 Presentation 1:

Signals and Systems and Introduction to Digital Signal

Processing.
 Presentation 2:
Transformation of Signals and Applications
 Presentation 3:
Speech Enhancement using Empirical Mode Decomposition
 Presentation 4:
A study on Bivariate Empirical Mode Decomposition
 Presentation 5:
Speech Enhancement using Bivariate EMD
 SIGNALS AND SYSTEMS
AND
INTRODUCTION TO DIGITAL
14/11/2012
SIGNAL PROCESSING
 Signal Processing deals with the enhancement, extraction, and representation
of information for communication or analysis
 Many different fields of engineering rely upon signal processing technology
 Acoustics, telephony, radio, television, seismology, and radar are some
examples
 Initially, signal processing systems were implemented exclusively with analog
hardware
 However, recent advances in high-speed digital technology have made discrete
signal processing systems more popular.
 Digital systems have an advantage over analog systems in that they can
process signals with an extraordinary degree of precision
 Unlike the resistive and capacitive networks of analog systems, digital systems
can be built numerically with the simple operations of addition and
multiplication.
 Digital Signal Processing is a field of numerical mathematics that is concerned
with the processing of discrete signals
 This area of mathematics deals with the principles that underlie all digital
systems
 Examples:
 Brain signals (EEG)
 Cardiac signals (ECG)
 Medical images (x-ray, PET, MRI)
 Goals:
 Detect abnormal activity (heart attack)
 Help physicians with diagnosis
 Tools: Filtering, Fourier Transform
 Identifying a person using physiological
characteristics
 Examples:
 Fingerprint Identification
 Face Recognition
 Voice Recognition
 Active noise cancellation: Adaptive
filtering
 Headphones used in cockpits
 Digital Audio Effects
 Add special music effects such as delay,
echo, reverb
 Audio signal separation
 Separate speech from interference
o Signals
o Sampling
o Systems
o Periodic signals
o Discrete time sinusoidal signals
o Exponential Signals
o Linear Time Invariant Systems
o Unit Impulse
o Convolution
o Correlation
 A signal is any physical phenomenon
which conveys information
 Systems respond to signals and produce
new signals
 Excitation signals are applied at system
inputs and response signals are produced
at system outputs
• Time is often the independent variable
for a signal. x(t) will be used to represent
a signal that is a function of time, t.

• A temporal signal is defined by the

relationship of its amplitude
(the dependent variable) to time
(the independent variable).
• An independent variable can be 1D (time),
2D (space), 3D (space) or even something
more complicated.

• The signal is described as a function of this

variable.

• There are many types of functions that can

be used to describe signals (continuous,
discrete, random).
 Signals can be analog or digital.
 Analog signals can have an infinite
number of values in a range.
 Digital signals can have only a limited
number of values.
• Most of the signals in the physical world are CT signals, since
the time scale is infinitesimally fine (e.g., voltage, pressure,
temperature, velocity).
• Often, the only way we can view these signals is through a
transducer, a device that converts a CT signal to an
electrical signal.
• Common transducers are the ears, the eyes, the nose… but
these are a little complicated.
• Simpler transducers are voltmeters, microphones, and
pressure sensors.
Amplitude
5
4
3

x(t) = 5 cos (2.pi.f.t)

2
1
0
-10 -5 -1 0 5 10
-2

Phase
-3
-4
-5

x(t) = 5 cos (2.pi.f.t + 3.14) 3

-10 -5 -1 0 5 10

Frequency -3

-5

x(t) = 5 cos (3. 2.pi.f.t + 5

3.14) -10 -5
1

-1 0 5 10
-3

-5
 The Independent Variable is Time
 The Dependent Variable is the Amplitude
 Most of the Information is Hidden in the Frequency Content
1 1

0.5 0.5

Magnitude
2 Hz
Magnitude

0 0
10 Hz
-0.5 -0.5

-1 -1
0 0.5 1 0 0.5 1
Time Time
1 4

2 Hz +
20 Hz
0.5 2
Magnitude

Magnitude

0 0 10 Hz +
-0.5 -2 20Hz
-1 -4
0 0.5 1 0 0.5 1
Time Time
 Stationary signal
• Frequency content of stationary signals do
not change in time.
• All frequency components exist at all times
 Non-Stationary signal
• Frequency content of stationary signals
change in time.
Magnit
ude

20 Hz 80 Hz 120 Hz
 Speech is the message content or information conveyed
Noise is the unwanted signal that interfere
Speech
3000

2000

• Broad band (100-10000Hz)

1000

Amplitude
0

• Non-stationary -1000

-2000

-3000
0 1 2 3 4 5 6
Time (sec.)

-40 -40
Crowd Factory
-60 -60

Noise
PSD

PSD
-80 -80

-100 -100

• Unknown -120
0 1000 2000 3000 4000 5000
-120
0 1000 2000 3000 4000 5000
Frequency Frequency

• Slowly varying (fan) or

-40 -40
Station F16 cockpit
-60 -60

non-stationary (babble)
PSD

PSD
-80 -80

• wideband or narrow band

-100 -100

-120 -120
0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000
Frequency Frequency
• For voiced speech, when SNR is low- periodicity
and structure of speech are affected.
• For unvoiced, the low energy parts are highly amplified,
looking like voiced.
Voiced speech Unvoiced speech

3000 1500
2000 1000
Amplitude

Amplitude
1000 500
0 0
-1000 -500
-2000 -1000
-3000 -1500
0 50 100 150 200 0 50 100 150 200
Sample number Sample number

Voiced noisy speech Unvoiced noisy speech

3000 1500
2000 1000
Amplitude
Amplitude

1000 500
0 0
-1000 -500
-2000 -1000
-3000 -1500
0 50 100 150 200 0 50 100 150 200
Sample number Sample number
Almost all biological signals are non-stationary. Some of the most
famous ones are ECG (electrical activity of the heart,
electrocardiograph), EEG (electrical activity of the brain,
electroencephalogram), and EMG (electrical activity of the
muscles, electromyogram).

ECG

EEG

EMG
 ADC
Microphone converts acoustic to
electrical energy. It’s a transducer.
Continuously varying electrical energy is an
analog of the sound pressure wave.
ADC (Analog to Digital Converter) converts
analog to digital electrical signal.
Digital signal transmits binary numbers.
DAC (Digital to Analog Converter) converts digital
signal in computer to analog for your headphones.
Instantaneous amplitudes of
continuous analog signal, measured
at equally spaced points in time.

A series of “snapshots”
Sampling Rate
How often analog signal is measured
[samples per second, Hz]

Sampling Resolution
[ “sample word length,” “bit depth”]
Precision of numbers used for measurement:
the more bits, the higher the resolution.
Example: 16 bit
 Determines the highest frequency that you can
represent with a digital signal.
Sampling/Nyquist Theorem:
Sampling rate must be at least twice as high as
the highest frequency you want to represent.

Capturing just the crest and trough of a sine

wave will represent the wave exactly.
Speech BW=0-5/6kHz, Telephone Speech BW=0-4kHz, Music BW=0-22kHz
What happens if sampling rate not high
enough? A high frequency signal

sampled at too low a rate

looks like …

… a lower frequency signal.

That’s called aliasing or foldover. An ADC has a low-

pass anti-aliasing filter to prevent this.
Synthesis software can cause aliasing.
An anti-aliasing filter removes
frequencies that are higher than half the
sampling rate using what is called a low
pass filter. A low pass filter lets the low
frequencies “pass" and "cut" the high
frequencies. Low pass filters are
sometimes called high cut filters.
Which rates can represent the range of
frequencies audible by (fresh) ears?
Sampling Rate Uses
44.1 kHz (44100) CD, DAT
48 kHz (48000) DAT, DV, DVD-Video
96 kHz (96000) DVD-Audio
22.05 kHz (22050) Old samplers

Most software can handle all

these rates.
 Sampling results in a series of pulses of
varying amplitude values ranging between two
limits: a min and a max.
 The amplitude values are infinite between
the two limits.
 We need to map the infinite amplitude values
onto a finite set of known values.
 This is achieved by dividing the distance
between min and max into L zones, each of
height ∆.
∆ = (max - min)/L
 Each zone is then assigned a binary code.
 The number of bits required to encode the
zones, or the number of bits per sample as it
is commonly referred to, is obtained as
follows:
nb = log2 L

 Given our example, nb = 3

 The 8 zone (or level) codes are therefore:

000, 001, 010, 011, 100, 101, 110, and 111
 A 3-bit binary (base 2) number has 23 = 8 values.

7
6
5
Amplitude

4
3
2
1
0
Time — measure amp. at each tick of sample clock
 A 4-bit binary number has 24 = 16 values.
14
12
10
Amplitude

8
6
4
2
0
Time — measure amp. at each tick of sample clock

A better approximation
 When a signal is quantized, we introduce an
error - the coded signal is an approximation
of the actual amplitude value.
 The difference between actual and coded
value (midpoint) is referred to as the
quantization error.
 The more zones, the smaller ∆ which results
in smaller errors.
 BUT, the more zones the more bits required
to encode the samples -> higher bit rate
Round-off error: difference between actual
signal and quantization to integer values…

Random errors:
sounds like low-
amplitude noise
Quantization
Quantization is the process of
converting the sampled analog voltages
into digital words.

Data coding
Data coding separates the digital words
so that they are more easily identified.
 To reconstruct analog signal, hold each sample value for
one clock tick; convert it to steady voltage.

7
6
5
Amplitude

4
3
2
1
0
Time
 Apply an analog low-pass filter to the output of the
sample-and-hold unit: averages “stair steps” into a smooth
curve.

7
6
5
Amplitude

4
3
2
1
0
Time
 Discrete-time signals are represented by sequence of
numbers
 The nth number in the sequence is represented with x[n]
 Often times sequences are obtained by sampling of
continuous-time signals
 In this case x[n] is value of the analog signal at xc(nT)
 Where T is the sampling period
10

-10
0 20 40 60 80 100 t (ms)
10

-10
0 10 20 30 40 50 n (samples)
 Even

x ( n) = x ( − n)

 Odd

x ( n) = − x ( − n)
• Any signals can be expressed as a
sum of even and odd signals. That is:

x(n) = xeven (n) + xodd (n)

where :
xeven (n) = [ x(n) + x(−n)] / 2
xodd (n) = [ x(n) − x(−n)] / 2

• This is demonstrated to the right

for a signal referred to as a unit
step.
 Unit sample sequence  Unit step sequence

1.5
1
0.5
δ ( n) = u( n) − u( n − 1)
0 ∞
-10 -5 0 5 10
1.5
u( n) = ∑ δ ( n − m )
1
m =0
0.5
0
-10 -5 0 5 10  Exponential sequence
1

0.5 x[n] = Aαn

0
-10 -5 0 5 10
 δ (n) unit sample sequence

0.8
Amplitude

0.6

0.4

0.2

0
-10 -5 0 5 10 15 20

δ ( n − 5)
n

0.8
Amplitude

0.6

0.4

0.2

0
-10 -5 0 5 10 15 20
n
u(n) unit step sequence

0.8
Amplitude

0.6

0.4

0.2

0
-5 0 5 10 15 20

u( n − 5)
n

0.8
Amplitude

0.6

0.4

0.2

0
-5 0 5 10 15 20
n
 Operations on sequence
 Time-shifting operation
where is an integer

delaying operation
Unit delay x (n) z-1 y( n) = x ( n − 1)

advance operation
Unit x (n) z y( n) = x ( n + 1)
advance
Time-shifting operation
Time-shifting
0.2
original sequence

operation
0.2 × 0.8 n u( n)
Amplitude

0.1

0 n
-10 -5 0 5 10 15 20 25 30
delayed sequence
0.2

0.2 × 0.8 n − 5 u( n − 5)
Amplitude

0.1

0
-10 -5 0 5 10 15 20 25 30 n
advanced sequence
0.2

0.2 × 0.8 n + 5 u( n + 5)
Amplitude

0.1

0 n
-10 -5 0 5 10 15 20 25 30
 Time-reversal (folding) operation

 Addition operation
Sample-by-sample addition

Adder x (n) y( n) = x ( n) + w ( n)

w (n)
 folding operation
olding operation 1
original sequence

0.8 n u( n)
0.8
Amplitude

0.6

0.4

0.2

0
-20 -15 -10 -5 0 5 10 15 20
n
folding sequence
−n
1
0.8 u( − n)
0.8
Amplitude

0.6

0.4

0.2

0
-20 -15 -10 -5 0 5 10 15 20
n
addition operation
1
x1(n) 0.8 n u( n)
Amplitude

0.5

0
0 5 10 15 20 25 30 35 40 n
x2(n)
1 cos(0.2n)u( n)
Amplitude

-1
0 5 10 15 20 25 30 35 40 n
x1(n)+x2(n)
2 0.8 n u( n) + cos(0.2n)u( n)
Amplitude

-1
0 5 10 15 20 25 30 35 40 n
 Scaling operation

Multiplier x (n) A y( n) = Ax ( n)

 Product (modulation) operation

Sample-by-sample multiplication

modulator x (n) y( n) = x ( n) ⋅ w ( n)

w (n)
modulation operation x1(n) 0.1 sin 0.0125nπ
0.1
Amplitude

-0.1
0 20 40 60 80 100 120 140 160
x2(n)
1
sin 0.125nπ
Amplitude

-1
0 20 40 60 80 100 120 140 160
0.1
x1(n)*x2(n)
x1 ( n) ⋅ x 2 ( n)
Amplitude

-0.1
0 20 40 60 80 100 120 140 160
 Decimation---down-sampling

y ( m ) = x ( mN )
x(n)
 Decimation---down-sampling

y ( m ) = x ( mN )
 Decimation---down-sampling

y ( m ) = x ( mN )

y(m)
 Interpolation --- up-sampling

y ( m) = x ( m / M )
 Interpolation --- up-sampling

y ( m) = x ( m / M )
1.5 × cos( 0.05 × 2π n )

Sin u so id al seq u en ce
2

1
Amplitude

-1

-2
0 10 20 30 40 50 60

1.5 × sin( 0.05 × 2π n )

n

1
Amplitude

-1

-2
0 10 20 30 40 50 60
n
  Introduction

A discrete-time system processes a given input

sequence x(n) to generate an output sequence y(n)
with more desirable properties.

Mathematically, an operation T [ • ] is used.

y(n) = T [ x(n) ]
x(n): excitation, input signal
y(n): response, output signal
 Classification

 Linear System
 Time-Invariant (Shift-Invariant) System
 Linear Time-Invariant (LTI) System
 Causal System
 Stable System
 Memory System
 Linear System

A system is called linear if it has two mathematical

properties: homogeneity and additivity.

T [a1 x1 ( n) + a 2 x 2 ( n)] = a1T [ x1 ( n)] + a 2T [ x 2 ( n)]

 Time-Invariant (Shift-Invariant) System

Input output characteristics do not

change with time.

if T [ x ( n)] = y( n)
then T [ x ( n − n0 )] = y( n − n0 )
 Linear Time-Invariant (LTI) System

A system satisfying both the linearity and the time-

invariance properties is called an LTI system.

LTI systems are mathematically easy to analyze and

characterize, and consequently, easy to design.
The output of an LTI system is called
linear convolution sum
+∞ ∆
y( n) = LTI [ x ( n)] = ∑ x(k )h(n − k ) =x(n) * h(n)
k = −∞

An LTI system is completely characterized in the

time domain by the impulse response h(n).
 x(n) x ( n) = R10 ( n)
1.5
Amplitude
1

0.5

0
0 5 10 15 20 25 30 35 40 45 50
h(n)
1.5
h( n) = 0.9 n u( n)
Amplitude

0.5

0
0 5 10 15 20 25 30 35 40 45 50
y(n)
10
y( n) = x ( n) ∗ h( n)
Amplitude

0
0 5 10 15 20 25 30 35 40 45 50

Copyright © 2005. Shi Ping

return
CUC
  Causal System
In a causal system, the -th output sample
depends only on input samples for and
does not depend on input samples for

e.g.

For a causal system, changes in output samples

do not precede changes in the input samples.
An LTI system will be a causal system if and
only if :
h( n) = 0, n<0

An ideal low-pass filter is not a causal system !

A sequence is called a causal sequence if :

 Stable System
A system is said to be bounded-input bounded-
output (BIBO) stable if every bounded input
produces a bounded output, i.e.

An LTI system will be a stable system if and

only if : ∞
S= ∑ h(n) < ∞
n = −∞
 Memory:
 A system is memoryless if y[n] = f ( x[n] )
▪ i.e. it sees only present values.
 A system has memory if y [n] depends on
previous values
▪ it can also depend on present and future values!
 Consider the DT SISO system:
System
x[ n] y[ n]

 If the input signal is x[n] = δ [n] and the system has no

energy at y[ n] = h[ n] , the output n = 0
is called the impulse response of the system

δ [ n] System h[n]
Linear Time-Invariant Systems
and Convolution
Linear Time-Invariant Systems
and Convolution
Linear Time-Invariant Systems
and Convolution
Linear Time-Invariant Systems
and Convolution
Linear Time-Invariant Systems
and Convolution
Linear Time-Invariant Systems
and Convolution
 Correlation addresses the question: “to what
degree is signal A similar to signal B.”
 An intuitive answer can be developed by
comparing deterministic signals with
stochastic signals.
 Deterministic = a predictable signal equivalent to
that produced by a mathematical function
 Stochastic = an unpredictable signal equivalent to
that produced by a random process
 Correlation is maximum when
 Two signals are similar in shape
 And are in phase (or unshifted)
 Correlation is measure of similarity
between two signals as a function of
time shift between them
 Correlation functions shows how similar two
signals are, and how long they remain similar when
one is shifted with respect to the other
  Correlating a signal with itself

∞
rxx ( l ) = ∑ x ( n ) x ( n − l ) = r ( −l )
n =−∞
xx l = 0, ±1, ±2,
Cross-correlation of x(n) and y(n) is a
sequence, rxy(l)
∞
rxy ( l ) = ∑ x (n) y (n − l )
n = −∞
l = 0, ±1, ±2,
∞
rxy ( l ) = ∑ x (n − l ) y (n)
n = −∞
l = 0, ±1, ±2,
Thanks for your
attention