01OGGPT - Signal analysis and processing

PROF. FABIO DOVIS

Prof. Fabio Dovis

 Politecnico di Torino - Department of Electronics and Telecommunications
Digital Signal Processing

 Digital signal processing (DSP) is the application of an appropriate

sequence of arithmetic or logical operations (algorithm) to a series of
digital values series (e.g. binary digits) which represents (exactly or
sufficiently approximate) a signal (which if, generally, originally
analog).

 Purposes of the processing:

 Modify the signal
 Improve its quality
 Extract information from it

 DSP is now extensively used in the largest part of electronic devices

Master Veicoli Speciali

1
From DSP to PC …

 Until a few years ago, the ENS was the almost exclusive preserve of
microprocessors specifically dedicated to the purpose: the DSP (Digital Signal
Processor).
 DSP devices use a particular architecture specially designed to perform many mathematical
operations (also parallel) in extremely short times.
 The speed of calculation is essential when it is necessary to process a (digital) signal in real
time.

 Recently, the unstoppable rise in performance of Personal Computers made

many algorithms that were designed for DSPs usable on CPUs of normal PC.

 Image processing
 Audio processing
 …

Master Veicoli Speciali

2
Examples of DSP applications

 VOICE SIGNALS PROCESSING

 Storage of signals (CD, DVD, MP3)
 Signal transmission (Mobile telephony)
 Improvement of certain qualities of a signal (for example, the intelligibility of the human
voice)
 Synthesis of signals that simulate human voice.
 Recognition of sentences spoken by a speaker.
 Identification of the speaker.

 PROCESSING OF SEISMIC SIGNALS

 Location and measurement of the intensity of earthquakes.
 Assistance to drilling operations for oil exploration.

 PROCESSING OF RADAR SIGNALS

 Location and calculation of the speed and trajectory of an object

Master Veicoli Speciali

3
Examples of DSP applications

 IMAGE PROCESSING
 Signal storage (DVD, JPEG, MPEG)
 Signal transmission (digital terrestrial TV, satellite TV)
 Photographs, taken from space probes, of the earth, the moon and other
planets.
 Weather maps.
 Reconstruction of images from projections (TAC)

 The image captured by the lens of the camera is transformed into a

digital signal, then manipulated, generally to improve its readability,
i.e. to facilitate the extraction of that information for which the image
was taken

Master Veicoli Speciali

4
Starting for analog signals…

 DSP works with digital signals, i.e. represented by vectors of integer

numbers (stored in RAM and binary-encoded)
BUT
 The "real" signals, normally processed by "classic" electronics are by
their nature strictly analog:
 Electrical signals, representative of signals of other nature (for example music
and / or voice) that vary continuously over time
 They can take any value between a minimum and a maximum, continuously
varying over time.

Master Veicoli Speciali

5
… to digital signals

 The "understandable" signals to a PC or DSP are instead numbers

(generally encoded in binary code), whose value represents the signal
at a well-defined instant of time, ie those signals called digital.

 To apply DSP techniques to "real" signals, it is necessary to convert

them into digital format (and vice versa) analog analog-to-digital
(ADC) and digital-to-analog (DAC) converters.

Analog Digital DAC

signal signal
ADC DSP

Feature
extraction
Sampling Theorem

Master Veicoli Speciali

6
Analog-to-digital conversion

Analog Sampling

t
t
Quantization 1010
1010
1001
1001
1000
Digital
1000 0111
0111 0110
0110 0101
0101 0100
0100 0011
0011 0010
0010 0001
0001 0000
0000
t
t 0011 – 0011 – 0100 – 0100 – 0100 – 0100 – 0010 - 0010

7
Discrete-Time signals

 In general, discrete-time signals are defined with respect to an

independent variable that takes only integer values, generically
indicated with the letter 𝑛.

 These signals are therefore represented by a sequence of numbers

indexed by the discrete time variable 𝑛 : 𝑥(𝑛), 𝑥[𝑛], 𝑥𝑛

 𝑥(𝑛) is called "numeric" (or "digital") if it assumes only discrete

amplitudes.

Master Veicoli Speciali

8
Discrete-Time signals

 «Intrinsically» Discrete signals:

 stock quotes, defined at regular time intervals
 numbers deriving from lottery draws
 temperature measurements made at discrete time intervals.

 Discrete signals obtained by "sampling" analog signals:

 samples of a musical signal stored on a CD or on the hard disk of our PC
 samples of an image transmitted over the network.

Master Veicoli Speciali

9
Duration of a sequence

 A sequence can have:

 finite duration, if it is identically zero outside a finite interval of time [𝑛1, 𝑛2]
 infinite duration, if it lasts in an infinite time interval, which can be bilateral (−∞, +∞) or
monolateral [𝑛1, + ∞) 𝑜𝑟 (−∞, 𝑛2).
 The temporal support of a sequence of finite duration is equal to 𝑁 = 𝑛2 − 𝑛1 +
1.

x n 

n1 n2 n

Master Veicoli Speciali

10
Causality

 A sequence is said to be causal if it is identically null for values 𝑛 < 0,

anti-casual if it is identically null for values 𝑛 ≥ 0.
 A sequence that presents non-zero samples in both the positive and
negative axes of the discrete time is called the bilateral.

x n  x n 
causal anti-causal

n n

11
Parity

 A real sequence 𝑥(𝑛) (i.e. made of real values) is:

 Even (symmetric) , if 𝑥(𝑛) = 𝑥(−𝑛)
 Odd (anti-symmetric), if 𝑥(𝑛) = −𝑥(−𝑛)

x n  x n 

even odd

n
n

12
Parity

 A complex sequence 𝑥(𝑛) (i.e. made of complex values) is:

 conjugate symmetric, if 𝑥(𝑛) = 𝑥 ∗ (−𝑛)
 conjugate anti-symmetric, if 𝑥(𝑛) = −𝑥 ∗ (−𝑛)

 Any complex sequence 𝑥(𝑛) can be written as the sum of a conjugate

symmetric sequence 𝑥𝑠 (𝑛) and a congjugate anti-symmetric one 𝑥𝑎𝑠 (𝑛):

1 1 ∗
𝑥𝑠 𝑛 = 𝑥 𝑛 + 𝑥 −𝑛
2 2

1 1 ∗
𝑥𝑎𝑠 𝑛 = 𝑥 𝑛 − 𝑥 −𝑛
2 2

13
Parity

 conjugate symmetric,
if 𝑥(𝑛) = 𝑥 ∗ (−𝑛)

14
Periodicity

 A sequence 𝑥(𝑛) is periodic if it is possible to find a discrete time

interval N so that 𝑥 𝑛 = 𝑥(𝑛 ± 𝑁)

 The period is the smallest positive integer value 𝑁 for which the
previous relation is valid.
 Note that if a sequence is periodic of period 𝑁 it is periodic also of 𝑘𝑁 (being
𝑘 a positive integer value).
x n 

15
Sequences with limited amplitude

 A sequence 𝑥(𝑛) is limited if for any dicrete time value 𝑛 it takes

values in a finite interval, i.e.:

xn   X 0   , n

being 𝑋0 a real finite constant and positive value.

16
Summable sequences

Sequences summable in absolute sense



 xn  
n  

Sequences summable in quadratic sense



 xn  
2

n  

17