit into a single chip
I flexibility, parameters can easily be varied in software
I digital processing is insensitive to component tolerances, aging,
environmental conditions, electromagnetic interference
But:
I discrete-time processing artifacts (aliasing)
I can require significantly more power (battery, cooling)
I digital clock and switching cause interference

5 / 231
Some DSP applications

communication systems geophysics

modulation/demodulation, channel seismology, oil exploration
equalization, echo cancellation
astronomy
consumer electronics VLBI, speckle interferometry
perceptual coding of audio and video (DAB,
DVB, DVD), speech synthesis, speech transportation
recognition radar, radio navigation

music security
synthetic instruments, audio effects, noise steganography, digital watermarking, biometric
reduction identification, surveillance systems, signals
intelligence, electronic warfare
medical diagnostics
magnetic-resonance and ultrasonic imaging, engineering
X-ray computed tomography, ECG, EEG, MEG, control systems, feature extraction for pattern
AED, audiology recognition, sensor-data evaluation

6 / 231
Objectives

By the end of the course, you should be able to

I apply basic properties of time-invariant linear systems
I understand sampling, aliasing, convolution, filtering, the pitfalls of
spectral estimation
I explain the above in time and frequency domain representations
I use filter-design software
I visualise and discuss digital filters in the z-domain
I use the FFT for convolution, deconvolution, filtering
I implement, apply and evaluate simple DSP applications in MATLAB
I apply transforms that reduce correlation between several signal sources
I understand the basic principles of several widely-used modulation and
image-coding techniques.

7 / 231
Textbooks

I R.G. Lyons: Understanding digital signal processing. 3rd ed.,

Prentice-Hall, 2010. (£68)
I A.V. Oppenheim, R.W. Schafer: Discrete-time signal processing. 3rd
ed., Prentice-Hall, 2007. (£47)
I J. Stein: Digital signal processing – a computer science perspective.
Wiley, 2000. (£133)
I S.W. Smith: Digital signal processing – a practical guide for
engineers and scientists. Newness, 2003. (£48)
I K. Steiglitz: A digital signal processing primer – with applications to
digital audio and computer music. Addison-Wesley, 1996. (£67)
I Sanjit K. Mitra: Digital signal processing – a computer-based
approach. McGraw-Hill, 2002. (£38)

8 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Sequences and systems

A discrete sequence {xn }∞

n=−∞ is a sequence of numbers

. . . , x−2 , x−1 , x0 , x1 , x2 , . . .

where xn denotes the n-th number in the sequence (n ∈ Z). A discrete

sequence maps integer numbers onto real (or complex) numbers.
We normally abbreviate {xn }∞ n=−∞ to {xn }, or to {xn }n if the running index is not obvious.
The notation is not well standardized. Some authors write x[n] instead of xn , others x(n).

Where a discrete sequence {xn } samples a continuous function x(t) as

xn = x(ts · n) = x(n/fs ),

we call ts the sampling period and fs = 1/ts the sampling frequency.

A discrete system T receives as input a sequence {xn } and transforms it
into an output sequence {yn } = T {xn }:
discrete
. . . , x2 , x1 , x0 , x−1 , . . . . . . , y2 , y1 , y0 , y−1 , . . .
system T

9 / 231
Some simple sequences

un
Unit-step sequence: 1
(
0, n < 0
un =
1, n ≥ 0

. . . −3 −2 −1 0 1 2 3 ... n

δn
Impulse sequence: 1
(
1, n = 0
δn =
6 0
0, n =
= un − un−1
. . . −3 −2 −1 0 1 2 3 ... n

10 / 231
Sinusoidial sequences
A cosine wave, amplitude A, frequency f , phase offset ϕ:
x(t) = A · cos( 2πf t + ϕ )
| {z }
phase

Sampling it at sampling rate fs results in the discrete sequence {xn }:

xn = A · cos(2πf n/fs + ϕ) = A · cos(ω̇n + ϕ)
where ω̇ = 2πf /fs is the frequency expressed in radians per sample.
MATLAB/Octave example:
1

n=0:40; fs=8000; 0.8

0.6
f=400; x=cos(2*pi*f*n/fs); 0.4

stem(n, x); ylim([-1.1 1.1]) 0.2

This shows 41 samples (≈ 1/200 s = 5 ms) 0

of an f = 400 Hz sine wave, sampled at −0.2

fs = 8 kHz. −0.4

Exercise: Try f = 0, 1000, 2000, 3000, 4000, −0.6

5000 Hz. Try negative f . Try sine instead of co- −0.8

sine. Try adding phase offsets ϕ of ±π/4, ±π/2, −1

and ±π. 0 5 10 15 20 25 30 35 40

11 / 231
Properties of sequences
A sequence {xn } is
periodic ⇔ ∃k > 0 : ∀n ∈ Z : xn = xn+k

Is a continuous function with period tp still periodic after sampling?

∞
X
absolutely summable ⇔ |xn | < ∞
n=−∞
X∞
square summable ⇔ |xn |2 < ∞ ⇔ “energy signal”
n=−∞
| {z }
“energy”
k
1 X
0 < lim |xn |2 < ∞ ⇔ “power signal”
k→∞ 1 + 2k
n=−k
| {z }
“average power”

This energy/power terminology reflects that if U is a voltage supplied to a load

resistor R, then P = U I = U 2 /R is the power consumed, and P (t) dt the energy. It
R
is used even if we drop physical units (e.g., volts) for simplicity in calculations.
12 / 231
Properties of sequences
A sequence {xn } is
periodic ⇔ ∃k > 0 : ∀n ∈ Z : xn = xn+k

Is a continuous function with period tp still periodic after sampling? Only if tp /ts ∈ Q.

∞
X
absolutely summable ⇔ |xn | < ∞
n=−∞
X∞
square summable ⇔ |xn |2 < ∞ ⇔ “energy signal”
n=−∞
| {z }
“energy”
k
1 X
0 < lim |xn |2 < ∞ ⇔ “power signal”
k→∞ 1 + 2k
n=−k
| {z }
“average power”

This energy/power terminology reflects that if U is a voltage supplied to a load

resistor R, then P = U I = U 2 /R is the power consumed, and P (t) dt the energy. It
R
is used even if we drop physical units (e.g., volts) for simplicity in calculations.
12 / 231
A brief excursion into measuring signal intensity
Root-mean-square (RMS) signal strength
DC = direct current (constant), AC = alternating current (zero mean)
Consider a time-variable signal f (t) over time interval [t1 , t2 ]:
Z t2
1
DC component = mean voltage = f (τ ) dτ
t2 − t1 t1
AC component = f (t) − DC component
How can we state the strength of an AC signal?
The root-mean-square signal strength (voltage, etc.)
s Z t2
1
rms = f 2 (τ ) dτ
t 2 − t 1 t1
is the strength of a DC signal of equal average power.
RMS of a sine wave:
s
Z 2πk
1 A
[A · sin(τ + ϕ)]2 dτ = √ for all k ∈ N, A, ϕ ∈ R
2πk 0 2
13 / 231
Perception of signal strength
Sensation limit (SL) = lowest intensity stimulus that can still be perceived
Difference limit (DL) = smallest perceivable stimulus difference at given
intensity level

Weber’s law
Difference limit ∆φ is proportional to the intensity φ of the stimulus
(except for a small correction constant a, to describe deviation of
experimental results near SL):

∆φ = c · (φ + a)

Fechner’s scale
Define a perception intensity scale ψ using the sensation limit φ0 as the
origin and the respective difference limit ∆φ = c · φ as a unit step. The
result is a logarithmic relationship between stimulus intensity and scale
value:
φ
ψ = logc
φ0
14 / 231
Fechner’s scale matches older subjective intensity scales that follow
differentiability of stimuli, e.g. the astronomical magnitude numbers for
star brightness introduced by Hipparchos (≈150 BC).

Stevens’ power law

A sound that is 20 DL over SL is perceived as more than twice as loud as
one that is 10 DL over SL, i.e. Fechner’s scale does not describe well
perceived intensity. A rational scale attempts to reflect subjective
relations perceived between different values of stimulus intensity φ.
Stanley Smith Stevens observed that such rational scales ψ follow a
power law:
ψ = k · (φ − φ0 )a
Example coefficients a: brightness 0.33, loudness 0.6, heaviness 1.45,
temperature (warmth) 1.6.

15 / 231
Units and decibel

Communications engineers often use logarithmic units:

I Quantities often vary over many orders of magnitude → difficult to
agree on a common SI prefix (nano, micro, milli, kilo, etc.)
I Quotient of quantities (amplification/attenuation) usually more
interesting than difference
I Signal strength usefully expressed as field quantity (voltage, current,
pressure, etc.) or power, but quadratic relationship between these
two (P = U 2 /R = I 2 R) rather inconvenient
I Perception is logarithmic (Weber/Fechner law → slide 14)
Plus: Using magic special-purpose units has its own odd attractions (→ typographers, navigators)

Neper (Np) denotes the natural logarithm of the quotient of a field

quantity F and a reference value F0 . (rarely used today)
Bel (B) denotes the base-10 logarithm of the quotient of a power P and
a reference power P0 . Common prefix: 10 decibel (dB) = 1 bel.

16 / 231
Decibel
Where P is some power and P0 a 0 dB reference power, or equally where
F is a field quantity and F0 the corresponding reference level:
P F
10 dB · log10 = 20 dB · log10
P0 F0
Common reference values are indicated with a suffix after “dB”:
0 dBW = 1 W
0 dBm = 1 mW = −30 dBW
0 dBµV = 1 µV
√
0 dBu = 0.775 V = 600 Ω × 1 mW
0 dBSPL = 20 µPa (sound pressure level)
0 dBSL = perception threshold (sensation limit)
0 dBFS = full scale (clipping limit of analog/digital converter)
Remember:
3 dB = 2× power, 6 dB = 2× voltage/pressure/etc.
10 dB = 10× power, 20 dB = 10× voltage/pressure/etc.
W.H. Martin: Decibel – the new name for the transmission unit. Bell Syst. Tech. J., Jan. 1929.
ITU-R Recommendation V.574-4: Use of the decibel and neper in telecommunication.
17 / 231
Types of discrete systems
discrete
. . . , x2 , x1 , x0 , x−1 , . . . . . . , y2 , y1 , y0 , y−1 , . . .
system T

A causal system cannot look into the future:

yn = f (xn , xn−1 , xn−2 , . . .)
A memory-less system depends only on the current input value:
yn = f (xn )
A delay system shifts a sequence in time:
yn = xn−d
T is a time-invariant system if for any d
{yn } = T {xn } ⇐⇒ {yn−d } = T {xn−d }.
T is a linear system if for any pair of sequences {xn } and {x0n }
T {a · xn + b · x0n } = a · T {xn } + b · T {x0n }.
18 / 231
Example: M -point moving average system

M −1
1 X xn−M +1 + · · · + xn−1 + xn
yn = xn−k =
M M
k=0

It is causal, linear, time-invariant, with memory. With M = 4:

x
y

19 / 231
Example: exponential averaging system

∞
X
yn = α · xn + (1 − α) · yn−1 =α (1 − α)k · xn−k
k=0

It is causal, linear, time-invariant, with memory. With α = 12 :

x
y

20 / 231
Example: accumulator system

n
X
yn = xk
k=−∞

It is causal, linear, time-invariant, with memory.

x
y

21 / 231
Example: backward difference system

yn = xn − xn−1

It is causal, linear, time-invariant, with memory.

x
y

22 / 231
Other examples
Time-invariant non-linear memory-less systems:

yn = x2n , yn = log2 xn , yn = max{min{b256xn c, 255}, 0}

Linear but not time-invariant systems:

(
xn , n ≥ 0
yn = = xn · un
0, n<0
yn = xbn/4c
yn = xn · <(eω̇ jn )

Linear time-invariant non-causal systems:

1
yn = (xn−1 + xn+1 )
2
9
X sin(πkω)
yn = xn+k · · [0.5 + 0.5 · cos(πk/10)]
πkω
k=−9

23 / 231
Constant-coefficient difference equations
Of particular practical interest are causal linear time-invariant systems of
the form
xn b0 yn
N
X
yn = b0 · xn − ak · yn−k z −1
−a1
k=1
yn−1

z −1
Block diagram representation −a2
of sequence operations: yn−2
x0n
z −1
−a3
xn xn + x0n
Addition: yn−3

Multiplication xn a axn
by constant: The ak and bm are
constant coefficients.
xn xn−1
Delay: z −1
24 / 231
or
xn xn−1 xn−2 xn−3
z −1 z −1 z −1
M
X
yn = bm · xn−m b0 b1 b2 b3
m=0

yn

or the combination of both:

xn b0 a−1
0 yn

z −1 z −1
b1 −a1
N
X M
X xn−1 yn−1
ak · yn−k = bm · xn−m
k=0 m=0 z −1 z −1
b2 −a2
xn−2 yn−2

z −1 z −1
b3 −a3
xn−3 yn−3

The MATLAB function filter is an efficient implementation of the last variant.

25 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Convolution
Another example of a LTI systems is
∞
X
yn = ak · xn−k
k=−∞

where {ak } is a suitably chosen sequence of coefficients.

This operation over sequences is called convolution and is defined as
∞
X
{pn } ∗ {qn } = {rn } ⇐⇒ ∀n ∈ Z : rn = pk · qn−k .
k=−∞

26 / 231
Convolution examples

A B C D

E F A∗B A∗C

C∗A A∗E D∗E A∗F

27 / 231
Convolution
Another example of a LTI systems is
∞
X
yn = ak · xn−k
k=−∞

where {ak } is a suitably chosen sequence of coefficients.

This operation over sequences is called convolution and is defined as
∞
X
{pn } ∗ {qn } = {rn } ⇐⇒ ∀n ∈ Z : rn = pk · qn−k .
k=−∞

← 26 / 231
Convolution
Another example of a LTI systems is
∞
X
yn = ak · xn−k
k=−∞

where {ak } is a suitably chosen sequence of coefficients.

This operation over sequences is called convolution and is defined as
∞
X
{pn } ∗ {qn } = {rn } ⇐⇒ ∀n ∈ Z : rn = pk · qn−k .
k=−∞

If {yn } = {an } ∗ {xn } is a representation of an LTI system T , with

{yn } = T {xn }, then we call the sequence {an } the P impulse response of
T , because {an } = T {δn }, as {an } ∗ {δn } = {an }, k ak · δn−k = an .

← 26 / 231
Properties of convolution
For arbitrary sequences {pn }, {qn }, {rn } and scalars a, b:
I Convolution is associative

({pn } ∗ {qn }) ∗ {rn } = {pn } ∗ ({qn } ∗ {rn })

I Convolution is commutative

{pn } ∗ {qn } = {qn } ∗ {pn }

I Convolution is linear

{pn } ∗ {a · qn + b · rn } = a · ({pn } ∗ {qn }) + b · ({pn } ∗ {rn })

I The impulse sequence (slide 10) is neutral under convolution

{pn } ∗ {δn } = {δn } ∗ {pn } = {pn }

I Sequence shifting is equivalent to convolving with a shifted impulse

{pn−d }n = {pn } ∗ {δn−d }n

28 / 231
Convolution
Another example of a LTI systems is
∞
X
yn = ak · xn−k
k=−∞

where {ak } is a suitably chosen sequence of coefficients.

This operation over sequences is called convolution and is defined as
∞
X
{pn } ∗ {qn } = {rn } ⇐⇒ ∀n ∈ Z : rn = pk · qn−k .
k=−∞

If {yn } = {an } ∗ {xn } is a representation of an LTI system T , with

{yn } = T {xn }, then we call the sequence {an } the P impulse response of
T , because {an } = T {δn }, as {an } ∗ {δn } = {an }, k ak · δn−k = an .

← 26 / 231
Convolution
Another example of a LTI systems is
∞
X
yn = ak · xn−k
k=−∞

where {ak } is a suitably chosen sequence of coefficients.

This operation over sequences is called convolution and is defined as
∞
X
{pn } ∗ {qn } = {rn } ⇐⇒ ∀n ∈ Z : rn = pk · qn−k .
k=−∞

If {yn } = {an } ∗ {xn } is a representation of an LTI system T , with

{yn } = T {xn }, then we call the sequence {an } the P impulse response of
T , because {an } = T {δn }, as {an } ∗ {δn } = {an }, k ak · δn−k = an .

If f and g are continuous functions, their convolution is defined similarly as the integral
Z ∞
(f ∗ g)(t) = f (s)g(t − s)ds.
−∞
But what is the continuous equivalent of {δn }? More on that later . . .
← 26 / 231
All LTI systems just apply convolution

Proof:
Any sequence {xn } can be decomposed into a weighted sum of shifted
impulse sequences:
X∞
{xn } = xk · {δn−k }
k=−∞

Let’s see what happens if we apply a linear(∗) time-invariant(∗∗) system T

to such a decomposed sequence:
∞
! ∞
X (∗) X
T {xn } = T xk · {δn−k } = xk · T {δn−k }
k=−∞ k=−∞
∞ ∞
!
(∗∗) X X
= xk · {δn−k } ∗ T {δn } = xk · {δn−k } ∗ T {δn }
k=−∞ k=−∞

= {xn } ∗ T {δn } q.e.d.

⇒ The impulse response T {δn } fully characterizes an LTI system.

29 / 231
Direct form I and II implementations

xn b0 a−1
0 yn xn a−1
0 b0 yn

z −1 z −1 z −1
b1 −a1 −a1 b1
xn−1 yn−1

z −1 z −1 = z −1
b2 −a2 −a2 b2
xn−2 yn−2

z −1 z −1 z −1
b3 −a3 −a3 b3
xn−3 yn−3

The block diagram representation of the constant-coefficient difference

equation on slide 25 is called the direct form I implementation.
The number of delay elements can be halved by using the commutativity
of convolution to swap the two feedback loops, leading to the direct form
II implementation of the same LTI system.
These two forms are only equivalent with ideal arithmetic (no rounding errors and range limits).
30 / 231
Convolution: optics example
If a projective lens is out of focus, the blurred image is equal to the
original image convolved with the aperture shape (e.g., a filled circle):

∗ =

as image plane focal plane

Point-spread function h (disk, r = 2f ):

1
x2 + y 2 ≤ r 2

,
h(x, y) = r2 π
0, x2 + y 2 > r 2
a
Original image I, blurred image B = I ∗ h, i.e.
ZZ
0 0 0 0 0 0
B(x, y) = I(x−x , y−y )·h(x , y )·dx dy s

f 31 / 231
Convolution: electronics example

R
Uin

Uin C Uout

Uout

t
Any passive network (resistors, capacitors, inductors) convolves its input
voltage Uin with an impulse response function h, leading to
Uout = Uin ∗ h, that is
Z ∞
Uout (t) = Uin (t − τ ) · h(τ ) · dτ
−∞

In the above example:

−t
Uin − Uout 1

dUout · e RC , t≥0
=C· , h(t) = RC
R dt 0, t<0

32 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Adding sine waves
Adding together sine waves of equal frequency, but arbitrary amplitude
and phase, results in another sine wave of the same frequency:

A1 · sin(ωt + ϕ1 ) + A2 · sin(ωt + ϕ2 ) = A · sin(ωt + ϕ)

Why?

33 / 231
Adding sine waves
Adding together sine waves of equal frequency, but arbitrary amplitude
and phase, results in another sine wave of the same frequency:

A1 · sin(ωt + ϕ1 ) + A2 · sin(ωt + ϕ2 ) = A · sin(ωt + ϕ)

Why?

Think of A · sin(ωt + ϕ) as the height of

ω
an arrow of length A, rotating 2π times per second,
with start angle ϕ (radians) at t = 0. A

A · sin(ωt + ϕ)
ωt ϕ

33 / 231
Adding sine waves
Adding together sine waves of equal frequency, but arbitrary amplitude
and phase, results in another sine wave of the same frequency:

A1 · sin(ωt + ϕ1 ) + A2 · sin(ωt + ϕ2 ) = A · sin(ωt + ϕ)

Why?

Think of A · sin(ωt + ϕ) as the height of

ω
an arrow of length A, rotating 2π times per second,
A2
with start angle ϕ (radians) at t = 0. A
Consider two more such arrows, ϕ2
of length A1 and A2 , A1
with start angles ϕ1 and ϕ2 . ωt ϕ
A1 and A2 stuck together are as high as A, ϕ1
all three rotating at the same frequency.

33 / 231
Adding sine waves
Adding together sine waves of equal frequency, but arbitrary amplitude
and phase, results in another sine wave of the same frequency:

A1 · sin(ωt + ϕ1 ) + A2 · sin(ωt + ϕ2 ) = A · sin(ωt + ϕ)

Why?

Think of A · sin(ωt + ϕ) as the height of

ω
an arrow of length A, rotating 2π times per second,
A2
with start angle ϕ (radians) at t = 0. A
Consider two more such arrows, ϕ2
of length A1 and A2 , A1
with start angles ϕ1 and ϕ2 . ωt ϕ
A1 and A2 stuck together are as high as A, ϕ1
all three rotating at the same frequency.

But adding sine waves as vectors (A1 , ϕ1 ) and (A2 , ϕ2 ) in polar coordinates is cumbersome:
q A1 sin ϕ1 + A2 sin ϕ2
A = A21 + A22 + 2A1 A2 cos(ϕ2 − ϕ1 ), tan ϕ =
A1 cos ϕ1 + A2 cos ϕ2
33 / 231
Cartesian coordinates for sine waves
Sine waves of any amplitude A and phase (start angle) ϕ can be
represented as linear combinations of sin(ωt) and cos(ωt):

A · sin(ωt + ϕ) = x · sin(ωt) + y · cos(ωt) cos(ωt) = sin(ωt + 90◦ )

where
A
x = A · cos(ϕ), y = A · sin(ϕ) A · sin(ϕ)
ωt ϕ
and A · cos(ϕ)
p y
A= x2 + y 2 , tan ϕ = .
x
Base: two rotating arrows with start angles 0◦ [height = sin(ω)] and 90◦ [height = cos(ω)].

Adding two sine waves as vectors in Cartesian coordinates is simple:

f1 (t) = x1 · sin(ω) + y1 · cos(ω)
f2 (t) = x2 · sin(ω) + y2 · cos(ω)
f1 (t) + f2 (t) = (x1 + x2 ) · sin(ω) + (y1 + y2 ) · cos(ω)
34 / 231
Why are sine waves useful?
1) Sine-wave sequences form a family of discrete
sequences that is closed under convolution with
arbitrary sequences.
Convolution of a discrete sequence {xn } with another sequence {hn } is
nothing but adding together scaled and delayed copies of {xn }.
Think again of {hn } as decomposed into a sum of impulses:
X X
{xn } ∗ {hn } = {xn } ∗ hk · {δn−k }n = hk · ({xn } ∗ {δn−k }n )
k k
X
= hk · {xn−k }n
k

If {xn } is a sampled sine wave of frequency f , i.e.

xn = Ax · sin(2πf t + φx )
P
then {yn } = {xn } ∗ {hn } = k hk · {xn−k }n is another sampled sine
wave of frequency f , i.e. for each {hn } there exists a pair (Ay , φy ) with
yn = Ay · sin(2πf t + φy )
The equivalent applies for continuous sine waves and convolution. 35 / 231
Why are sine waves useful?
2) Sine waves are orthogonal to each other
The term “orthogonal” is used here in the context of an (infinitely dimensional)
vector space, where the “vectors” are functions of the form f : R → R
(or f : R → C) and the scalar product is defined as
Z
f · g = f (t) · g(t) dt.

Over integer (half-)periods:

Z π
m, n ∈ N, m 6= n ⇒ sin(nt) sin(mt)dt = 0
0
Z π
m, n ∈ N ⇒ sin(nt) cos(mt)dt = 0
−π

We can even (with some handwaving) extend this to improper integrals:

Z ∞
sin(ω1 t + ϕ1 ) · sin(ω2 t + ϕ2 ) dt “=” 0
−∞
⇐⇒ ω1 6= ω2 ∨ ϕ1 − ϕ2 = (2k + 1)π/2 (k ∈ Z)
They can be used to form an orthogonal function basis for a transform.
36 / 231
 1

sin(1t)⋅sin(2t)
sin(1t)
sin(2t)
−1
0 1.5708 3.1416 4.7124 6.2832
t

37 / 231
 1

sin(2t)⋅sin(3t)
sin(2t)
sin(3t)
−1
0 1.5708 3.1416 4.7124 6.2832
t
 1

sin(3t)⋅sin(4t)
sin(3t)
sin(4t)
−1
0 1.5708 3.1416 4.7124 6.2832
t
 1

sin(2t)⋅sin(4t)
sin(2t)
sin(4t)
−1
0 1.5708 3.1416 4.7124 6.2832
t
 1

sin(t)⋅cos(t)
sin(t)
cos(t)
−1
0 1.5708 3.1416 4.7124 6.2832
t
Why are exponential functions useful?
Adding together two exponential functions with the same base z, but
different scale factor and offset, results in another exponential function
with the same base:

A1 · z t+ϕ1 + A2 · z t+ϕ2 = A1 · z t · z ϕ1 + A2 · z t · z ϕ2
= (A1 · z ϕ1 + A2 · z ϕ2 ) · z t = A · z t

Likewise, if we convolve a sequence {xn } of values

. . . , z −3 , z −2 , z −1 , 1, z, z 2 , z 3 , . . .

xn = z n with an arbitrary sequence {hn }, we get {yn } = {z n } ∗ {hn },

∞
X ∞
X ∞
X
yn = xn−k · hk = z n−k
· hk = z ·n
z −k · hk = z n · H(z)
k=−∞ k=−∞ k=−∞

where H(z) is independent of n.

Exponential sequences are closed under convolution with
arbitrary sequences.
The same applies in the continuous case.
38 / 231
Why are complex numbers so useful?
1) They give us all n solutions (“roots”)
√ of equations involving
polynomials up to degree n (the “ −1 = j ” story).
2) They give us the “great unifying theory” that combines sine and
exponential functions:
1 jθ
e + e− jθ


cos(θ) =
2
1
e jθ − e− jθ


sin(θ) =
2j
or
1  j(ωt+ϕ) 
cos(ωt + ϕ) = e + e− j(ωt+ϕ)
2
or

cos(ω̇n + ϕ) = <(e j(ω̇n+ϕ) ) = <[(e jω̇ )n · e jϕ ]

sin(ω̇n + ϕ) = =(e j(ω̇n+ϕ) ) = =[(e jω̇ )n · e jϕ ]

Notation: <(a + jb) := a, =(a + jb) := b and (a + jb)∗ := a − jb, where j2 = −1 and a, b ∈ R.
Then <(x) = 12 (x + x∗ ) and =(x) = 21j (x − x∗ ) for all x ∈ C.
39 / 231
We can now represent sine waves as projections of a rotating complex
vector. This allows us to represent sine-wave sequences as exponential
sequences with basis e jω̇ .
A phase shift in such a sequence corresponds to a rotation of a complex
vector.
3) Complex multiplication allows us to modify the amplitude and phase
of a complex rotating vector using a single operation and value.
Rotation of a 2D vector in (x, y)-form is notationally slightly messy, but
fortunately j2 = −1 does exactly what is required here:
     
x3 x2 −y2 x1
= ·
y3 y2 x2 y1 (x3 , y3 )
 
x1 x2 − y1 y2
= (−y2 , x2 )
x1 y2 + x2 y1
(x2 , y2 )
z1 = x1 + jy1 , z2 = x2 + jy2 (x1 , y1 )

z1 · z2 = x1 x2 − y1 y2 + j(x1 y2 + x2 y1 )

40 / 231
Complex phasors
Amplitude and phase are two distinct characteristics of a sine function
that are inconvenient to keep separate notationally.
Complex functions (and discrete sequences) of the form

(A · e jϕ ) · e jωt = A · e j(ωt+ϕ) = A · [cos(ωt + ϕ) + j · sin(ωt + ϕ)]

(where j2 = −1) are able to represent both amplitude A ∈ R+ and phase

ϕ ∈ [0, 2π) in one single algebraic object A · e jϕ ∈ C.
Thanks to complex multiplication, we can also incorporate in one single
factor both a multiplicative change of amplitude and an additive change
of phase of such a function. This makes discrete sequences of the form

xn = e jω̇n

eigensequences with respect to an LTI system T , because for each ω̇,

there is a complex number (eigenvalue) H(ω̇) such that

T {xn } = H(ω̇) · {xn }

In the notation of slide 38, where the argument of H is the base, we would write H(e jω̇ ).
41 / 231
Recall: Fourier transform

We define the Fourier integral transform and its inverse as

Z ∞
F{g(t)}(f ) = G(f ) = g(t) · e−2π jf t dt
−∞
Z ∞
F −1 {G(f )}(t) = g(t) = G(f ) · e2π jf t df
−∞

Many equivalent forms of the Fourier transform are used in the literature. There is no strong
consensus on whether the forward transform uses e−2π jf t and the backwards transform e2π jf t , or
vice versa. The above form uses the ordinary frequency f , whereas some authors prefer the angular
frequency ω = 2πf :
Z ∞
∓ jωt
F {h(t)}(ω) = H(ω) = α h(t) · e dt
−∞

Z ∞
± jωt
F −1 {H(ω)}(t) = h(t) = β H(ω)· e dω
−∞

This substitution introduces factors α and β such that αβ = 1/(2π). Some authors set α = 1
and β = 1/(2π), to keep √ the convolution theorem free of a constant prefactor; others prefer the
unitary form α = β = 1/ 2π, in the interest of symmetry.

42 / 231
Properties of the Fourier transform
If
x(t) •−◦ X(f ) and y(t) •−◦ Y (f )
are pairs of functions that are mapped onto each other by the Fourier
transform, then so are the following pairs.

Linearity:

ax(t) + by(t) •−◦ aX(f ) + bY (f )

Time scaling:
 
1 f
x(at) •−◦ X
|a| a

Frequency scaling:
 
1 t
x •−◦ X(af )
|a| a
43 / 231
Time shifting:

x(t − ∆t) •−◦ X(f ) · e−2π jf ∆t

Frequency shifting:

x(t) · e2π j∆f t •−◦ X(f − ∆f )

Parseval’s theorem (total energy):

Z ∞ Z ∞
2
|x(t)| dt = |X(f )|2 df
−∞ −∞

44 / 231
Fourier transform example: rect and sinc

The Fourier transform of the “rectangular function”

1

 1 if |t| < 2
 1
1
rect(t) = 2 if |t| = 12
 0
0 otherwise

− 12 0 1
2
is the “(normalized) sinc function”
1
sin πf
Z 2
F{rect(t)}(f ) = e−2π jf t dt = = sinc(f )
− 12 πf

and vice versa

F{sinc(t)}(f ) = rect(f ).
Some noteworthy properties of these functions:
I ∞ sinc(t) dt = 1 = ∞ rect(t) dt
R R
−∞ −∞ 1
I sinc(0) = 1 = rect(0)
I ∀n ∈ Z \ {0} : sinc(n) = 0 0
−3 −2 −1 0 1 2 3
45 / 231
Convolution theorem
Convolution in the time domain is equivalent to (complex) scalar
multiplication in the frequency domain:

F{(f ∗ g)(t)} = F{f (t)} · F {g(t)}

z(r)e− jωr dr = r s x(s)y(r − s)e− jωr dsdr =

R R R R
Proof: z(r) = s x(s)y(r − s)ds ⇐⇒ r
t:=r−s
x(s) r y(r − s)e− jωr drds = s x(s)e− jωs r y(r − s)e− jω(r−s) drds =
R R R R
s

x(s)e− jωs t y(t)e− jωt dtds = s x(s)e− jωs ds · t y(t)e− jωt dt.
R R R R
s

Convolution in the frequency domain corresponds to scalar multiplication

in the time domain:

F{f (t) · g(t)} = F{f (t)} ∗ F {g(t)}

This second form is also called “modulation theorem”, as it describes what happens in the
frequency domain with amplitude modulation of a signal (see slide 53).
The proof is very similar to the one above.
Both equally work for the inverse Fourier transform:
−1 −1 −1
F {(F ∗ G)(f )} = F {F (f )} · F {G(f )}
−1 −1 −1
F {F (f ) · G(f )} = F {F (f )} ∗ F {G(f )}
46 / 231
Dirac delta function
The continuous equivalent of the impulse sequence {δn } is known as
Dirac delta function δ(x). It is a generalized function, defined such that


0, x 6= 0 1
δ(x) =
∞, x = 0
Z ∞
δ(x) dx = 1
−∞
0 x
and can be thought of as the limit of function sequences such as

0, |x| ≥ 1/n
δ(x) = lim
n→∞ n/2, |x| < 1/n
or
n 2 2
δ(x) = lim √ e−n x
n→∞ π
The delta function is mathematically speaking not a function, but a distribution, that is an
expression that is only defined when integrated.

47 / 231
Some properties of the Dirac delta function:
Z ∞
f (x)δ(x − a) dx = f (a)
−∞
Z ∞
e±2π jxa dx = δ(a)
−∞

∞ ∞
X 1 X
e±2π jixa = δ(x − i/a)
i=−∞
|a| i=−∞
1
δ(ax) = δ(x)
|a|

Fourier transform:
Z ∞
F{δ(t)}(f ) = δ(t) · e−2π jf t dt = e0 = 1
−∞
Z ∞
F −1 {1}(t) = 1 · e2π jf t df = δ(t)
−∞

48 / 231
Linking the Dirac delta with the Fourier transform
The Fourier transform of 1 follows from the Dirac delta’s ability to
sample inside an integral:
g(t) = F −1 (F(g))(t)
Z ∞ Z ∞ 
−2π jf s
= g(s) · e · ds · e2π jf t · df
−∞ −∞
Z ∞ Z ∞ 
−2π jf s 2π jf t
= e ·e · df · g(s) · ds
−∞ −∞
Z ∞ Z ∞ 
= e−2π jf (t−s) · df ·g(s) · ds
−∞ −∞
| {z }
δ(t−s)

So if δ has the property

Z ∞
g(t) = δ(t − s) · g(s) · ds
−∞
then Z ∞
e−2π jf (t−s) df = δ(t − s)
−∞
49 / 231
Z ∞ P10
e2π jtf df = δ(t) i=1 cos(2πfi t) ≈ δ(t)
−∞
f1 , . . . , f10 ∈ [0, 3] chosen uniformly at random
1

0.5

−0.5

−1
−4 −3 −2 −1 0 1 2 3 4

10

−5

−10
−4 −3 −2 −1 0 1 2 3 4 50 / 231
Z ∞ P100
e2π jtf df = δ(t) i=1 cos(2πfi t) ≈ δ(t)
−∞
f1 , . . . , f100 ∈ [0, 10] chosen uniformly at random
1

0.5

−0.5

−1
−4 −3 −2 −1 0 1 2 3 4

100

80

60

40

20

−20

−40
−4 −3 −2 −1 0 1 2 3 4
∞
X ∞
X
e±2π jnt = δ(t − n)
n=−∞ n=−∞
5
X ∞
X
cos(2πnt) ≈ δ(t − n)
n=1 n=−∞
1

0.5

−0.5

−1
−4 −3 −2 −1 0 1 2 3 4

−1

−2
−4 −3 −2 −1 0 1 2 3 4
Sine and cosine in the frequency domain

1 2π jf0 t 1 −2π jf0 t 1 2π jf0 t 1

cos(2πf0 t) = e + e sin(2πf0 t) = e − e−2π jf0 t
2 2 2j 2j
1 1
F{cos(2πf0 t)}(f ) = δ(f − f0 ) + δ(f + f0 )
2 2
j j
F{sin(2πf0 t)}(f ) = − δ(f − f0 ) + δ(f + f0 )
2 2
< <
1 1
2 2
= =
1 1
2 j 2 j

−f0 f0 f −f0 f0 f

As any x(t) ∈ R can be decomposed into sine and cosine functions, the spectrum of any
real-valued signal will show the symmetry X(−f ) = [X(f )]∗ , where ∗ denotes the complex
conjugate (i.e., negated imaginary part).
51 / 231
Fourier transform symmetries

We call a function x(t)

odd if x(−t) = −x(t)

even if x(−t) = x(t)

and ·∗ is the complex conjugate, such that (a + jb)∗ = (a − jb).

Then
x(t) is real ⇔ X(−f ) = [X(f )]∗
x(t) is imaginary ⇔ X(−f ) = −[X(f )]∗
x(t) is even ⇔ X(f ) is even
x(t) is odd ⇔ X(f ) is odd
x(t) is real and even ⇔ X(f ) is real and even
x(t) is real and odd ⇔ X(f ) is imaginary and odd
x(t) is imaginary and even ⇔ X(f ) is imaginary and even
x(t) is imaginary and odd ⇔ X(f ) is real and odd

52 / 231
Example: amplitude modulation

Communication channels usually permit only the use of a given frequency

interval, such as 300–3400 Hz for the analog phone network or 590–598
MHz for TV channel 36. Modulation with a carrier frequency fc shifts
the spectrum of a signal x(t) into the desired band.
Amplitude modulation (AM):

y(t) = A · cos(2πtfc ) · x(t)

X(f ) Y (f )
∗ =

−fl 0 fl f −fc fc f −fc 0 fc f

The spectrum of the baseband signal in the interval −fl < f < fl is
shifted by the modulation to the intervals ±fc − fl < f < ±fc + fl .
How can such a signal be demodulated?

53 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Sampling using a Dirac comb

The loss of information in the sampling process that converts a

continuous function x(t) into a discrete sequence {xn } defined by

xn = x(ts · n) = x(n/fs )

can be modelled through multiplying x(t) by a comb of Dirac impulses

∞
X
s(t) = ts · δ(t − ts · n)
n=−∞

to obtain the sampled function

x̂(t) = x(t) · s(t)

The function x̂(t) now contains exactly the same information as the
discrete sequence {xn }, but is still in a form that can be analysed using
the Fourier transform on continuous functions.

54 / 231
The Fourier transform of a Dirac comb
∞
X ∞
X
s(t) = ts · δ(t − ts · n) = e2π jnt/ts
n=−∞ n=−∞

is another Dirac comb

∞
( )
X
S(f ) = F ts · δ(t − ts n) (f ) =
n=−∞
Z∞ X
∞ ∞  
X n
ts · δ(t − ts n) e−2π jf t dt = δ f− .
ts
−∞ n=−∞ n=−∞

s(t) S(f )

−2ts −ts 0 ts 2ts t −2fs −fs 0 fs 2fs f

55 / 231
Sampling and aliasing

sample
cos(2π tf)
cos(2π t(k⋅ fs± f))

Sampled at frequency fs , the function cos(2πtf ) cannot be distinguished

from cos[2πt(kfs ± f )] for any k ∈ Z.
56 / 231
Frequency-domain view of sampling

x(t) s(t) x̂(t)

· =
... ... ... ...
0 t −1/fs 0 1/fs t −1/fs 0 1/fs t
X(f ) S(f ) X̂(f )

∗ =
... ... ... ...
0 f −fs fs f −fs 0 fs f
Sampling a signal in the time domain corresponds in the frequency
domain to convolving its spectrum with a Dirac comb. The resulting
copies of the original signal spectrum in the spectrum of the sampled
signal are called “images”.

57 / 231
Discrete-time Fourier transform (DTFT)
The Fourier transform of a sampled signal
∞
X
x̂(t) = ts · xn · δ(t − ts · n)
n=−∞

is
Z ∞ ∞
X f
F{x̂(t)}(f ) = X̂(f ) = x̂(t) · e−2π jf t dt = ts · xn · e−2π j fs n
−∞ n=−∞

The inverse transform is

Z ∞ Z fs /2 f
x̂(t) = X̂(f ) · e2π jf t df or xm = X̂(f ) · e2π j fs m df.
−∞ −fs /2

The DTFT is also commonly expressed using the normalized frequency

ω̇ = 2π ffs (radians per sample), and the notation
X
X(e jω̇ ) = xn · e− jω̇n
n
is customary, to highlight both the periodicity of the DTFT and its
relationship with the z-transform of {xn } (see slide 118).
58 / 231
 1 8
DTFT real
DTFT imag
0.8 6

0.6 4

0.4 2

0.2 0

0 -2
-5 0 5 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

1 8
DTFT real
DTFT imag
0.8 6

0.6 4

0.4 2

0.2 0

0 -2
-5 0 5 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

59 / 231
 1 8
DTFT real
DTFT imag
0.8 6

0.6 4

0.4 2

0.2 0

0 -2
-5 0 5 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

1 8
DTFT real
DTFT imag
0.8 6

0.6 4

0.4 2

0.2 0

0 -2
-5 0 5 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

60 / 231
 1 8
DTFT real
DTFT imag
0.8 6

0.6 4

0.4 2

0.2 0

0 -2
-5 0 5 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

1 8
DTFT real
DTFT imag
6
0.5

4
0
2

-0.5
0

-1 -2
-5 0 5 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

61 / 231
 1 8
DTFT real
DTFT imag
6
0.5

4
0
2

-0.5
0

-1 -2
-5 0 5 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

1 8
DTFT real
DTFT imag
0.8 6

0.6 4

0.4 2

0.2 0

0 -2
-5 0 5 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

62 / 231
Properties of the DTFT
The DTFT is periodic:

X̂(f ) = X̂(f + kfs ) or X(e jω̇ ) = X(e j(ω̇+2πk) ) ∀k ∈ Z

Beyond that, the DTFT is just the Fourier transform applied to a discrete
sequence, and inherits the properties of the continuous Fourier transform,
e.g.
I Linearity
I Symmetries
I Convolution and modulation theorem:

{xn } ∗ {yn } = {zn } ⇐⇒ X(e jω̇ ) · Y (e jω̇ ) = Z(e jω̇ )

and
Z π
0 0
xn · yn = zn ⇐⇒ X(e jω̇ ) · Y (e j(ω̇−ω̇ ) ) dω̇ 0 = Z(e jω̇ )
−π

63 / 231
Nyquist limit and anti-aliasing filters

If the (double-sided) bandwidth of a signal to be sampled is larger than

the sampling frequency fs , the images of the signal that emerge during
sampling may overlap with the original spectrum.
Such an overlap will hinder reconstruction of the original continuous
signal by removing the aliasing frequencies with a reconstruction filter.
Therefore, it is advisable to limit the bandwidth of the input signal to the
sampling frequency fs before sampling, using an anti-aliasing filter.
In the common case of a real-valued base-band signal (with frequency
content down to 0 Hz), all frequencies f that occur in the signal with
non-zero power should be limited to the interval −fs /2 < f < fs /2.
The upper limit fs /2 for the single-sided bandwidth of a baseband signal
is known as the “Nyquist limit”.

64 / 231
Nyquist limit and anti-aliasing filters

Without anti-aliasing filter With anti-aliasing filter

single-sided Nyquist
X(f ) bandwidth X(f ) limit = fs /2
anti-aliasing filter

0 f −fs 0 fs f
double-sided bandwidth
reconstruction filter
X̂(f ) X̂(f )

−2fs −fs 0 fs 2fs f −2fs −fs 0 fs 2fs f

Anti-aliasing and reconstruction filters both suppress frequencies outside |f | < fs /2.

65 / 231
Reconstruction of a continuous band-limited waveform

The ideal anti-aliasing filter for eliminating any frequency content above
fs /2 before sampling with a frequency of fs has the Fourier transform
(
1 if |f | < f2s
H(f ) = = rect(ts f ).
0 if |f | > f2s

This leads, after an inverse Fourier transform, to the impulse response

 
sin πtfs 1 t
h(t) = fs · = · sinc .
πtfs ts ts

The original band-limited signal can be reconstructed by convolving this

with the sampled signal x̂(t), which eliminates the periodicity of the
frequency domain introduced by the sampling process:

x(t) = h(t) ∗ x̂(t)

Note that sampling h(t) gives the impulse function: h(t) · s(t) = δ(t).

66 / 231
Impulse response of ideal low-pass filter with cut-off frequency fs /2:

−3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5 3

t⋅ fs
67 / 231
Reconstruction filter example

sampled signal
interpolation result
scaled/shifted sin(x)/x pulses

1 2 3 4 5
68 / 231
If before being sampled with xn = x(t/fs ) the signal x(t) satisfied the
Nyquist limit
Z ∞
F{x(t)}(f ) = x(t) · e−2π jf t dt = 0 for all |f | ≥ f2s
−∞
 
1 t
then it can be reconstructed by interpolation with h(t) = ts sinc ts :
Z ∞
x(t) = h(s) · x̂(t − s) · ds
−∞
Z ∞   ∞
1 s X
= sinc · ts xn · δ(t − s − ts · n) · ds
−∞ ts ts n=−∞
∞ Z ∞  
X s
= xn · sinc · δ(t − s − ts · n) · ds
n=−∞ −∞ t s
∞ ∞
t − ts · n
X   X
= xn · sinc = xn · sinc(t/ts − n)
n=−∞
ts n=−∞
∞
X sin π(t/ts − n)
= xn ·
n=−∞
π(t/ts − n)

69 / 231
Reconstruction filters

The mathematically ideal form of a reconstruction filter for suppressing

aliasing frequencies interpolates the sampled signal xn = x(ts · n) back
into the continuous waveform
∞
X sin π(t/ts − n)
x(t) = xn · .
n=−∞
π(t/ts − n)

Choice of sampling frequency

Due to causality and economic constraints, practical analog filters can only
approximate such an ideal low-pass filter. Instead of a sharp transition between the
“pass band” (< fs /2) and the “stop band” (> fs /2), they feature a “transition band”
in which their signal attenuation gradually increases.
The sampling frequency is therefore usually chosen somewhat higher than twice the
highest frequency of interest in the continuous signal (e.g., 4×). On the other hand,
the higher the sampling frequency, the higher are CPU, power and memory
requirements. Therefore, the choice of sampling frequency is a tradeoff between signal
quality, analog filter cost and digital subsystem expenses.

70 / 231
Band-pass signal sampling
Sampled signals can also be reconstructed if their spectral components
remain entirely within the interval n · fs /2 < |f | < (n + 1) · fs /2 for some
n ∈ N. (The baseband case discussed so far is just n = 0.)
X(f ) anti-aliasing filter X̂(f ) reconstruction filter

− 54 fs 0 5
4 fs f −fs −fs 0 fs fs f
2 2
n=2

In this case, the aliasing copies of the positive and the negative
frequencies will interleave instead of overlap, and can therefore be
removed again later by a reconstruction filter.
The ideal reconstruction filter for this sampling technique will only allow frequencies in the interval
[n · fs /2, (n + 1) · fs /2] to pass through. The impulse response of such a band-pass filter can be
obtained by amplitude modulating a low-pass filter, or by subtracting two low-pass filters:
sin πtfs /2 2n + 1 sin πt(n + 1)fs sin πtnfs
 
h(t) = fs · cos 2πtfs = (n + 1)fs − nfs .
πtfs /2 4 πt(n + 1)fs πtnfs

71 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Spectrum of a periodic signal

A signal x(t) that is periodic with frequency fp can be factored into a

single period ẋ(t) convolved with an impulse comb p(t). This
corresponds in the frequency domain to the multiplication of the
spectrum of the single period with a comb of impulses spaced fp apart.
x(t) ẋ(t) p(t)

= ∗
... ... ... ...
−1/fp 0 1/fp t t −1/fp 0 1/fp t
X(f ) Ẋ(f ) P (f )

= ·
... ...
−fp 0 fp f f −fp 0 fp f

72 / 231
Spectrum of a sampled signal

A signal x(t) that is sampled with frequency fs has a spectrum that is

periodic with a period of fs .
x(t) s(t) x̂(t)

· =
... ... ... ...
0 t −1/fs 0 1/fs t −1/fs 0 1/fs t
X(f ) S(f ) X̂(f )

∗ =
... ... ... ...
0 f −fs fs f −fs 0 fs f

73 / 231
Continuous vs discrete Fourier transform

I Sampling a continuous signal makes its spectrum periodic

I A periodic signal has a sampled spectrum
We sample a signal x(t) with fs , getting x̂(t). We take n consecutive
samples of x̂(t) and repeat these periodically, getting a new signal ẍ(t)
with period n/fs . Its spectrum Ẍ(f ) is sampled (i.e., has non-zero
value) at frequency intervals fs /n and repeats itself with a period fs .
Now both ẍ(t) and its spectrum Ẍ(f ) are finite vectors of length n.
ẍ(t) Ẍ(f )

... ... ... ...

−n/fs fs−1 0 fs−1 n/fs t −fs −fs /n 0 fs /n fs f

74 / 231
If x(t) has period tp = n · ts , then after sampling it at rate ts we have
∞
X ∞ n−1
X X
ẍ(t) = x(t)·s(t) = ts · xi ·δ(t−ts ·i) = ts · xi ·δ(t−ts ·(i+nl))
i=−∞ l=−∞ i=0

and the Fourier transform of that is

Z ∞
F{ẍ(t)}(f ) = Ẍ(f ) = ẍ(t) · e−2π jf t dt
−∞
∞ n−1 ∞ n−1
X X f X f X f
= ts · xi ·e−2π j fs ·(i+nl) = ts · e−2π j fs ·nl · xi ·e−2π j fs ·i
l=−∞ i=0 l=−∞ i=0
| {z }
1 l
P
ts n l δ(f − n fs )

e±2π jixa =
P∞ 1
P∞ n
Recall that i=−∞ |a| i=−∞ δ(x − i/a) and map x = f , a = fs and i = l.
f f f k
After substituting k := fp = fs n, i.e. fs = n and f = kfp
∞ n−1
1 X X ki
Ẍ(kfp ) = · δ(kfp − lfp ) · xi · e−2π j n
n i=0
l=−∞
|  {z } | {z }
Xk
δ(0) if k ∈ Z

=
0

if k ∈
/Z
Show that Xk = Xk±n for all k ∈ Z. 75 / 231
Discrete Fourier Transform (DFT)

n−1 n−1
X ik 1 X ik
Xk = xi · e−2π j n xk = · Xi · e2π j n
i=0
n i=0
The n-point DFT multiplies a vector with an n × n matrix
1 1 1 1 ··· 1
 
1
−2π j n 2
−2π j n 3
−2π j n −2π j n−1

 1 e e e ··· e n


2 4 6 2(n−1)
−2π j n −2π j n −2π j n −2π j n
 

 1 e e e ··· e 

3(n−1)
Fn = 
 1 e−2π j n
3
e−2π j n
6
e−2π j n
9
··· e −2π j n 

 
 .. .. .. .. .. .. 

 . . . . . .


n−1 2(n−1) 3(n−1) (n−1)(n−1)
1 e−2π j n e−2π j n e−2π j n · · · e−2π j n

x0 X0 X0 x0
       
 x1   X1   X1   x1 
 x   X 
    1  X   x 
   
Fn ·  2  =  2  , · Fn∗ ·  2  =  2 
 .   .  n  .   . 
 ..   ..   ..   .. 
xn−1 Xn−1 Xn−1 xn−1

76 / 231
Discrete Fourier Transform visualized

     
x0 X0
x1 X1
     
     
     


 
  x2  
  X2 

   x3   X3 
· =
     
x4 X4
 
     
     


 
  x5  
  X5 



 
  x6  
  X6 

x7 X7

The n-point DFT of a signal {xi } sampled at frequency fs contains in

the elements X0 to Xn/2 of the resulting frequency-domain vector the
frequency components 0, fs /n, 2fs /n, 3fs /n, . . . , fs /2, and contains in
Xn−1 downto Xn/2 the corresponding negative frequencies. Note that
for a real-valued input vector, both X0 and Xn/2 will be real, too.
Why is there no phase information recovered at fs /2?

77 / 231
Inverse DFT visualized

 
   
  X0 x0
X1 x1
    
    
X2 x2
    
    
1 
X3   x3 
· · =
  
8  X4  
  x4 

X5 x5
    
    
X6 x6
    
 
X7 x7
 

78 / 231
Fast Fourier Transform (FFT)

  n−1
X ik
Fn {xi }n−1
i=0 = xi · e−2π j n
k
i=0
n −1 n −1
2 2
X ik
−2π j n/2 k X ik
−2π j n/2
= x2i · e + e−2π j n x2i+1 · e
i=0 i=0
n −1  n −1 
  k

2 −2π j n 2 n
 F n2 {x2i }i=0 k + e · F n2 {x2i+1 }i=0 , k<

 2
k
=  n −1   n −1 
k
2
 F n {x2i }i=0 + e−2π j n · F n2 {x2i+1 }i=0
2
k≥ n


2
, 2
k− n
2
k− n
2

The DFT over n-element vectors can be reduced to two DFTs over
n/2-element vectors plus n multiplications and n additions, leading to
log2 n rounds and n log2 n additions and multiplications overall,
compared to n2 for the equivalent matrix multiplication.
A high-performance FFT implementation in C with many processor-specific optimizations and
support for non-power-of-2 sizes is available at http://www.fftw.org/.

79 / 231
Efficient real-valued FFT
The symmetry properties of the Fourier transform applied to the discrete
Fourier transform {Xi }n−1 n−1
i=0 = Fn {xi }i=0 have the form

∀i : xi = <(xi ) ⇐⇒ ∀i : Xn−i = Xi∗

∀i : xi = j · =(xi ) ⇐⇒ ∀i : Xn−i = −Xi∗

These two symmetries, combined with the linearity of the DFT, allows us to
calculate two real-valued n-point DFTs

{Xi0 }n−1 0 n−1

i=0 = Fn {xi }i=0 {Xi00 }n−1 00 n−1
i=0 = Fn {xi }i=0

simultaneously in a single complex-valued n-point DFT, by composing its input

as
xi = x0i + j · x00i
and decomposing its output as
1 1
Xi0 = ∗
(Xi + Xn−i ) Xi00 = ∗
(Xi − Xn−i )
2 2j
where Xn = X0 .
To optimize the calculation of a single real-valued FFT, use this trick to calculate the two half-size
real-value FFTs that occur in the first round.
80 / 231
Fast complex multiplication

Calculating the product of two complex numbers as

(a + jb) · (c + jd) = (ac − bd) + j(ad + bc)

involves four (real-valued) multiplications and two additions.

The alternative calculation
α = a(c + d)
(a + jb) · (c + jd) = (α − β) + j(α + γ) with β = d(a + b)
γ = c(b − a)

provides the same result with three multiplications and five additions.
The latter may perform faster on CPUs where multiplications take three
or more times longer than additions.
This “Karatsuba multiplication” is most helpful on simpler microcontrollers. Specialized
signal-processing CPUs (DSPs) feature 1-clock-cycle multipliers. High-end desktop processors use
pipelined multipliers that stall where operations depend on each other.

81 / 231
FFT-based convolution
Calculating the convolution of two finite sequences {xi }m−1 n−1
i=0 and {yi }i=0
of lengths m and n via
min{m−1,i}
X
zi = xj · yi−j , 0≤i<m+n−1
j=max{0,i−(n−1)}

takes mn multiplications.
Can we apply the FFT and the convolution theorem to calculate the
convolution faster, in just O(m log m + n log n) multiplications?

{zi } = F −1 (F{xi } · F {yi })

There is obviously no problem if this condition is fulfilled:

{xi } and {yi } are periodic, with equal period lengths
In this case, the fact that the DFT interprets its input as a single period
of a periodic signal will do exactly what is needed, and the FFT and
inverse FFT can be applied directly as above.
82 / 231
In the general case, measures have to be taken to prevent a wrap-over:
A B F−1[F(A)⋅F(B)]

A’ B’ F−1[F(A’)⋅F(B’)]

Both sequences are padded with zero values to a length of at least m + n − 1.

This ensures that the start and end of the resulting sequence do not overlap.

83 / 231
Zero padding is usually applied to extend both sequence lengths to the
next higher power of two (2dlog2 (m+n−1)e ), which facilitates the FFT.
With a causal sequence, simply append the padding zeros at the end.
With a non-causal sequence, values with a negative index number are
wrapped around the DFT block boundaries and appear at the right end.
In this case, zero-padding is applied in the center of the block, between
the last and first element of the sequence.
Thanks to the periodic nature of the DFT, zero padding at both ends has
the same effect as padding only at one end.

If both sequences can be loaded entirely into RAM, the FFT can be
applied to them in one step. However, one of the sequences might be too
large for that. It could also be a realtime waveform (e.g., a telephone
signal) that cannot be delayed until the end of the transmission.
In such cases, the sequence has to be split into shorter blocks that are
separately convolved and then added together with a suitable overlap.

84 / 231
Each block is zero-padded at both ends and then convolved as before:

∗ ∗ ∗

= = =

The regions originally added as zero padding are, after convolution, aligned to
overlap with the unpadded ends of their respective neighbour blocks. The
overlapping parts of the blocks are then added together.

85 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Deconvolution
A signal u(t) was distorted by convolution with a known impulse
response h(t) (e.g., through a transmission channel or a sensor problem).
The “smeared” result s(t) was recorded.
Can we undo the damage and restore (or at least estimate) u(t)?

∗ =

∗ =

86 / 231
The convolution theorem turns the problem into one of multiplication:
Z
s(t) = u(t − τ ) · h(τ ) · dτ

s = u∗h

F{s} = F{u} · F {h}

F{u} = F{s}/F{h}

u = F −1 {F{s}/F{h}}

In practice, we also record some noise n(t) (quantization, etc.):

Z
c(t) = s(t) + n(t) = u(t − τ ) · h(τ ) · dτ + n(t)

Problem – At frequencies f where F{h}(f ) approaches zero, the noise

will be amplified (potentially enormously) during deconvolution:

ũ = F −1 {F{c}/F{h}} = u + F −1 {F{n}/F{h}}

87 / 231
Typical workarounds:
I Modify the Fourier transform of the impulse response, such that
|F{h}(f )| >  for some experimentally chosen threshold .
I If estimates of the signal spectrum |F{s}(f )| and the noise
spectrum |F{n}(f )| can be obtained, then we can apply the
“Wiener filter” (“optimal filter”)

|F{s}(f )|2
W (f ) =
|F{s}(f )|2 + |F{n}(f )|2

before deconvolution:

ũ = F −1 {W · F {c}/F{h}}

88 / 231
Exercise 13: Use MATLAB to deconvolve the blurred stars from slide 31.
The files stars-blurred.png with the blurred-stars image and stars-psf.png
with the impulse response (point-spread function) are available on the course-
material web page. You may find the MATLAB functions imread, double,
imagesc, circshift, fft2, ifft2 of use.
Try different ways to control the noise (slide 88) and distortions near the margins
(windowing). [The MATLAB image processing toolbox provides ready-made
“professional” functions deconvwnr, deconvreg, deconvlucy, edgetaper, for
such tasks. Do not use these, except perhaps to compare their outputs with the
results of your own attempts.]
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
 Vowel "A" sung at varying pitch
8000

7000

6000
Frequency (Hz)

5000

4000

3000

2000

1000

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time
[w,fs, bits] = auread('sing.au');
specgram(w,2048,fs);
ylim([0 8e3]); xlim([0 4.5]);
saveas(gcf, 'sing.eps', 'eps2c');

89 / 231
 Different vovels at constant pitch
8000

7000

6000
Frequency (Hz)

5000

4000

3000

2000

1000

0
0.5 1 1.5 2 2.5 3 3.5 4
Time

90 / 231
 3.5 -105

-110
3
-115

-120

Power/frequency (dB/Hz)
2.5
Frequency (MHz)

-125
2
-130

1.5 -135

-140
1
-145
0.5 -150

-155
0
20 40 60 80 100 120
Time (ms)
f = fopen('iq-fm-97M-3.6M.dat', 'r', 'ieee-le');
c = fread(f, [2,inf], '*float32');
fclose(f);
z = c(1,:) + j*c(2,:);
fs = 3.6e6; % IQ sampling frequency
fciq = 97e6; % center frequency of IQ downconverter
spectrogram(double(z(1:5e5)), 1024, 512, 1024, fs, 'yaxis');
colormap(gray)
Spectral estimation
cos(2 *[0:15]/16*4)
1 12
DTFT mag
10 DFT mag
0.5
8

0 6

4
-0.5
2

-1 0
0 5 10 15 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

cos(2 *[0:15]/16*4.2)
1 12
DTFT mag
10 DFT mag
0.5
8

0 6

4
-0.5
2

-1 0
0 5 10 15 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)
91 / 231
We introduced the DFT as a special case of the continuous Fourier
transform, where the input is sampled and periodic.
If the input is sampled, but not periodic, the DFT can still be used to
calculate an approximation of the Fourier transform of the original
continuous signal. However, there are two effects to consider. They are
particularly visible when analysing pure sine waves.
Sine waves whose frequency is a multiple of the base frequency (fs /n) of
the DFT are identical to their periodic extension beyond the size of the
DFT. They are, therefore, represented exactly by a single sharp peak in
the DFT. All their energy falls into one single frequency “bin” in the
DFT result.
Sine waves with other frequencies, which do not match exactly one of the
output frequency bins of the DFT, are still represented by a peak at the
output bin that represents the nearest integer multiple of the DFT’s base
frequency. However, such a peak is distorted in two ways:
I Its amplitude is lower (down to 63.7%).
I Much signal energy has “leaked” to other frequencies.

92 / 231
 35

30

25

20

15

10

0
0 16
5 10 15 15.5
20 25 30 15
input freq.
DFT index
The leakage of energy to other frequency bins not only blurs the estimated spectrum.
The peak amplitude also changes significantly as the frequency of a tone changes from
that associated with one output bin to the next, a phenomenon known as scalloping.
In the above graphic, an input sine wave gradually changes from the frequency of bin
15 to that of bin 16 (only positive frequencies shown).

93 / 231
Windowing
Sine wave Discrete Fourier Transform
1 300

200
0
100

−1 0
0 200 400 0 200 400
Sine wave multiplied with window function Discrete Fourier Transform
1 100

0 50

−1 0
0 200 400 0 200 400
94 / 231
The reason for the leakage and scalloping losses is easy to visualize with the
help of the convolution theorem:
The operation of cutting a sequence of the size of the DFT input vector out of
a longer original signal (the one whose continuous Fourier spectrum we try to
estimate) is equivalent to multiplying this signal with a rectangular function.
This destroys all information and continuity outside the “window” that is fed
into the DFT.
Multiplication with a rectangular window of length T in the time domain is
equivalent to convolution with sin(πf T )/(πf T ) in the frequency domain.
The subsequent interpretation of this window as a periodic sequence by the
DFT leads to sampling of this convolution result (sampling meaning
multiplication with a Dirac comb whose impulses are spaced fs /n apart).
Where the window length was an exact multiple of the original signal period,
sampling of the sin(πf T )/(πf T ) curve leads to a single Dirac pulse, and the
windowing causes no distortion. In all other cases, the effects of the convolution
become visible in the frequency domain as leakage and scalloping losses.

95 / 231
Some better window functions

0.8

0.6

0.4

0.2
Rectangular window
Triangular window
0
Hann window
Hamming window

0 0.2 0.4 0.6 0.8 1

All these functions are 0 outside the interval [0,1].
96 / 231
 cos(2 *[0:15]/16*4.2)
1 12
DTFT mag
10 DFT mag
0.5
8

0 6

4
-0.5
2

-1 0
0 5 10 15 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

cos(2 *[0:15]/16*4.2).*hann(16)
1 12
DTFT mag
10 DFT mag
0.5
8

0 6

4
-0.5
2

-1 0
0 5 10 15 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

97 / 231
 Rectangular window (64−point) Triangular window

20 20
Magnitude (dB)

Magnitude (dB)
0 0

−20 −20

−40 −40

−60 −60
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)
Hann window Hamming window

20 20
Magnitude (dB)

Magnitude (dB)
0 0

−20 −20

−40 −40

−60 −60
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

98 / 231
Numerous alternatives to the rectangular window have been proposed
that reduce leakage and scalloping in spectral estimation. These are
vectors multiplied element-wise with the input vector before applying the
DFT to it. They all force the signal amplitude smoothly down to zero at
the edge of the window, thereby avoiding the introduction of sharp jumps
in the signal when it is extended periodically by the DFT.
Three examples of such window vectors {wi }n−1
i=0 are:
Triangular window (Bartlett window):
 
 i 
wi = 1 − 1 −

n/2 

Hann window (raised-cosine window, Hanning window):

 
i
wi = 0.5 − 0.5 × cos 2π
n−1

Hamming window:
 
i
wi = 0.54 − 0.46 × cos 2π
n−1

99 / 231
Does zero padding increase DFT resolution?
The two figures below show two spectra of the 16-element sequence

si = cos(2π · 3i/16) + cos(2π · 4i/16), i ∈ {0, . . . , 15}.

The left plot shows the DFT of the windowed sequence

xi = si · wi , i ∈ {0, . . . , 15}

and the right plot shows the DFT of the zero-padded windowed sequence

si · wi , i ∈ {0, . . . , 15}
x0i =
0, i ∈ {16, . . . , 63}

where wi = 0.54 − 0.46 × cos (2πi/15) is the Hamming window.

DFT without zero padding DFT with 48 zeros appended to window

4 4

2 2

0 0
0 5 10 15 0 20 40 60
100 / 231
cos(2 *[0:15]/16*3.3) + cos(2 *[0:15]/16*4)
2 8
DTFT mag
DFT mag
1 6

0 4

-1 2

-2 0
0 5 10 15 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

zero-padded to 64 samples
2 8
DTFT mag
DFT mag
1 6

0 4

-1 2

-2 0
0 20 40 60 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

101 / 231
cos(2 *[0:15]/16*3.3) + cos(2 *[0:15]/16*4)
2 8
DTFT mag
DFT mag
1 6

0 4

-1 2

-2 0
0 5 10 15 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

with 64 actual samples

2 35
DTFT mag
30 DFT mag
1
25

20
0
15

10
-1
5

-2 0
0 20 40 60 - -¾ -½ -¼ 0 ¼ ½ ¾
time-domain samples DTFT frequency (1 period)

102 / 231
Applying the discrete Fourier transform (DFT) to an n-element long
real-valued sequence samples the DTFT of that sequence at n/2 + 1
discrete frequencies.
The DTFT spectrum has already been distorted by multiplying the
(hypothetically longer) signal with a windowing function that limits its
length to n non-zero values and forces the waveform down to zero
outside the window. Therefore, appending further zeros outside the
window will not affect the DTFT.
The frequency resolution of the DFT is the sampling frequency divided by
the block size of the DFT. Zero padding can therefore be used to increase
the frequency resolution of the DFT, to sample the DTFT at more
places. But that does not change the limit imposed on the frequency
resolution (i.e., blurriness) of the DTFT by the length of the window.
Note that zero padding does not add any additional information to the
signal. The DTFT has already been “low-pass filtered” by being
convolved with the spectrum of the windowing function. Zero padding in
the time domain merely causes the DFT to sample the same underlying
DTFT spectrum at a higher resolution, thereby making it easier to
visually distinguish spectral lines and to locate their peak more precisely.

103 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
FIR filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Digital filters

Filter: supresses (removes, attenuates) unwanted signal components.

I low-pass filter – suppress all frequencies above a cut-off frequency
I high-pass filter – suppress all frequencies below a cut-off frequency,
including DC (direct current = 0 Hz)
I band-pass filter – suppress signals outside a frequency interval (=
passband)
I band-stop filter (aka: band-reject filter) – suppress signals inside a single
frequency interval (= stopband)
I notch filter – narrow band-stop filter, ideally suppressing only a single
frequency

For digital filters, we also distinguish

I finite impulse response (FIR) filters
I infinite impulse response (IIR) filters

depending on how far their memory reaches back in time.

104 / 231
Window-based design of FIR filters

Recall that the ideal continuous low-pass filter with cut-off frequency fc
has the frequency characteristic
  
1 if |f | < fc f
H(f ) = = rect
0 if |f | > fc 2fc

and the impulse response

sin 2πtfc
h(t) = 2fc = 2fc · sinc(2fc · t).
2πtfc
Sampling this impulse response with the sampling frequency fs of the
signal to be processed will lead to a periodic frequency characteristic,
that matches the periodic spectrum of the sampled signal.
There are two problems though:
I the impulse response is infinitely long
I this filter is not causal, that is h(t) 6= 0 for t < 0

105 / 231
Solutions:
I Make the impulse response finite by multiplying the sampled h(t)
with a windowing function
I Make the impulse response causal by adding a delay of half the
window size
The impulse response of an n-th order low-pass filter is then chosen as

sin[2π(i − n/2)fc /fs ]

hi = 2fc /fs · · wi
2π(i − n/2)fc /fs

where {wi } is a windowing sequence, such as the Hamming window

wi = 0.54 − 0.46 × cos (2πi/n)

with wi = 0 for i < 0 and i > n.

Note that for fc = fs /4, we have hi = 0 for all even values of i. Therefore, this special case
requires only half the number of multiplications during the convolution. Such “half-band” FIR
filters are used, for example, as anti-aliasing filters wherever a sampling rate needs to be halved.

106 / 231
FIR low-pass filter design example
z Plane Impulse Response
1
1
Imaginary Part

Amplitude
0.5
30 0.5
0
−0.5
0
−1
−1 0 1 0 10 20 30
Real Part n (samples)
0 0

Phase (degrees)
Magnitude (dB)

−20 −500

−40 −1000

−60 −1500
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

order: n = 30, cutoff frequency (−6 dB): fc = 0.25 × fs /2, window: Hamming

107 / 231
Filter performance

An ideal filter has a gain of 1 in the pass-band and a gain of 0 in the stop
band, and nothing in between.
A practical filter will have
I frequency-dependent gain near 1 in the passband
I frequency-dependent gain below a threshold in the stopband
I a transition band between the pass and stop bands
We truncate the ideal, infinitely-long impulse response by multiplication
with a window sequence.
In the frequency domain, this will convolve the rectangular frequency
response of the ideal low-pass filter with the frequency characteristic of
the window.
The width of the main lobe determines the width of the transition band,
and the side lobes cause ripples in the passband and stopband.

108 / 231
Low-pass to band-pass filter conversion (modulation)

To obtain a band-pass filter that attenuates all frequencies f outside the

range fl < f < fh , we first design a low-pass filter with a cut-off
frequency (fh − fl )/2. We then multiply its impulse response with a sine
wave of frequency (fh + fl )/2, effectively amplitude modulating it, to
shift its centre frequency. Finally, we apply a window function:

sin[π(i − n/2)(fh − fl )/fs ]

hi = (fh − fl )/fs · · cos[πi(fh + fl )/fs ] · wi
π(i − n/2)(fh − fl )/fs

H(f )
= ∗

−fh −fl 0 fl fh f − fh −f
2
l fh −fl
2
f − fh +f
2
l
0 fh +fl
2
f

109 / 231
Low-pass to high-pass filter conversion (freq. inversion)
In order to turn the spectrum X(f ) of a real-valued signal xi sampled at
fs into an inverted spectrum X 0 (f ) = X(fs /2 − f ), we merely have to
shift the periodic spectrum by fs /2:
X 0 (f ) X(f )

= ∗
... ... ... ...
−fs 0 fs f −fs 0 fs f − f2s 0 fs
2
f
This can be accomplished by multiplying the sampled sequence xi with
yi = cos πfs t = cos πi, which is nothing but multiplication with the
sequence
. . . , 1, −1, 1, −1, 1, −1, 1, −1, . . .

So in order to design a discrete high-pass filter that attenuates all

frequencies f outside the range fc < |f | < fs /2, we merely have to
design a low-pass filter that attenuates all frequencies outside the range
−fc < f < fc , and then multiply every second value of its impulse
response with −1.
110 / 231
Linear phase filters

A filter where the Fourier transform H(f ) of its impulse response h(t) is
real-valued will not affect the phase of the filtered signal at any
frequency. Only the amplitudes will be affected.

∀f ∈ R : H(f ) ∈ R ⇐⇒ ∀t ∈ R : h(t) = [h(−t)]∗

A phase-neutral filter with a real-valued frequency response will have an

even impulse response, and will therefore usually be non-causal.
To make such a filter causal, we have to add a delay ∆t (half the length
of the impulse response). This corresponds to multiplication with
e−2π jf ∆t in the frequency domain:

h(t − ∆t) •−◦ H(f ) · e−2π jf ∆t

Filters that delay the phase of a signal at each frequency by the time ∆t
therefore add to the phase angle a value −2π jf ∆t, which increases
linearly with f . They are therefore called linear-phase filters.
This is the closest one can get to phase-neutrality with causality.

111 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Finite impulse response (FIR) filter

M
X
yn = bm · xn−m
m=0
M = 3:
xn xn−1 xn−2 xn−3
z −1 z −1 z −1
b0 b1 b2 b3

yn
(see slide 25)

Transposed implementation:
xn

b3 b2 b1 b0

z −1 z −1 z −1
yn
112 / 231
Infinite impulse response (IIR) filter
N
X M
X
ak · yn−k = bm · xn−m Usually normalize: a0 = 1
k=0 m=0
M N
!
X X
yn = bm · xn−m − ak · yn−k /a0
m=0 k=1

Direct form I implementation:

xn b0 a−1
0 yn

z −1 z −1
b1 −a1
xn−1 yn−1

z −1 z −1
b2 −a2
xn−2 yn−2

z −1 z −1
b3 −a3
xn−3 yn−3
113 / 231
Infinite impulse response (IIR) filter – direct form II

M N
!
X X
yn = bm · xn−m − ak · yn−k /a0
m=0 k=1

Direct form II: Transposed direct form II:

xn a−1
0 b0 yn xn b0 a−1
0 yn

z −1 z −1
−a1 b1
b1 −a1

z −1 z −1
−a2 b2
b2 −a2

z −1 z −1
−a3 b3
b3 −a3

114 / 231
Polynomial representation of sequences

We can represent sequences {xn } as polynomials:

∞
X
X(v) = xn v n
n=−∞

Example of polynomial multiplication:

(1 + 2v + 3v 2 ) · (2 + 1v)
2
2 + 4v + 6v
+ 1v + 2v 2 + 3v 3
= 2 + 5v + 8v 2 + 3v 3

Compare this with the convolution of two sequences (in MATLAB):

conv([1 2 3], [2 1]) equals [2 5 8 3]

115 / 231
Convolution of sequences is equivalent to polynomial multiplication:
∞
X
{hn } ∗ {xn } = {yn } ⇒ yn = hk · xn−k
k=−∞
↓ ↓
∞ ∞
! !
X X
n n
H(v) · X(v) = hn v · xn v
n=−∞ n=−∞
∞
X ∞
X
= hk · xn−k · v n
n=−∞ k=−∞

Note how the Fourier transform of a sequence can be accessed easily

from its polynomial form:
∞
X
X(e− jω̇ ) = xn e− jω̇n
n=−∞

116 / 231
Example of polynomial division:
∞
1 X
= 1 + av + a2 v 2 + a3 v 3 + · · · = an v n
1 − av n=0

1 + av + a2 v 2 + ···
1 − av 1
1 − av
xn yn av
av − a2 v 2
a2 v 2
v
a a2 v 2 − a3 v 3
yn−1 ···

Rational functions (quotients of two polynomials) can provide a

convenient closed-form representations for infinitely-long exponential
sequences, in particular the impulse responses of IIR filters.

117 / 231
The z-transform

The z-transform of a sequence {xn } is defined as:

∞
X
X(z) = xn z −n
n=−∞

Note that this differs only in the sign of the exponent from the polynomial representation discussed
on the preceeding slides.

Recall that the above X(z) is exactly the factor with which an
exponential sequence {z n } is multiplied, if it is convolved with {xn }:

{z n } ∗ {xn } = {yn }

∞
X ∞
X
⇒ yn = z n−k n
xk = z · z −k xk = z n · X(z)
k=−∞ k=−∞

118 / 231
The z-transform defines for each sequence a continuous complex-valued
surface over the complex plane C.
For finite sequences, its value is defined across the entire complex plane
(except possibly at z = 0 or |z| = ∞).
For infinite sequences, it can be shown that the z-transform converges
only for the region
   
 xn+1 
 < |z| < lim  xn+1 
 
lim 
n→∞ xn  n→−∞  xn 
The z-transform identifies a sequence unambiguously only in conjunction with a given region of
convergence. In other words, there exist different sequences, that have the same expression as their
z-transform, but that converge for different amplitudes of z.

The z-transform is a generalization of the discrete-time Fourier

transform, which it contains on the complex unit circle (|z| = 1):
∞
X
jω̇
t−1
s · F {x̂(t)}(f ) = X(e ) = xn e− jω̇n
n=−∞

where ω̇ = 2π ffs .

119 / 231
Properties of the z-transform
If X(z) is the z-transform of {xn }, we write here {xn } •−◦ X(z).
If {xn } •−◦ X(z) and {yn } •−◦ Y (z), then:

Linearity:

{axn + byn } •−◦ aX(z) + bY (z)

Convolution:

{xn } ∗ {yn } •−◦ X(z) · Y (z)

Time shift:

{xn+k } •−◦ z k X(z)

Remember in particular: delaying by one sample is multiplication with z −1 .

120 / 231
Time reversal:

{x−n } •−◦ X(z −1 )

Multiplication with exponential:

{a−n xn } •−◦ X(az)

Complex conjugate:

{x∗n } •−◦ X ∗ (z ∗ )

Real/imaginary value:
1
{<{xn }} •−◦ (X(z) + X ∗ (z ∗ ))
2
1
{={xn }} •−◦ (X(z) − X ∗ (z ∗ ))
2j
Initial value:

x0 = lim X(z) if xn = 0 for all n < 0

z→∞

121 / 231
Some example sequences and their z-transforms:

xn X(z)
δn 1
z 1
un =
z−1 1 − z −1
z 1
an un =
z−a 1 − az −1
z
nun
(z − 1)2
z(z + 1)
n2 un
(z − 1)3
z
ean un
z − ea
n − 1 a(n−k)
 
1
e un−k
k−1 (z − ea )k
z 2 sin(ϕ) + z sin(ω̇ − ϕ)
sin(ω̇n + ϕ)un
z 2 − 2z cos(ω̇) + 1
122 / 231
Example:
What is the z-transform of the impulse response {hn }
of the discrete system yn = xn + ayn−1 ?
xn yn
yn = xn + ayn−1
Y (z) = X(z) + az −1 Y (z) z −1
a
Y (z) − az −1 Y (z) = X(z) yn−1
Y (z)(1 − az −1 ) = X(z)
Y (z) 1 z
= =
X(z) 1 − az −1 z−a
Since {yn } = {hn } ∗ {xn }, we have Y (z) = H(z) · X(z) and therefore

Y (z) z
H(z) = = = 1 + az −1 + a2 z −2 + · · ·
X(z) z−a

where polynomial long division returns the causal impulse response

h0 = 1, h1 = a, h2 = a2 , . . . , hn = an for all n ≥ 0

We have applied here the linearity of the z-transform, and its time-shift and convolution properties.
123 / 231
z-transform of recursive filter structures

Consider the discrete system defined by xn b0 a−1

0 yn

k m z −1 z −1
X X b1 −a1
al · yn−l = bl · xn−l xn−1 yn−1
l=0 l=0
z −1 z −1
··· ···
or equivalently ··· ···

k m z −1 z −1
X X bm −ak
a0 yn + al · yn−l = bl · xn−l xn−m yn−k
l=1 l=0
m k
!
X X
yn = a−1
0 · bl · xn−l − al · yn−l
l=0 l=1

What is the z-transform H(z) of its impulse response {hn }, where

{yn } = {hn } ∗ {xn }?

124 / 231
Using the linearity and time-shift property of the z-transform:

k
X m
X
al · yn−l = bl · xn−l
l=0 l=0
k
X m
X
al z −l · Y (z) = bl z −l · X(z)
l=0 l=0
k
X m
X
Y (z) al z −l = X(z) bl z −l
l=0 l=0
Pm
Y (z) bl z −l
H(z) = = Pkl=0
X(z) l=0 al z
−l

b0 + b1 z −1 + b2 z −2 + · · · + bm z −m
H(z) =
a0 + a1 z −1 + a2 z −2 + · · · + ak z −k

125 / 231
The z-transform of the impulse re- xn b0 a−1
0 yn
sponse {hn } of the causal LTI system
defined by z −1 z −1
b1 −a1
xn−1 yn−1
k
X m
X
al · yn−l = bl · xn−l z −1
··· ···
z −1
l=0 l=0 ··· ···

with {yn } = {hn } ∗ {xn } is the z −1

z −1
bm −ak
rational function xn−m yn−k

b0 + b1 z −1 + b2 z −2 + · · · + bm z −m
H(z) =
a0 + a1 z −1 + a2 z −2 + · · · + ak z −k
(bm 6= 0, ak =
6 0) which can also be written as
Pm
z k l=0 bl z m−l z k b0 z m + b1 z m−1 + b2 z m−2 + · · · + bm
H(z) = k
= · .
z m a0 z k + a1 z k−1 + a2 z k−2 + · · · + ak
P
z m l=0 al z k−l

H(z) has m zeros and k poles at non-zero locations in the z plane, plus
k − m zeros (if k > m) or m − k poles (if m > k) at z = 0.

126 / 231
This function can be converted into the form
m
Y m
Y
(1 − cl · z −1 ) (z − cl )
b0 l=1 b0 k−m l=1
H(z) = · k = ·z · k
a0 Y a0 Y
−1
(1 − dl · z ) (z − dl )
l=1 l=1

where the cl are the non-zero positions of zeros (H(cl ) = 0) and the dl
are the non-zero positions of the poles (i.e., z → dl ⇒ |H(z)| → ∞) of
H(z). Except for a constant factor, H(z) is entirely characterized by the
position of these zeros and poles.
On the unit circle z = e jω̇ , H(e jω̇ ) is the discrete-time Fourier transform
of {hn } (ω̇ = πf / f2s ). The DTFT amplitude can also be expressed in
terms of the relative position of e jω̇ to the zeros and poles:
  Qm
jω̇
 b0  |e jω̇ − cl |
|H(e )| =   · Qkl=1
a0 l=1 |e
jω̇ − d |
l

127 / 231
Example: a single-pole filter
Consider this IIR filter: Its z-transform

xn 0.8 yn 0.8 0.8z

H(z) = =
1 − 0.2 · z −1 z − 0.2

z −1 has one pole at z = d1 = 0.2 and one

0.2 zero at z = 0.
yn−1
a0 = 1, a1 = −0.2, Amplitude |H(z)|:
b0 = 0.8 2
1.75
xn = δn ⇒ yn = 1.5
Impulse Response 1.25
0.8
1
0.6 0.75
Amplitude

0.5
0.4
0.25
0.2 0
1
0.5 1
0 0 0.5
0 2 4 0
n (samples) −0.5 −0.5
−1 −1
imaginary real 128 / 231
 0.8 0.8z
H(z) = 1−0.2·z −1 = z−0.2 (cont’d) Magnitude Response
1

0.95
2
1.75
0.9
1.5

Magnitude
1.25
0.85
1
0.75
0.8
0.5
0.25 0.75
0
1
0.5 1 0.7
0 0.5
−0.5 0
−0.5
−1 −1
imaginary real 0 0.2 0.4 0.6 0.8
Normalized Frequency (×π rad/sample)

Run this LTI filter at sampling frequency fs and test it with sinusoidial
input (frequency f , amplitude 1): xn = cos(2πf n/fs )
Output: yn = A(f ) · cos(2πf n/fs + θ(f ))
What are the gain A(f ) and phase delay θ(f ) at frequency f ?
Answer: A(f ) = |H(e j2πf /fs )|
={H(e jπf /fs )}
θ(f ) = ∠H(e j2πf /fs ) = tan−1
<{H(e jπf /fs )}
Example: fs = 8 kHz, f = 2 kHz (normalized frequency f / f2s = 0.5) ⇒ Gain A(2 kHz) =
  0.8 j(− j−0.2)   0.8−0.16 j  q 0.82 +0.162
|H(e jπ/2 )| = |H( j)| =  0.8 j  = 

= = = 0.784. . .
j−0.2 ( j−0.2)(− j−0.2) 1+0.04 1.042

129 / 231
Visual verification in MATLAB:
n = 0:15; x
fs = 8000; 1.5 y (time domain)
f = 1500; y (z−transform)
x = cos(2*pi*f*n/fs);
b = [0.8]; a = [1 -0.2]; 1
y1 = filter(b,a,x);
z = exp(j*2*pi*f/fs);
H = 0.8*z/(z-0.2);
A = abs(H); 0.5
theta = atan(imag(H)/real(H));
y2 = A*cos(2*pi*f*n/fs+theta);
plot(n, x, 'bx-', ... 0
n, y1, 'go-', ...
n, y2, 'r+-')
legend('x', ...
'y (time domain)', ... −0.5
'y (z-transform)')
ylim([-1.1 1.8])
−1
0 5 10 15
130 / 231
 z 1
H(z) = z−0.7 = 1−0.7·z −1 How do poles affect time domain?
z Plane Impulse Response
1 1
Imaginary Part

Amplitude
0 0.5

−1 0
−1 0 1 0 10 20 30
Real Part n (samples)
z 1
H(z) = z−0.9 = 1−0.9·z −1

z Plane Impulse Response

1 1
Imaginary Part

Amplitude
0 0.5

−1 0
−1 0 1 0 10 20 30
Real Part n (samples)
131 / 231
 z 1
H(z) = z−1 = 1−z −1

z Plane Impulse Response

1 1
Imaginary Part

Amplitude
0 0.5

−1 0
−1 0 1 0 10 20 30
Real Part n (samples)
z 1
H(z) = z−1.1 = 1−1.1·z −1

z Plane Impulse Response

1 20
Imaginary Part

Amplitude
0 10

−1 0
−1 0 1 0 10 20 30
Real Part n (samples)
132 / 231
 z2 1
H(z) = (z−0.9·e jπ/6 )·(z−0.9·e− jπ/6 )
= 1−1.8 cos(π/6)z −1 +0.92 ·z −2

z Plane Impulse Response

1 2
Imaginary Part

Amplitude
1
2
0
0

−1 −1
−1 0 1 0 10 20 30
Real Part n (samples)
z2 1
H(z) = (z−e jπ/6 )·(z−e− jπ/6 )
= 1−2 cos(π/6)z −1 +z −2

z Plane Impulse Response

1 5
Imaginary Part

Amplitude
2
0 0

−1 −5
−1 0 1 0 10 20 30
Real Part n (samples)
133 / 231
 z2 1 1
H(z) = (z−0.9·e jπ/2 )·(z−0.9·e− jπ/2 )
= 1−1.8 cos(π/2)z −1 +0.92 ·z −2 = 1+0.92 ·z −2

z Plane Impulse Response

1 1
Imaginary Part

Amplitude
2
0 0

−1 −1
−1 0 1 0 10 20 30
Real Part n (samples)
z 1
H(z) = z+1 = 1+z −1

z Plane Impulse Response

1 1
Imaginary Part

Amplitude
0 0

−1 −1
−1 0 1 0 10 20 30
Real Part n (samples)
134 / 231
IIR filter design goals

The design of a filter starts with specifying the desired parameters:

I The passband is the frequency range where we want to approximate
a gain of one.
I The stopband is the frequency range where we want to approximate
a gain of zero.
I The order of a filter is the number of poles it uses in the z-domain,
and equivalently the number of delay elements necessary to
implement it.
I Both passband and stopband will in practice not have gains of
exactly one and zero, respectively, but may show several deviations
from these ideal values, and these ripples may have a specified
maximum quotient between the highest and lowest gain.
I There will in practice not be an abrupt change of gain between
passband and stopband, but a transition band where the frequency
response will gradually change from its passband to its stopband
value.

135 / 231
IIR filter design techniques

The designer can then trade off conflicting goals such as: small transition
band, low order, low ripple amplitude or absence of ripples.
Design techniques for making these tradeoffs for analog filters (involving
capacitors, resistors, coils) can also be used to design digital IIR filters:
Butterworth filters: Have no ripples, gain falls monotonically across the pass
and p
transition band. Within the passband, the gain drops slowly down to
1 − 1/2 (−3 dB). Outside the passband, it drops asymptotically by a factor
2N per octave (N · 20 dB/decade).
Chebyshev type I filters: Distribute the gain error uniformly throughout the
passband (equiripples) and drop off monotonically outside.
Chebyshev type II filters: Distribute the gain error uniformly throughout the
stopband (equiripples) and drop off monotonically in the passband.
Elliptic filters (Cauer filters): Distribute the gain error as equiripples both in
the passband and stopband. This type of filter is optimal in terms of the
combination of the passband-gain tolerance, stopband-gain tolerance, and
transition-band width that can be achieved at a given filter order.

136 / 231
IIR filter design in MATLAB

The aforementioned filter-design techniques are implemented in the

MATLAB Signal Processing Toolbox in the functions butter, cheby1,
cheby2, and ellip. They output the coefficients an and bn of the
difference equation that describes the filter.
These can be applied with filter
to a sequence, or can be visual-
ized with zplane as poles/zeros in
the z-domain, with impz as an im-
pulse response, and with freqz as
an amplitude and phase spectrum.
The commands sptool and
fdatool provide interactive GUIs
to design digital filters.

MATLAB fdatool
137 / 231
Cascade of filter sections
Higher-order IIR filters can be numerically unstable (quantization noise).
A commonly used trick is to split a higher-order IIR filter design into a
cascade of l second-order (biquad) filter sections of the form:
xn b0 yn

z −1 b0 + b1 z −1 + b2 z −2
−a1 b1 H(z) =
1 + a1 z −1 + a2 z −2

z −1
−a2 b2
Filter sections H1 , H2 , . . . , Hl are then applied sequentially to the input
sequence, resulting in a filter
l l
Y Y bk,0 + bk,1 z −1 + bk,2 z −2
H(z) = Hk (z) =
1 + ak,1 z −1 + ak,2 z −2
k=1 k=1

Each section implements one pair of poles and one pair of zeros. Jackson’s algorithm for pairing
poles and zeros into sections: pick the pole pair closest to the unit circle, and place it into a
section along with the nearest pair of zeros; repeat until no poles are left.
138 / 231
Butterworth filter design example
z Plane Impulse Response
1 1
Imaginary Part

0.5

Amplitude
0.5
0

−0.5
0
−1
−1 0 1 0 10 20
Real Part n (samples)
0 0

Phase (degrees)
Magnitude (dB)

−20
−50
−40

−60 −100
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

order: 1, cutoff frequency (−3 dB): 0.25 × fs /2

139 / 231
Butterworth filter design example
z Plane Impulse Response
1 1
Imaginary Part

0.5

Amplitude
0.5
0

−0.5
0
−1
−1 0 1 0 10 20
Real Part n (samples)
0 0

Phase (degrees)
Magnitude (dB)

−20 −200

−40 −400

−60 −600
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

order: 5, cutoff frequency (−3 dB): 0.25 × fs /2

140 / 231
Chebyshev type I filter design example
z Plane Impulse Response
1 1
Imaginary Part

0.5

Amplitude
0.5
0

−0.5
0
−1
−1 0 1 0 10 20
Real Part n (samples)
0 0

Phase (degrees)
Magnitude (dB)

−20 −200

−40 −400

−60 −600
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

order: 5, cutoff frequency: 0.5 × fs /2, pass-band ripple: −3 dB

141 / 231
Chebyshev type II filter design example
z Plane Impulse Response
1 1
Imaginary Part

0.5

Amplitude
0.5
0

−0.5
0
−1
−1 0 1 0 10 20
Real Part n (samples)
0 100

Phase (degrees)
Magnitude (dB)

0
−20
−100
−40
−200

−60 −300
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

order: 5, cutoff frequency: 0.5 × fs /2, stop-band ripple: −20 dB

142 / 231
Elliptic filter design example
z Plane Impulse Response
1 1
Imaginary Part

0.5

Amplitude
0.5
0

−0.5
0
−1
−1 0 1 0 10 20
Real Part n (samples)
0 0

Phase (degrees)
Magnitude (dB)

−100
−20
−200
−40
−300

−60 −400
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

order: 5, cutoff frequency: 0.5 × fs /2, pass-band ripple: −3 dB, stop-band ripple: −20 dB
143 / 231
Notch filter design example
z Plane Impulse Response
1 1
Imaginary Part

0.5

Amplitude
0.5
0

−0.5
0
−1
−1 0 1 0 10 20
Real Part n (samples)
0 0

Phase (degrees)
Magnitude (dB)

−100
−20
−200
−40
−300

−60 −400
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

order: 2, cutoff frequency: 0.25 × fs /2, −3 dB bandwidth: 0.05 × fs /2

144 / 231
Peak filter design example
z Plane Impulse Response
1 1
Imaginary Part

0.5

Amplitude
0.5
0

−0.5
0
−1
−1 0 1 0 10 20
Real Part n (samples)
0 100

Phase (degrees)
Magnitude (dB)

50
−20
0
−40
−50

−60 −100
0 0.5 1 0 0.5 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

order: 2, cutoff frequency: 0.25 × fs /2, −3 dB bandwidth: 0.05 × fs /2

145 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Random variables, vectors, and processes
Let S be the set of all possible outcomes (results) of some experiment.
We call S the sample space of that experiment.
A random variable X is a function

X:S→E

that assigns to each outcome ζ ∈ S a value X(ζ) ∈ E, where usually

E ⊆ R or E ⊆ C.
~
A random vector X(ζ) = (x1 (ζ), x2 (ζ), . . . , xn (ζ))T is a vector of n
random variables, or equivalently a random variable that outputs vectors,
~
e.g. X(ζ) ∈ Rn .
A continuous-time random process X : S → E R is a function that maps
each experimental outcome ζ ∈ S onto a continuous-time function xζ (t),
and a discrete-time random process X : S → E Z maps each outcome ζ
onto a discrete sequence {xζ,n }n .
The ensemble of a random process is the set of all functions (or
sequences) from which it picks its output.
In the following, we will usually omit outcome parameter ζ from random variables, etc., for
notational convenience, and use boldface roman to distinguish random variables from samples.
146 / 231
Random sequences
A discrete-time random process or random sequence {xn } can also be
thought of as a discrete sequence of random variables

. . . , x−2 , x−1 , x0 , x1 , x2 , . . .

Each time we repeat an experiment, we observe one realization or sample

sequence
. . . , x−2 , x−1 , x0 , x1 , x2 , . . .
of that random process. (We cannot observe the outcome ζ directly.)
Each individual random variable xn in a random sequence is
characterized by its probability distribution function

Pxn (a) = Prob(xn ≤ a)

and the entire random process is characterized completely by all joint

probability distribution functions

Pxn1 ,...,xnk (a1 , . . . , ak ) = Prob(xn1 ≤ a1 ∧ . . . ∧ xnk ≤ ak )

for all possible sets {xn1 , . . . , xnk } and all k > 0.

147 / 231
Two random variables xn and xm are called independent if

Pxn ,xm (a, b) = Pxn (a) · Pxm (b)

The derivative pxn (a) = Px0 n (a) is called the probability density function.
This helps us to define quantities such as the
R
I expected value E(xn ) = apxn (a) da
I mean-square value (average power) E(|xn |2 ) = |a|2 pxn (a) da
R

I variance Var(xn ) = E[|xn − E(xn )|2 ] = E(|xn |2 ) − |E(xn )|2

I correlation Cor(xn , xm ) = E(xn · x∗m )
I covariance Cov(xn , xm ) = E[(xn − E(xn )) · (xm − E(xm ))∗ ] =
E(xn x∗m ) − E(xn )E(xm )∗

The expected value E(·) is a linear operator: E(ax) = aE(x) and

E(x + y) = E(x) + E(y).
Variance is not linear, but Var(ax) = a2 Var(x) and, if x and y are
independent, Var(x + y) = Var(x) + Var(y).

148 / 231
Stationary processes

A random sequence is called strict-sense stationary if

Pxn1 +l ,...,xnk +l (a1 , . . . , ak ) = Pxn1 ,...,xnk (a1 , . . . , ak )

for any shift l and any number k, that is if all joint probability
distributions are time invariant.
If the above condition holds at least for k = 1, then the mean

E(xn ) = mx ,

and variance
E(|xn − mx |2 ) = σx2
are constant over all n. (σx is also called standard deviation).

If the above condition holds in addition also for k = 2, we call the

random sequence wide-sense stationary (WSS).
If a sequence is strict-sense stationary, it is always also wide-sense stationary, but not vice versa.

149 / 231
A wide-sense stationary random process {xn } can not only be
characterized by its mean mx = E(xn ) and variance σx2 = E(|xn − mx |2 )
over all sample positions n.
It can, in addition, also be characterized by its autocorrelation sequence
φxx (k) = E(xn+k · x∗n )

The autocorrelation sequence of a zero-mean version of a sequence is

called the autocovariance sequence
γxx (k) = E[(xn+k − mx ) · (xn − mx )∗ ] = φxx (k) − |mx |2
where γxx (0) = σx2 .
A pair of stationary random processes {xn } and {yn } can, in addition,
be jointly wide-sense stationary and therefore be characterized by their
crosscorrelation sequence
φxy (k) = E(xn+k · yn∗ )
Their crosscovariance sequence is then
γxy (k) = E[(xn+k − mx ) · (yn − my )∗ ] = φxy (k) − mx m∗y
∗
The complex conjugates are only needed with complex-valued sequences.
150 / 231
Ergodic processes
If . . . , x−2 , x−1 , x0 , x1 , x2 , . . . is a WSS random sequence, then we can
estimate the mean value and auto-correlation sequence from these
random variables from any location n as

mx = E(xn )
φxx (k) = E(xn+k x∗n )

What if we have just one sample sequence . . . , x−2 , x−1 , x0 , x1 , x2 , . . .?

If we still can estimate mean and auto-correlation from that as
L N
1 X 1 X
mx = lim xn ≈ xn for large N
L→∞ 2L + 1 N n=1
n=−L
L N
1 X 1 X
φxx (k) = lim xn+k x∗n ≈ xn+k x∗n
L→∞ 2L + 1 N n=1
n=−L

then we call the process mean ergodic and correlation ergodic, resp.
Ergodicity means that single-sample-sequence time averages are identical to averages over the
entire ensemble for a random process, or, in other words, variation along the time axis looks similar
to variation across the ensemble.
151 / 231
Deterministic crosscorrelation sequence
For deterministic finite-energy sequences {xn } and {yn }, we can define
their crosscorrelation sequence as
∞
X ∞
X
cxy (k) = xi+k · yi∗ = ∗
xi · yi−k .
i=−∞ i=−∞

If {xn } is similar to {yn }, but lags l elements behind (xn ≈ yn−l ), then cxy (l) will be a peak in
the crosscorrelation sequence. It can therefore be used to locate shifted versions of a known
sequence in another one.
MATLAB’s xcorr function calculates the crosscorrelation sequence for two finite sequences
(vectors) of equal length (or zero-pads the shorter one). Option unbiased divides for each lag
∗
through the length of the overlap, to estimate E(xn+k yn ) from a pair of sample sequences {xn }
and {yn } from jointly correlation-ergodic WSS processes {xn } and {yn }.

This crosscorrelation sequence is essentially just convolution, with the

second input sequence mirrored:
∗
{cxy (n)} = {xn } ∗ {y−n }
It can therefore be calculated equally easily via the DTFT:
Cxy (e jω̇ ) = X(e jω̇ ) · Y ∗ (e jω̇ )
Swapping the input sequences mirrors the output sequence: cxy (k) = c∗
yx (−k).
152 / 231
Demonstration of covert spread-spectrum communication:
n = randn(1,10000);
pattern=2*round(rand(1,1000))-1;
p1 = [zeros(1,2000), pattern, zeros(1,7000)];
p2 = [zeros(1,4000), pattern, zeros(1,5000)];
r = n + p1/3 - p2/3;
figure(1)
plot([n;p1/3-3;p2/3-4;r-6]');
figure(2)
plot(conv(r,fliplr(pattern)));
% or: plot(xcorr(r,pattern));

153 / 231
Deterministic autocorrelation sequence
Equivalently, we define the deterministic autocorrelation sequence in the
time domain as
X∞
cxx (k) = xi+k x∗i
i=−∞

This is just the sequence convolved with a time-reversed version of itself:

{cxx (k)} = {xi } ∗ {x∗−i }

This corresponds in the frequency domain to

Cxx (e jω̇ ) = X(e jω̇ ) · X ∗ (e jω̇ ) = |X(e jω̇ )|2 .

In other words, the DTFT Cxx (e jω̇ ) of the autocorrelation sequence

{cxx (n)} of a sequence {xn } is identical to the squared amplitudes of
the DTFT, or power spectrum, of {xn }.
This suggests, that the DTFT of the autocorrelation sequence of a
random process might be a suitable way for defining the power spectrum
of that random process.
What can we say about the phase in the Fourier spectrum of a time-invariant random process?
154 / 231
Power spectrum of a random sequence

For a zero-mean wide-sense stationary random sequence {xn } with

absolutely summable autocorrelation sequence

φxx (k) = E(xn+k · x∗n )

we call the DTFT

∞
X
Φxx (e jω̇ ) = φxx (n) · e− jω̇n
n=−∞

of its autocorrelation sequence the power density spectrum (PDS) or

power spectrum of {xn }.
The power spectrum is real, even∗ , non-negative and periodic.
∗
for real-valued sequences

155 / 231
The autocorrelation of a sequence {xn } with power spectrum Φxx (e jω̇ ) is
Z π
1
φxx (k) = Φxx (e jω̇ )e jkω̇ dω̇
2π −π

Since the variance of {xn } is

π
1
Z
Var(xn ) = φxx (0) = Φxx (e jω̇ )dω̇
2π −π

we can interpret
f
2π fhs
1
Z
Φxx (e jω̇ )dω̇
π f
2π fsl

as the variance of the output of an ideal band-pass filter applied to {xn }

with cut-off frequencies 0 ≤ fl < fh .

156 / 231
Filtered random sequences
Let {xn } be a random sequence from a WSS random process. The output
∞
X ∞
X
yn = hk · xn−k = hn−k · xk
k=−∞ k=−∞

of an LTI applied to it will then be another random sequence, characterized by

∞
! ∞ ∞
X X X
my = E(yn ) = E hk · xn−k = hk · E(xn−k ) = mx hk
k=−∞ k=−∞ k=−∞

and
∞
X φxx (k) = E(xn+k · x∗n )
φyy (k) = φxx (k − i)chh (i), where P∞ ∗
i=−∞ chh (k) = i=−∞ hi+k hi .

In other words:
{φyy (n)} = {chh (n)} ∗ {φxx (n)}
{yn } = {hn } ∗ {xn } ⇒
Φyy (e jω̇ ) = |H(e jω̇ )|2 · Φxx (e jω̇ )
Similarly:
{φyx (n)} = {hn } ∗ {φxx (n)}
{yn } = {hn } ∗ {xn } ⇒
jω̇
Φyx (e ) = H(e jω̇ ) · Φxx (e jω̇ )
157 / 231
Summary:
∗{hn } ∗{h∗
−n }
{yn } = {hn } ∗ {xn } ⇒ {φxx (n)} −→ {φyx (n)} −→ {φyy (n)}

Proofs:
∞
!
X
φyx (l) = E(x∗n · yn+l ) = E x∗n · hk · xn+l−k =
k=−∞
∞ ∞
WSS
X X
= hk · E(x∗n · xn+l−k ) = hk · φxx (l − k)
k=−∞ k=−∞

∞ ∞
!
X X
φyy (l) = E(yn∗ · yn+l ) = E h∗k · x∗n−k · hm · xn+l−m =
k=−∞ m=−∞
∞ ∞
WSS
X X
= h∗k · hm · E(x∗n−k · xn+l−m ) =
k=−∞ m=−∞
∞ ∞
i:=m−k
X X
= h∗k · hm · φxx (l + k − m) =
k=−∞ m=−∞
∞
X ∞
X ∞
X ∞
X
= h∗k · hk+i · φxx (l − i) = φxx (l − i) h∗k · hk+i
k=−∞ i=−∞ i=−∞ k=−∞
| {z }
chh (i) 158 / 231
White noise
A random sequence {xn } is a white noise signal, if mx = 0 and
φxx (k) = σx2 δk .
The power spectrum of a white noise signal is flat:
Φxx (e jω̇ ) = σx2 .
A commonly used form of white noise is white Gaussian noise (WGN),
where each random variable xn is independent and identically distributed
(i.i.d.) according to the normal-distribution probability density function
(x−mx )2
1 − 2
pxn (a) = p e 2σx

2πσx2
Application example:
Where an LTI {yn } = {hn } ∗ {xn } can be observed to operate on white
noise {xn } with φxx (k) = σx2 δk , the crosscorrelation between input and
output will reveal the impulse response of the system:
φyx (k) = σx2 · hk
where φyx (k) = φ∗xy (−k) = E(yn+k · x∗n ).
159 / 231
 Demonstration of covert spread-spectrum radar:
x = randn(1,10000)
h = [0 0 0.4 0 0 0.3 0 0 0.2 0 0];
y = conv(x, h);
figure(1)
plot(1:length(x), x, 1:length(y), y-5);
figure(2)
c = conv(fliplr(x),y);
stem(c(length(c)/2-20:length(c)/2+20));

4000

2000

−2000
0 10 20 30 40 50
160 / 231
Spectral estimation: periodogram
Estimate amplitude spectrum of the noisy discrete sequence

xk = sin(2π jk × 8/64) + sin(2π jk × 14.32/64) + ni with φnn (i) = 4δi

n = 64; % block length
m = 1000; % blocks averaged
k = 1:(n*m);
x = 2*randn(1, n*m) + ...
sin(2*pi*k*8/n) + ...
sin(2*pi*k*14.32/n);

s1 = abs(fft(x(1:n))/n);
s1 s2 s2 = abs(fft(x(1:8*n))/(8*n));

s1 Absolute values of a single 64-element DFT of {xn }64n=1 (rect. window).

The flat spectrum of white noise is only an expected value. In a single
discrete Fourier transform of such a sequence, the significant variance of
the noise spectrum becomes visible. It almost drowns the two peaks from
the sine waves.
s2 Absolute values of a single 512-element DFT of {xn }512
n=1 (rect. window).
With an 8× larger window, the bandwidth of each frequency bin is now
reduced 8×, so the sine functions stand out better from the noise.
However, the variance in each frequency bin relative to the expected value
remains the same.
161 / 231
Spectral estimation: averaging
Estimate amplitude spectrum of the noisy discrete sequence

xk = sin(2π jk × 8/64) + sin(2π jk × 14.32/64) + ni with φnn (i) = 4δi

n =
64; % block length
m =
1000; % blocks averaged
k =
1:(n*m);
x =
2*randn(1, n*m) + ...
sin(2*pi*k*8/n) + ...
sin(2*pi*k*14.32/n);
xx = reshape(x, n, m)';

s3 = mean(abs(fft(xx, n, 2)/n),1);
s3 s4 s4 = abs(mean(fft(xx, n, 2)/n,1));

s3 {xn }64000
n=1 cut into 1000 consecutive 64-sample windows, showing
the average of the absolute values of the DFT of each window.
Non-coherent averaging: discard phase information first.
This better approximates the shape of the power spectrum: with a flat noise floor.

s4 Same 1000 windows, but this time the complex values of the DFTs
averaged before the absolute value was taken ⇒ coherent averaging.
Because DFT is linear, this is identical to first averaging all 1000 windows and then applying
a single DFT and taking its absolute value.
The windows start 64 samples apart. Only periodic waveforms with a period length that
divides 64 are not averaged away. This periodic averaging step suppresses both the noise
and the second sine wave.
162 / 231
Welch’s method for estimating PSD
“Periodogram”: Single-rectanglar-window DTFTPpower spectrum of a
N −1
random sequence {xn }: |X(ω̇)|2 with X(ω̇) = n=0 xn · e−2π jnω̇ .
Var[|X(ω̇)|2 ]
Problem: E[|X(ω̇)|2 ] does not drop with increasing window length N .
“Welch’s method” for estimating the PSD makes three improvements:
I Reduce leakage using a non-rectangular window sequence {wi }
(“modified periodogram”)
I To reduce the variance, average K periodograms of length N .
I Triangular, Hamming, Hanning, etc. windows can be used with 50%
overlap (L = N/2), such that all samples contribute with equal
weight.
0≤k<K
xk,n = xk·L+n · wn ,
0≤n<N
N
X −1
Xk (ω̇) = xk,n · e−2π jnω̇
n=0

K−1
1 X
P (ω̇) = |Xk (ω̇)|2
K
k=0
163 / 231
Periodic averaging
If a signal x(t) has a periodic component with period length tp , then we
can isolate this periodic component from discrete sequence xn = x(n/fs )
by periodic averaging
L N
1 X 1 X
x̄n = lim xn+pi ≈ xn+pi , n ∈ {0, . . . , p − 1}
L→∞ 2L + 1 N i=1
i=−L

but only if the period length in samples p = tp · fs is an integer.

Otherwise {xn } may need to be interpolated and resampled at an integer multiple of t−1
p first.

Periodic averaging of x(t) corresponds in the

Ptime domain to convolution
with a windowed Dirac comb a(t) = w(t) · i δ(t − tp i):
Z
x̄(t) = x(t − s) · a(s)ds
s

In the frequency domain, this means multiplication with an t−1

p spaced
Dirac comb that has been convolved with W (f ).

164 / 231
Parametric models of the power spectrum
If we understand the physical process that generates a random sequence,
we may be able to model and estimate its power spectrum more
accurately, with fewer parameters.
If {xn } can be modeled as white noise filtered by an LTI system H(e jω̇ ),
then
Φxx (e jω̇ ) = σw
2
|H(e jω̇ )|2 .
Often such an LTI can be modeled as an IIR filter with
b0 + b1 z −1 + b2 z −2 + · · · + bm z −m
H(e jω̇ ) = .
a0 + a1 z −1 + a2 z −2 + · · · + ak z −k

The auto-regressive moving-average model ARMA(k, m) is

m
X k
X
xn = bl · wn−l − al · xn−l
l=0 l=1

2
where {wn } is stationary white noise with variance σw .
Pk
There is also the simpler AR(k) model xn = wn − l=1 al · xn−l .
165 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
IQ sampling
12 Audiovisual data compression
IQ sampling / downconversion / complex baseband signal

Consider signal x(t) ∈ R in which only frequencies fl < |f | < fh are of

interest. This band has a centre frequency of fc = (fl + fh )/2 and a
bandwidth B = fh − fl . It can be sampled efficiently (at the lowest
possible sampling frequency) by downconversion:
I Shift its spectrum by −fc :

y(t) = x(t) · e−2π jfc t

I Low-pass filter it with a cut-off frequency of B/2:

Z ∞
z(t) = B y(τ ) · sinc((t − τ )B) · dτ •−◦ Z(f ) = Y (f ) · rect(f /B)
−∞

I Sample the result at sampling frequency fs ≥ B:

zn = z(n/fs )

166 / 231
 X(f ) δ(f + fc )

∗
−fc 0 fc f −fc 0 fc f

Y (f ) anti-aliasing filter Ẑ(f ) Z(f )

sample
−→

−2fc −fc −B 0 B fc f −2fc −fc 0 B fc f

2 2

Shifting the center frequency fc of the interval of interest to 0 Hz (DC)

makes the spectrum asymmetric. This leads to a complex-valued
time-domain representation
(∃f : Z(f ) 6= [Z(−f )]∗ =⇒ ∃t : z(t) ∈ C \ R).

167 / 231
IQ upconversion / interpolation
Given a discrete sequence of downconverted samples zn ∈ C recorded
with sampling frequency fs at centre frequency fc (as on slide 166), how
can we reconstruct a continuous waveform x̃(t) ∈ R that matches the
original signal x(t) within the frequency interval fl to fh ?
Reconstruction steps:
I Interpolation of complex baseband signal (remove aliases):
∞
X
z̃(t) = zn · sinc(t · fs − n)
n=−∞

I Upconvert by modulating a complex phasor at carrier frequency fc .

Then discard the imaginary part (to reconstruct the negative
frequency components of the original real-valued signal):
 
x̃(t) = 2< z̃(t) · e2π jfc t
 




= 2< < z̃(t) + j= z̃(t) · cos 2πfc t + j sin 2πfc t



= 2< z̃(t) · cos 2πfc t − 2= z̃(t) · sin 2πfc t
Recall that 2<(c) = c + c∗ for all c ∈ C. 168 / 231
Example: IQ downconversion of a sine wave
What happens if we downconvert the input signal
A 2π jf t+ jφ A −2π jf t− jφ
x(t) = A · cos(2πf t + φ) = ·e + ·e
2 2
using centre frequency fc and bandwidth B < 2fc with |f − fc | < B/2?

After frequency shift:

A 2π j(f −fc )t+ jφ A −2π j(f +fc )t− jφ
y(t) = x(t) · e−2π jfc t = ·e + ·e
2 2
After low-pass filter with cut-off frequency B/2 < fc < f + fc :

A 2π j(f −fc )t+ jφ

z(t) = ·e
2
After sampling:
A 2π j(f −fc )n/fs + jφ
zn = ·e
2

169 / 231
Software-defined radio (SDR) front end
IQ downconversion in SDR receiver:
⊗ sample Q

x(t) −90◦ y(t) z(t) zn

⊗ sample I

The real part <(z(t)) is also known as “in-phase” signal (I) and
cos(2πfc t) the imaginary part =(z(t)) as “quadrature” signal (Q).

IQ upconversion in SDR transmitter:

⊗ δ Q

x̃(t) +90◦ z̃(t) ẑ(t) zn

⊗ δ I

In SDR, x(t) is the antenna voltage and zn appears on the

cos(2πfc t) digital interface with the microprocessor.
170 / 231
SDR front-end hardware examples
Low-cost USB-dongle receivers: ≈£20 SDR front ends are also
Realtek RTL2832U/R820T (RTL-SDR) commonly used today in
USB2, fs < 2.5 MHz, fc = 24–1776 MHz, 8-bit IQ samples
http://osmocom.org/projects/sdr/wiki/rtl-sdr military radios, spectrum
surveillance, amateur-radio
stations, mobile-phone base
stations, MRI machines,
radars, etc.

Mid range transceivers: £250–£2k High-end measurement kit: £3k–£40k

HackRF One, Ettus USRP B200/N200, etc. National Instruments, Rohde&Schwarz, etc.
USB3 or 1-Gbit Ethernet, fs = 10–50 MHz, 10 Gbit/PCIe, FPGA, B, fs = 60–1000 MHz,
fc = 0–6 GHz, 16-bit IQ samples fc = 0–14 GHz, float32 IQ samples in volts

171 / 231
Visualization of IQ representation of sine waves
⊗ Q
Q
x(t) −90◦ y(t) z(t)

⊗ I

cos(2πfc t)
I
Recall these products of sine and cosine functions:
I cos(x) · cos(y) = 1 cos(x − y) + 1 cos(x + y)
2 2
I sin(x) · sin(y) = 1 1
2 cos(x − y) − 2 cos(x + y)
I sin(x) · cos(y) = 1 1
2 sin(x − y) + 2 sin(x + y)
Consider: (with x = 2πfc t)
I sin(x) = cos(x − 1 π)
2
I cos(x) · cos(x) = 1 1
2 + 2 cos 2x
I sin(x) · sin(x) = 1 1
2 − 2 cos 2x
I sin(x) · cos(x) = 0 + 1
2 sin 2x
I cos(x) · cos(x − ϕ) = 1 1
2 cos(ϕ) + 2 cos(2x − ϕ)
I sin(x) · cos(x − ϕ) = 1 1
2 sin(ϕ) + 2 sin(2x − ϕ)
172 / 231
IQ representation of amplitude-modulated signals
Assume voice signal s(t) contains only frequencies below B/2.
Antenna signal amplitude-modulated with carrier frequency fc :
x(t) = s(t) · A · cos(2πfc t + ϕ)
After IQ downconversion with centre frequency fc0 ≈ fc :
A 0
z(t) = · s(t) · e2π j(fc −fc )t+ jϕ
2 =[z(t)]
With perfect receiver tuning fc0 = fc :
A
z(t) = · s(t) · e jϕ
2 <[z(t)]
Reception techniques:
I Non-coherent demodulation (requires s(t) ≥ 0):
2
s(t) = A |z(t)|

I Coherent demodulation (requires knowing ϕ and fc0 = fc ):

2
s(t) = A <[z(t) · e− jϕ ]
173 / 231
IQ representation of frequency-modulated signals

In frequency modulation, the voice signal s(t) changes the

carrier frequency fc : fc (t) = fc + k · s(t)
Compared to a constant-frequency carrier signal cos(2πfc t + ϕ), to allow
R variable frequency, we
need to replace the phase-accumulating term 2πfc t with an integral 2π fc (t)dt.

Frequency-modulated antenna signal:

 Z t 
x(t) = A · cos 2π · [fc + k · s(τ )]dτ + ϕ
0
 Z t 
= A · cos 2πfc t + 2πk · s(τ )dτ + ϕ
0

After IQ downconversion from centre frequency fc :

A 2π jk R t s(τ )dτ + jϕ
z(t) = ·e 0
2

Therefore, s(t) is proportional to the rotational rate of z(t).

174 / 231
Frequency demodulating IQ samples
Rt
Determine s(t) from downconverted signal z(t) = A
2 · e2π jk 0
s(τ )dτ + jϕ
.

First idea: measure the angle ∠z(t), where the angle operator ∠ is
defined such that ∠ae jφ = φ (a, φ ∈ R, a > 0). Then take its derivative:
1 d
s(t) = ∠z(t)
2πk dt
Problem: angle ambiguity, ∠ works only for −π ≤ φ < π.
Ugly hack: MATLAB function unwrap removes 2π jumps from sample sequences

Better idea: first take the complex derivative

dz(t) A Rt
= · 2π jk · s(t) · e2π jk 0 s(τ )dτ + jϕ
dt 2
=[z(t)]
then divide by z(t): dz(t)
dt /z(t) = 2π jk · s(t)
Other practical approaches:
h i
I s(t) ∝ = dz(t) ∗
dt · z (t) /|z(t)|
2
<[z(t)]
z(t)
I s(t) ∝ ∠ z(t−∆t) /∆t
175 / 231
Digital modulation schemes
Pick zn ∈ C from a constellation of 2n symbols to encode n bits:

ASK BPSK QPSK

10 00

0 1 0 1

11 01

8PSK 101 16QAM FSK

1
111 100
00
0
110 000 01

11

010 001 10

011 00 01 11 10
176 / 231
Basic model of a modem

bits - symbols - P impulses - transmit filter - IQ

bj ∈ {0, 1} ai ∈ S ai δ(t − its ) ∗ht (t) upconversion

?
noise - LTI channel
+n(t) ∗hc (t)

?
bits  data  sampling  receive filter  IQ
slicer ∗hr (t) downconv.

IQ up/down conversion: only required for pass-band channels (e.g., radio)

177 / 231
Pulse Amplitude Modulation (PAM)
Baseband transmission (e.g., for wires), no IQ up/down conversion
I binary PAM: ai ∈ S = {−A, A} ⊂ R
1 bit/symbol ⇒ bit rate (bit/s) = symbol rate (baud)
I m-ary PAM: ai ∈ S = {A1 , . . . , Am } ⊂ R
k = log2 m bit/symbol ⇒ bit rate (bit/s) > symbol rate (baud)
I bit sequence {bj } → symbol sequence {ai },
ai = f (bki , . . . , bki+k−1 )
Pulse generator (symbol rate fs = t−1
s ):
X
x̂(t) = ai · δ(t − its )
i

Transmit filter: x = x̂ ∗ ht , X(f ) = X̂(f ) · Ht (f )

X
x(t) = ai · ht (t − its )
i

Channel: Z ∞
z(t) = hc (s)x(t − s)ds + n(t)
0
178 / 231
PAM reception
Receive filter applied to channel output z(t):
Z ∞
y(t) = hr (s)y(t − s)ds
0

Initial symbol pulses x̂(t) have now passed through three LTIs:

y = h ∗ x̂
h = ht ∗ hc ∗ hr
H(f ) = Ht (f ) · Hc (f ) · Hr (f )

Sample y(t) at times tn = nts + td with delay td where pulse magnitude

is largest:
X
yn = y(nts + td ) = ai h((n − i)ts + td ) + vn
i

where vn = v(nts + td ) is the sampled noise v = n ∗ hr .

Data slicer: compare yn against thresholds and convert detected nearest
symbol a0n ∈ S back into bits b0kn , . . . , b0kn+k−1 .
179 / 231
Synchronization
The receiver needs to know the times tn when to sample y(t).
I Local clock generators have temperature-dependent frequency drift.
I In some transmission systems, the transmitter provides the sample
clock on a separate wire (or wire pair).
For example: DVI and HDMI video cables contain four wire pairs: three transmitting
red/green/blue pixel bytes (using an 8b/10b line encoding), and one providing a pixel clock
signal, which the receiver multiplies 10× to get a bit clock.
I More commonly, the receiver has to extract the sample clock from
the received signal, for example by tracking the phase of transitions
(phase locked loop, PLL).
This works reliably only if there are regular transitions.
• Some systems use a line encoding (e.g., Manchester code, 8b/10b
encoding) to ensure regular transitions.
Some line encodings add a spectral line at the symbol rate, which the receiver can
extract with a band-pass filter, others first require a non-linear step, e.g. squaring.
• Others use a scrambler: the data bits bi are XORed with the output
of synchronized deterministic random-bit generators (e.g., a
maximum-length linear feedback shift register), in both the sender
and recipient, to make long runs of the same symbol unlikely.
180 / 231
Intersymbol interference
For notational convenience: set td = 0 and allow h(t) to be non-causal.
X
yn = an h(0) + ai h((n − i)ts ) + vn
i6=n

Ideally, we want (
1, i=0
h(its ) =
0, i 6= 0
otherwise yn will depend on other (mainly previous) symbols, not just on
an ⇒ intersymbol interference. (See also: interpolation function)
Nyquist ISI criterion
X
yn = an + vn ⇔ h(t) · δ(t − its ) = δ(t)
i
X
⇔ H(f ) ∗ fs δ(f − ifs ) = 1
i
X
⇔ H(f − ifs ) = ts
i

181 / 231
Some possible pulse-shape choices
I h(t) = rect(t/ts ) •−◦ H(f ) = ts sinc(f /fs )
Rectangular pulses may be practical on fibre optics and short cables, where there are no
bandwidth restriction. Not suitable for radio: bandwidth high compared to symbol rate.
I h(t) = sinc(t/ts ) •−◦ H(f ) = ts rect(f /fs )
Most bandwidth efficient pulse shape, but very long symbol waveform, very sensitive to
clock synchronization errors.
I Raised-cosine filter: rectangle with half-period cosine transitions

ts ,
 |f | ≤ ts /2 − β
2 π
H(f ) = ts cos 4β (|f | − ts /2 + β), ts /2 − β < |f | ≤ ts /2 + β

0, |f | > ts /2 + β


cos 2πβt
h(t) = sinc(t/ts )
1 − (4βt)2

Transition width (roll-off) β with 0 ≤ β ≤ ts /2; for β = 0 this is H(f ) = ts rect(f /fs ).
I Gaussian filter: both h(t) and H(f ) are Gaussians (self-Fourier)
Fastest transition without overshot in either time or frequency domain, but does not satisfy
Nyquist ISI criterion.

182 / 231
Optimal transmit and receive filtering
Nyquist ISI criterion dictates H(f ) = Ht (f ) · Hc (f ) · Hr (f ).
Bandwidth limits guide choice of Ht (f ), and channel dictates Hc (f ) and
N (f ).
How should we then choose Ht (f ) Hr (f )?
Select a received pulse spectrum Pr (f ), e.g. raised cosine. Then for some
arbitrary gain factor k > 0:
H(f ) = Ht (f ) · Hc (f ) · Hr (f ) = k · Pr (f )

Optimal filters
Minimize noise variance Var(vn ) = N (f )|Hr (f )|2 df at slicer relative to
R

symbol distance.
  12
 Pr (f ) 
|Hr (f )| =  p
 

 N (f )Hc (f ) 
 1
 P (f )pN (f )  2
 r
|Ht (f )| = k 

Hc (f )

 

If N (f ) and Hc (f ) are flat: |Hr (f )| = |Ht (f )|/k 0 , e.g. root raised cosine. 183 / 231
Outline
1 Sequences and systems
2 Convolution
3 Fourier transform
4 Sampling
5 Discrete Fourier transform
6 Deconvolution
7 Spectral estimation
8 Digital filters
9 IIR filters
10 Random signals
11 Digital communication
12 Audiovisual data compression
Audiovisual data compression

Structure of modern audiovisual communication systems:

sensor + - perceptual entropy channel

signal - - -
sampling coding coding coding

?
noise - channel

?
human perceptual  entropy channel
senses
 display  
decoding decoding decoding

184 / 231
Audio-visual lossy coding today typically consists of these steps:
I A transducer converts the original stimulus into a voltage.
I This analog signal is then sampled and quantized.
The digitization parameters (sampling frequency, quantization levels) are preferably chosen
generously beyond the ability of human senses or output devices.
I The digitized sensor-domain signal is then transformed into a
perceptual domain.
This step often mimics some of the first neural processing steps in humans.
I This signal is quantized again, based on a perceptual model of what level
of quantization-noise humans can still sense.
I The resulting quantized levels may still be highly statistically dependent.
A prediction or decorrelation transform exploits this and produces a less
dependent symbol sequence of lower entropy.
I An entropy coder turns that into an apparently-random bit string, whose
length approximates the remaining entropy.
The first neural processing steps in humans are in effect often a kind of decorrelation transform;
our eyes and ears were optimized like any other AV communications system. This allows us to use
the same transform for decorrelating and transforming into a perceptually relevant domain.

185 / 231
Outline of the remaining lectures

I Quick review of entropy coding

I Transform coding: techniques for converting sequences of
highly-dependent symbols into less-dependent lower-entropy
sequences.
• run-length coding
• decorrelation, Karhunen-Loève transform (PCA)
• Discrete cosine transform
I Introduction to some characteristics and limits of human senses
• perceptual scales and sensitivity limits
• colour vision
I Quantization techniques to remove information that is irrelevant to
human senses

186 / 231
Entropy coding review – Huffman

X 1
1.00 Entropy: H = p(α) · log2
p(α)
0 1 α∈A

= 2.3016 bit
0.40 0.60

0 1 0 1
v w
0.25
0.20 0.20 u
0.35 0 1

Mean codeword length: 2.35 bit x

0.10
0.15

Huffman’s algorithm constructs an optimal code-word tree for a set of 0 1

symbols with known probability distribution. It iteratively picks the two
elements of the set with the smallest probability and combines them into y z
a tree by adding a common root. The resulting tree goes back into the 0.05 0.05
set, labeled with the sum of the probabilities of the elements it combines.
The algorithm terminates when less than two elements are left.

187 / 231
Entropy coding review – arithmetic coding
Partition [0,1] according 0.0 0.35 0.55 0.75 0.9 0.95 1.0
to symbol probabilities: u v w x y z
Encode text wuvw . . . as numeric value (0.58. . . ) in nested intervals:
1.0 0.75 0.62 0.5885 0.5850
z z z z z
y y y y y

x x x x x

w w w w w

v v v v v

u u u u u

0.55
0.0 0.55 0.5745 0.5822

188 / 231
Arithmetic coding

Several advantages:
I Length of output bitstring can approximate the theoretical
information content of the input to within 1 bit.
I Performs well with probabilities > 0.5, where the information per
symbol is less than one bit.
I Interval arithmetic makes it easy to change symbol probabilities (no
need to modify code-word tree) ⇒ convenient for adaptive coding
Can be implemented efficiently with fixed-length arithmetic by rounding
probabilities and shifting out leading digits as soon as leading zeros
appear in interval size. Usually combined with adaptive probability
estimation.
Huffman coding remains popular because of its simplicity and lack of patent-licence issues.

189 / 231
Coding of sources with memory and correlated symbols
Run-length coding:

↓
5 7 12 3 3

Predictive coding:
encoder decoder
f(t) g(t) g(t) f(t)
− +

predictor predictor
P(f(t−1), f(t−2), ...) P(f(t−1), f(t−2), ...)

Delta coding (DPCM): P (x) = x

Xn
Linear predictive coding: P (x1 , . . . , xn ) = ai xi
i=1

190 / 231
Old (Group 3 MH) fax code
pixels white code black code
I Run-length encoding plus modified Huffman 0 00110101 0000110111
code 1 000111 010
I Fixed code table (from eight sample pages) 2 0111 11
3 1000 10
I separate codes for runs of white and black
4 1011 011
pixels 5 1100 0011
I termination code in the range 0–63 switches 6 1110 0010
between black and white code 7 1111 00011
I makeup code can extend length of a run by a 8 10011 000101
multiple of 64 9 10100 000100
10 00111 0000100
I termination run length 0 needed where run 11 01000 0000101
length is a multiple of 64 12 001000 0000111
I single white column added on left side before 13 000011 00000100
transmission 14 110100 00000111
I makeup codes above 1728 equal for black and 15 110101 000011000
16 101010 0000010111
white
... ... ...
I 12-bit end-of-line marker: 000000000001 (can 63 00110100 000001100111
be prefixed by up to seven zero-bits to reach 64 11011 0000001111
next byte boundary) 128 10010 000011001000
Example: line with 2 w, 4 b, 200 w, 3 b, EOL → 192 010111 000011001001
1000|011|010111|10011|10|000000000001 ... ... ...
1728 010011011 0000001100101

191 / 231
Modern (JBIG) fax code

Performs context-sensitive arithmetic coding of binary pixels. Both encoder and

decoder maintain statistics on how the black/white probability of each pixel
depends on these 10 previously transmitted neighbours:

Based on the counted numbers nblack and nwhite of how often each pixel value
has been encountered so far in each of the 1024 contexts, the probability for
the next pixel being black is estimated as
nblack + 1
pblack =
nwhite + nblack + 2

The encoder updates its estimate only after the newly counted pixel has been
encoded, such that the decoder knows the exact same statistics.
Joint Bi-level Expert Group: International Standard ISO 11544, 1993.
Example implementation: http://www.cl.cam.ac.uk/~mgk25/jbigkit/

192 / 231
Statistical dependence
Random variables X, Y are dependent iff ∃x, y:

P (X = x ∧ Y = y) 6= P (X = x) · P (Y = y).

If X, Y are dependent, then

⇒ ∃x, y : P (X = x | Y = y) 6= P (X = x) ∨
P (Y = y | X = x) 6= P (Y = y)
⇒ H(X|Y) < H(X) ∨ H(Y|X) < H(Y)

Application
Where x is the value of the next symbol to be transmitted and y is the
vector of all symbols transmitted so far, accurate knowledge of the
conditional probability P (X = x | Y = y) will allow a transmitter to
remove all redundancy.
An application example of this approach is JBIG, but there y is limited to
10 past single-bit pixels and P (X = x | Y = y) is only an estimate.
193 / 231
Practical limits of measuring conditional probabilities

The practical estimation of conditional probabilities, in their most general

form, based on statistical measurements of example signals, quickly
reaches practical limits. JBIG needs an array of only 211 = 2048 counting
registers to maintain estimator statistics for its 10-bit context.
If we wanted to encode each 24-bit pixel of a colour image based on its
statistical dependence of the full colour information from just ten
previous neighbour pixels, the required number of

(224 )11 ≈ 3 × 1080

registers for storing each probability will exceed the estimated number of
particles in this universe. (Neither will we encounter enough pixels to
record statistically significant occurrences in all (224 )10 contexts.)
This example is far from excessive. It is easy to show that in colour
images, pixel values show statistical significant dependence across colour
channels, and across locations more than eight pixels apart.
A simpler approximation of dependence is needed: correlation.

194 / 231
Correlation

Two random variables X ∈ R and Y ∈ R are correlated iff

E{[X − E(X)] · [Y − E(Y)]} =

6 0

where E(· · · ) denotes the expected value of a random-variable term.

Dependent but not correlated:

Correlation implies dependence, but 1

dependence does not always lead to
correlation (see example to the right).
However, most dependency in audio- 0
visual data is a consequence of corre-
lation, which is algorithmically much
easier to exploit.
−1
−1 0 1
Positive correlation: higher X ⇔ higher Y, lower X ⇔ lower Y
Negative correlation: lower X ⇔ higher Y, higher X ⇔ lower Y

195 / 231
Correlation of neighbour pixels
Values of neighbour pixels at distance 1 Values of neighbour pixels at distance 2
256 256

192 192

128 128

64 64

0 0
0 64 128 192 256 0 64 128 192 256
Values of neighbour pixels at distance 4 Values of neighbour pixels at distance 8
256 256

192 192

128 128

64 64

0 0
0 64 128 192 256 0 64 128 192 256

196 / 231
Covariance and correlation

We define the covariance of two random variables X and Y as

Cov(X, Y) = E{[X − E(X)] · [Y − E(Y)]} = E(X · Y) − E(X) · E(Y)

and the variance as Var(X) = Cov(X, X) = E{[X − E(X)]2 }.

The Pearson correlation coefficient

Cov(X, Y)
ρX,Y = p
Var(X) · Var(Y)

is a normalized form of the covariance. It is limited to the range [−1, 1].

If the correlation coefficient has one of the values ρX,Y = ±1, this
implies that X and Y are exactly linearly dependent, i.e. Y = aX + b,
with a = Cov(X, Y)/Var(X) and b = E(Y) − E(X).

197 / 231
Covariance Matrix

~ = (X1 , X2 , . . . , Xn ) ∈ Rn we define the

For a random vector X
covariance matrix
 
Cov(X)~ = E (X~ − E(X))
~ · (X ~ − E(X)) ~ T = (Cov(Xi , Xj )) =
i,j
 
Cov(X1 , X1 ) Cov(X1 , X2 ) Cov(X1 , X3 ) · · · Cov(X1 , Xn )
 Cov(X2 , X1 ) Cov(X2 , X2 ) Cov(X2 , X3 ) · · · Cov(X2 , Xn ) 
 
 Cov(X3 , X1 ) Cov(X3 , X2 ) Cov(X3 , X3 ) · · · Cov(X3 , Xn ) 
.. .. .. ..
 
 .. 
 . . . . . 
Cov(Xn , X1 ) Cov(Xn , X2 ) Cov(Xn , X3 ) · · · Cov(Xn , Xn )

~ are uncorrelated if and only if

The elements of a random vector X
~ is a diagonal matrix.
Cov(X)
Cov(X, Y) = Cov(Y, X), so all covariance matrices are symmetric:
~ = CovT (X).
Cov(X) ~

198 / 231
Decorrelation by coordinate transform
Neighbour−pixel value pairs Decorrelated neighbour−pixel value pairs
256 320

256
192
192

128 128

64
64
0

0 −64
0 64 128 192 256 −64 0 64 128 192 256 320
Probability distribution and entropy

correlated value pair (H = 13.90 bit)

Idea: Take the values of a group of cor-
decorrelated value 1 (H = 7.12 bit) related symbols (e.g., neighbour pixels) as
decorrelated value 2 (H = 4.75 bit)
a random vector. Find a coordinate trans-
form (multiplication with an orthonormal
matrix) that leads to a new random vector
whose covariance matrix is diagonal. The
vector components in this transformed co-
ordinate system will no longer be corre-
lated. This will hopefully reduce the en-
tropy of some of these components.
−64 0 64 128 192 256 320 199 / 231
Theorem: Let X ~ ∈ Rn and Y
~ ∈ Rn be random vectors that are linearly
~ ~
dependent with Y = AX + b, where A ∈ Rn×n and b ∈ Rn are
constants. Then
~
E(Y) ~ +b
= A · E(X)
~
Cov(Y) ~ · AT
= A · Cov(X)

Proof: The first equation follows from the linearity of the expected-value
~ · B) = A · E(X)
operator E(·), as does E(A · X ~ · B for matrices A, B.
With that, we can transform
 
Cov(Y)~ = E (Y ~ − E(Y))
~ · (Y~ − E(Y))
~ T
 
= E (AX ~ − AE(X))
~ · (AX ~ − AE(X))~ T
 
= E A(X ~ − E(X))
~ · (X ~ − E(X))
~ T AT
 
= A · E (X ~ − E(X))
~ · (X ~ − E(X))
~ T · AT

= ~ · AT
A · Cov(X)

200 / 231
Quick review: eigenvectors and eigenvalues

We are given a square matrix A ∈ Rn×n . The vector x ∈ Rn is an

eigenvector of A if there exists a scalar value λ ∈ R such that

Ax = λx.

The corresponding λ is the eigenvalue of A associated with x.

The length of an eigenvector is irrelevant, as any multiple of it is also an
eigenvector. Eigenvectors are in practice normalized to length 1.

Spectral decomposition
Any real, symmetric matrix A = AT ∈ Rn×n can be diagonalized into the
form
A = U ΛU T ,
where Λ = diag(λ1 , λ2 , . . . , λn ) is the diagonal matrix of the ordered
eigenvalues of A (with λ1 ≥ λ2 ≥ · · · ≥ λn ), and the columns of U are
the n corresponding orthonormal eigenvectors of A.

201 / 231
Karhunen-Loève transform (KLT)
We are given a random vector variable X ~ ∈ Rn . The correlation of the
~
elements of X is described by the covariance matrix Cov(X). ~
How can we find a transform matrix A that decorrelates X, ~ i.e. that
~ ~ T
turns Cov(AX) = A · Cov(X) · A into a diagonal matrix? A would
provide us the transformed representation Y ~ = AX
~ of our random
vector, in which all elements are mutually uncorrelated.
Note that Cov(X) ~ is symmetric. It therefore has n real eigenvalues
λ1 ≥ λ2 ≥ · · · ≥ λn and a set of associated mutually orthogonal
eigenvectors b1 , b2 , . . . , bn of length 1 with
~ i = λi bi .
Cov(X)b

We convert this set of equations into matrix notation using the matrix
B = (b1 , b2 , . . . , bn ) that has these eigenvectors as columns and the
diagonal matrix D = diag(λ1 , λ2 , . . . , λn ) that consists of the
corresponding eigenvalues:
~
Cov(X)B = BD

202 / 231
B is orthonormal, that is BB T = I.
Multiplying the above from the right with B T leads to the spectral
decomposition
Cov(X)~ = BDB T

of the covariance matrix. Similarly multiplying instead from the left with
B T leads to
~
B T Cov(X)B =D
and therefore shows with
~ =D
Cov(B T X)

that the eigenvector matrix B T is the wanted transform.

The Karhunen-Loève transform (also known as Hotelling transform or
Principal Component Analysis) is the multiplication of a correlated
random vector X~ with the orthonormal eigenvector matrix B T from the
spectral decomposition Cov(X)~ = BDB T of its covariance matrix. This
~ whose covariance matrix is
leads to a decorrelated random vector B T X
diagonal.

203 / 231
Karhunen-Loève transform example I

colour image red channel green channel blue channel

2
The colour image (left) has m = r pixels, each We can now define the mean colour vector
of which is an n = 3-dimensional RGB vector  
1 X
m 0.4839
Ix,y = (rx,y , gx,y , bx,y )
T S̄c = Sc,i , S̄ =  0.4456 
m i=1 0.3411

The three rightmost images show each of these and the covariance matrix
colour planes separately as a black/white
image. 1 X m
Cc,d = (Sc,i − S̄c )(Sd,i − S̄d )
We want to apply the KLT on a set of such Rn m − 1 i=1
colour vectors. Therefore, we reformat the  
image I into an n × m matrix of the form 0.0328 0.0256 0.0160
C =  0.0256 0.0216 0.0140 
  0.0160 0.0140 0.0109
r1,1 r1,2 r1,3 · · · rr,r
S =  g1,1 g1,2 g1,3 · · · gr,r 
b1,1 b1,2 b1,3 · · · br,r [“m − 1” because S̄c only estimates the mean]
204 / 231
Karhunen-Loève transform example I
The resulting covariance matrix C has three eigenvalues 0.0622, 0.0025, and 0.0006:
    
0.0328 0.0256 0.0160 0.7167 0.7167
 0.0256 0.0216 0.0140   0.5833  = 0.0622  0.5833 
0.0160 0.0140 0.0109 0.3822 0.3822
    
0.0328 0.0256 0.0160 −0.5509 −0.5509
 0.0256 0.0216 0.0140   0.1373  = 0.0025  0.1373 
0.0160 0.0140 0.0109 0.8232 0.8232
    
0.0328 0.0256 0.0160 −0.4277 −0.4277
 0.0256 0.0216 0.0140   0.8005  = 0.0006  0.8005 
0.0160 0.0140 0.0109 −0.4198 −0.4198

It can thus be diagonalized as

 
0.0328 0.0256 0.0160
 0.0256 0.0216 0.0140  = C = U · D · U T =
0.0160 0.0140 0.0109
   
0.7167 −0.5509 −0.4277 0.0622 0 0 0.7167 0.5833 0.3822
 0.5833 0.1373 0.8005   0 0.0025 0   −0.5509 0.1373 0.8232 
0.3822 0.8232 −0.4198 0 0 0.0006 −0.4277 0.8005 −0.4198

(e.g. using MATLAB’s singular-value decomposition function svd).

205 / 231
Karhunen-Loève transform example I

Before KLT: We finally apply the orthogonal 3 × 3 transform

matrix U , which we just used to diagonalize the
covariance matrix, to the entire image:
  
S̄1 S̄1 · · · S̄1
T
T = U · S −  S̄2 S̄2 · · · S̄2 
S̄3 S̄3 · · · S̄3
red green blue  
S̄1 S̄1 · · · S̄1
After KLT:
+  S̄2 S̄2 · · · S̄2 
S̄3 S̄3 · · · S̄3

The resulting transformed image

 
u1,1 u1,2 u1,3 · · · ur,r
u v w T =  v1,1 v1,2 v1,3 · · · vr,r 
Projections on eigenvector subspaces: w1,1 w1,2 w1,3 · · · wr,r

consists of three new “colour” planes whose

pixel values have no longer any correlation to
the pixels at the same coordinates in another
plane. [The bear disappeared from the last of
these (w), which represents mostly some of the
v=w=0 w=0 original green grass in the background.]

206 / 231
Spatial correlation

The previous example used the Karhunen-Loève transform in order to

eliminate correlation between colour planes. While this is of some
relevance for image compression, far more correlation can be found
between neighbour pixels within each colour plane.
In order to exploit such correlation using the KLT, the sample set has to
be extended from individual pixels to entire images. The underlying
calculation is the same as in the preceeding example, but this time the
columns of S are entire (monochrome) images. The rows are the
different images found in the set of test images that we use to examine
typical correlations between neighbour pixels.
In other words, we use the same formulas as in the previous example, but this time n is the
number of pixels per image and m is the number of sample images. The Karhunen-Loève
transform is here no longer a rotation in a 3-dimensional colour space, but it operates now in a
much larger vector space that has as many dimensions as an image has pixels.
To keep things simple, we look in the next experiment only at m = 9000 1-dimensional “images”
with n = 32 pixels each. As a further simplification, we use not real images, but random noise
that was filtered such that its amplitude spectrum is proportional to 1/f , where f is the frequency.
The result would be similar in a sufficiently large collection of real test images.

207 / 231
Karhunen-Loève transform example II
Matrix columns of S filled with samples of 1/f filtered noise

...
Covariance matrix C Matrix U with eigenvector columns

208 / 231
 Matrix U 0 with normalised KLT Matrix with Discrete Cosine
eigenvector columns Transform base vector columns

Breakthrough: Ahmed/Natarajan/Rao discovered the DCT as an

excellent approximation of the KLT for typical photographic images, but
far more efficient to calculate.
Ahmed, Natarajan, Rao: Discrete Cosine Transform. IEEE Transactions on Computers, Vol. 23,
January 1974, pp. 90–93.

209 / 231
Discrete cosine transform (DCT)

The forward and inverse discrete cosine transform

N −1
C(u) X (2x + 1)uπ
S(u) = s(x) cos
2N
p
N/2 x=0
N −1
X C(u) (2x + 1)uπ
s(x) = S(u) cos
2N
p
u=0
N/2

with
√1 u=0

C(u) = 2
1 u>0
is an orthonormal transform:
N −1
(2x + 1)uπ C(u0 ) (2x + 1)u0 π u = u0

X C(u) 1
cos ·p cos =
2N 2N 0 u 6= u0
p
x=0
N/2 N/2

210 / 231
DCT base vectors for N = 8:

4
u

0 1 2 3 4 5 6 7
x
211 / 231
Discrete cosine transform – 2D

The 2-dimensional variant of the DCT applies the 1-D transform on both
rows and columns of an image:

C(u) C(v)
S(u, v) = p p ·
N/2 N/2
N −1 N −1
X X (2x + 1)uπ (2y + 1)vπ
s(x, y) cos cos
x=0 y=0
2N 2N

s(x, y) =
N −1 N −1
X X C(u) C(v) (2x + 1)uπ (2y + 1)vπ
· S(u, v) cos cos
2N 2N
p p
u=0 v=0
N/2 N/2

A range of fast algorithms have been found for calculating 1-D and 2-D
DCTs (e.g., Ligtenberg/Vetterli).

212 / 231
Whole-image DCT

2D Discrete Cosine Transform (log10) Original image

4
50 50
3
100 100

150 2 150

200 1 200

250 0 250

300 300
−1
350 350
−2
400 400
−3
450 450
−4
500 500
100 200 300 400 500 100 200 300 400 500

213 / 231
Whole-image DCT, 80% coefficient cutoff

80% truncated 2D DCT (log10) 80% truncated DCT: reconstructed image

4
50 50
3
100 100

150 2 150

200 1 200

250 0 250

300 300
−1
350 350
−2
400 400
−3
450 450
−4
500 500
100 200 300 400 500 100 200 300 400 500

214 / 231
Whole-image DCT, 90% coefficient cutoff

90% truncated 2D DCT (log10) 90% truncated DCT: reconstructed image

4
50 50
3
100 100

150 2 150

200 1 200

250 0 250

300 300
−1
350 350
−2
400 400
−3
450 450
−4
500 500
100 200 300 400 500 100 200 300 400 500

215 / 231
Whole-image DCT, 95% coefficient cutoff

95% truncated 2D DCT (log10) 95% truncated DCT: reconstructed image

4
50 50
3
100 100

150 2 150

200 1 200

250 0 250

300 300
−1
350 350
−2
400 400
−3
450 450
−4
500 500
100 200 300 400 500 100 200 300 400 500

216 / 231
Whole-image DCT, 99% coefficient cutoff

99% truncated 2D DCT (log10) 99% truncated DCT: reconstructed image

4
50 50
3
100 100

150 2 150

200 1 200

250 0 250

300 300
−1
350 350
−2
400 400
−3
450 450
−4
500 500
100 200 300 400 500 100 200 300 400 500

217 / 231
Base vectors of 8×8 DCT
v
0 1 2 3 4 5 6 7
0

3
u

7
218 / 231
RGB video colour coordinates

Hardware interface (VGA): red, green, blue signals with 0–0.7 V

Electron-beam current and photon count of cathode-ray displays are
roughly proportional to (v − v0 )γ , where v is the video-interface or
control-grid voltage and γ is a device parameter that is typically in the
range 1.5–3.0. In broadcast TV, this CRT non-linearity is compensated
electronically in TV cameras. A welcome side effect is that it
approximates Stevens’ scale and therefore helps to reduce perceived noise.
Software interfaces map RGB voltage linearly to {0, 1, . . . , 255} or 0–1.
How numeric RGB values map to colour and luminosity depends at
present still highly on the hardware and sometimes even on the operating
system or device driver.
The new specification “sRGB” aims to standardize the meaning of an
RGB value with the parameter γ = 2.2 and with standard colour
coordinates of the three primary colours.
http://www.w3.org/Graphics/Color/sRGB, IEC 61966

219 / 231
YUV video colour coordinates

The human eye processes colour and luminosity at different resolutions.

To exploit this phenomenon, many image transmission systems use a
colour space with a luminance coordinate

Y = 0.3R + 0.6G + 0.1B

and colour (“chrominance”) components

V = R−Y = 0.7R − 0.6G − 0.1B

U = B − Y = −0.3R − 0.6G + 0.9B
220 / 231
YUV transform example

original Y channel U and V channels

The centre image shows only the luminance channel as a black/white
image. In the right image, the luminance channel (Y) was replaced with
a constant, such that only the chrominance information remains.
This example and the next make only sense when viewed in colour. On a black/white printout of
this slide, only the Y channel information will be present.

221 / 231
Y versus UV sensitivity example

original blurred U and V blurred Y channel

In the centre image, the chrominance channels have been severely
low-pass filtered (Gaussian impulse response ). But the human eye
perceives this distortion as far less severe than if the exact same filtering
is applied to the luminance channel (right image).
222 / 231
YCrCb video colour coordinates
Since −0.7 ≤ V ≤ 0.7 and −0.9 ≤ U ≤ 0.9, a more convenient
normalized encoding of chrominance is:
Y=0.1 Y=0.3 Y=0.5
1 1 1

0.8 0.8 0.8

0.6 0.6 0.6

Cr

Cr

Cr
0.4 0.4 0.4

Cb = + 0.5 0.2 0.2 0.2

2.0 0
0 0.5 1
0
0 0.5 1
0
0 0.5 1
Cb Cb Cb
Y=0.7 Y=0.9 Y=0.99
1 1 1
V
Cr = + 0.5 0.8 0.8 0.8

1.6 0.6 0.6 0.6

Cr

Cr

Cr
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
0 0.5 1 0 0.5 1 0 0.5 1
Cb Cb Cb

Modern image compression techniques operate on Y , Cr, Cb channels

separately, using half the resolution of Y for storing Cr, Cb.
Some digital-television engineering terminology:
If each pixel is represented by its own Y , Cr and Cb byte, this is called a “4:4:4” format. In the
compacter “4:2:2” format, a Cr and Cb value is transmitted only for every second pixel, reducing
the horizontal chrominance resolution by a factor two. The “4:2:0” format transmits in alternating
lines either Cr or Cb for every second pixel, thus halving the chrominance resolution both
horizontally and vertically. The “4:1:1” format reduces the chrominance resolution horizontally by
a quarter and “4:1:0” does so in both directions. [ITU-R BT.601]
223 / 231
Quantization

Uniform/linear quantization: Non-uniform quantization:

6 6
5 5
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
-5 -5
-6 -6
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6

Quantization is the mapping from a continuous or large set of values

(e.g., analog voltage, floating-point number) to a smaller set of (typically
28 , 212 or 216 ) values.
This introduces two types of error:
I the amplitude of quantization noise reaches up to half the maximum
difference between neighbouring quantization levels
I clipping occurs where the input amplitude exceeds the value of the
highest (or lowest) quantization level

224 / 231
Example of a linear quantizer (resolution R, peak value V ):
   
x 1
y = max −V, min V, R +
R 2

Adding a noise signal that is uniformly distributed on [0, 1] instead of adding 12 helps to spread the
frequency spectrum of the quantization noise more evenly. This is known as dithering.

Variant with even number of output values (no zero):

    
x 1
y = max −V, min V, R +
R 2

Improving the resolution by a factor of two (i.e., adding 1 bit) reduces

the quantization noise by 6 dB.
Linearly quantized signals are easiest to process, but analog input levels
need to be adjusted carefully to achieve a good tradeoff between the
signal-to-quantization-noise ratio and the risk of clipping. Non-uniform
quantization can reduce quantization noise where input values are not
uniformly distributed and can approximate human perception limits.

225 / 231
Logarithmic quantization

Rounding the logarithm of the signal amplitude makes the quantization

error scale-invariant and is used where the signal level is not very
predictable. Two alternative schemes are widely used to make the
logarithm function odd and linearize it across zero before quantization:
µ-law:
V log(1 + µ|x|/V )
y= sgn(x) for −V ≤ x ≤ V
log(1 + µ)
A-law:

A|x| V

1+log A sgn(x) for 0 ≤ |x| ≤ A
y= A|x|
V (1+log V ) V

1+log A sgn(x) for A ≤ |x| ≤ V

European digital telephone networks use A-law quantization (A = 87.6), North American ones use
µ-law (µ=255), both with 8-bit resolution and 8 kHz sampling frequency (64 kbit/s). [ITU-T
G.711]

226 / 231
 V -law (US)
A-law (Europe)
signal voltage

-V

-128 -96 -64 -32 0 32 64 96 128

byte value

227 / 231
Joint Photographic Experts Group – JPEG
Working group “ISO/TC97/SC2/WG8 (Coded representation of picture and audio information)”
was set up in 1982 by the International Organization for Standardization.

Goals:
I continuous tone gray-scale and colour images
I recognizable images at 0.083 bit/pixel
I useful images at 0.25 bit/pixel
I excellent image quality at 0.75 bit/pixel
I indistinguishable images at 2.25 bit/pixel
I feasibility of 64 kbit/s (ISDN fax) compression with late 1980s
hardware (16 MHz Intel 80386).
I workload equal for compression and decompression
The JPEG standard (ISO 10918) was finally published in 1994.
William B. Pennebaker, Joan L. Mitchell: JPEG still image compression standard. Van Nostrad
Reinhold, New York, ISBN 0442012721, 1993.
Gregory K. Wallace: The JPEG Still Picture Compression Standard. Communications of the ACM
34(4)30–44, April 1991, http://doi.acm.org/10.1145/103085.103089

228 / 231
Summary of the baseline JPEG algorithm

The most widely used lossy method from the JPEG standard:
I Colour component transform: 8-bit RGB → 8-bit YCrCb
I Reduce resolution of Cr and Cb by a factor 2
I For the rest of the algorithm, process Y , Cr and Cb components
independently (like separate gray-scale images)
The above steps are obviously skipped where the input is a gray-scale image.
I Split each image component into 8 × 8 pixel blocks
Partial blocks at the right/bottom margin may have to be padded by repeating the last
column/row until a multiple of eight is reached. The decoder will remove these padding
pixels.
I Apply the 8 × 8 forward DCT on each block
On unsigned 8-bit input, the resulting DCT coefficients will be signed 11-bit integers.

229 / 231
 I Quantization: divide each DCT coefficient with the corresponding
value from an 8 × 8 table, then round to the nearest integer:
The two standard quantization-matrix examples for luminance and chrominance are:
16 11 10 16 24 40 51 61 17 18 24 47 99 99 99 99
12 12 14 19 26 58 60 55 18 21 26 66 99 99 99 99
14 13 16 24 40 57 69 56 24 26 56 99 99 99 99 99
14 17 22 29 51 87 80 62 47 66 99 99 99 99 99 99
18 22 37 56 68 109 103 77 99 99 99 99 99 99 99 99
24 35 55 64 81 104 113 92 99 99 99 99 99 99 99 99
49 64 78 87 103 121 120 101 99 99 99 99 99 99 99 99
72 92 95 98 112 100 103 99 99 99 99 99 99 99 99 99
I apply DPCM coding to quantized DC coefficients from DCT
I read remaining quantized values from DCT in zigzag pattern
I locate sequences of zero coefficients (run-length coding)
I apply Huffman coding on zero run-lengths and magnitude of AC
values
I add standard header with compression parameters
http://www.jpeg.org/
Example implementation: http://www.ijg.org/

230 / 231
Outlook

Further topics that we have not covered in this brief introductory tour
through DSP, but for the understanding of which you should now have a
good theoretical foundation:
I multirate systems
I effects of rounding errors
I adaptive filters
I DSP hardware architectures
I sound effects

If you find any typo or mistake in these lecture notes, please email [email protected].

231 / 231
Some final thoughts about redundancy . . .
Aoccdrnig to rsceearh at Cmabrigde Uinervtisy, it deosn’t
mttaer in waht oredr the ltteers in a wrod are, the olny
iprmoetnt tihng is taht the frist and lsat ltteer be at
the rghit pclae. The rset can be a total mses and you can
sitll raed it wouthit porbelm. Tihs is bcuseae the huamn
mnid deos not raed ervey lteter by istlef, but the wrod as
a wlohe.

. . . and perception
Count how many Fs there are in this text:
FINISHED FILES ARE THE RE-
SULT OF YEARS OF SCIENTIF-
IC STUDY COMBINED WITH THE
EXPERIENCE OF YEARS