•  This course is about signals and systems •  Electrocardiogram (ECG) recording

that processes these signals instrument that generates various signals by

Examples combining multiple signals from electrodes
•  Music recorded in a CD is a signal and the attached to the body is a system
CD player that converts into a sound wave •  Automobile cruise control system uses the
is a system signal that is the difference between the
speed set by the driver and the actual speed
of the car to maintain a constant speed of
1 2
the car by adjusting the throttle
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  A system processes a specified signal or a •  The mathematical model of a signal is given

set of signals, called the input to generate as a function of one or more independent
one or more signals called the output to variables
carry out a specific task •  For example, the speech and the musical
•  Signals and systems are important in many sound signals are functions of a single
disciplines outside engineering such as variable which is time
biology, chemistry, physics, etc. •  The image captured by a camera is a
function of two spatial variables
3 4
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  The video signal is a function of three Continuous-Time and Discrete-Time

independent variables, two spatial variables Signals
and time •  The continuous-time signal, more
•  The algorithm for the processing of signals commonly known as the analog signal, is a
is thus a mathematical operation carried out function of one or more independent
on the assumed signal model continuous variables
•  We first consider the general classification •  It is defined at every value of the
of signals independent variable(s)
5 6
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

1
 •  The discrete-time signal, more commonly •  Figure below shows the two basic types of
known as the digital signal, is a function of signals
one or more independent discrete variables
•  It is defined at specific values, usually
integer-valued values, of the discrete
Continuous-time signal Discrete-time signal
variables
•  Value of the signal at a specific value(s) of
•  It is not defined for any other values of the the independent variable(s) is called its
discrete variables amplitude
7 8
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  Two special cases of the above two types of •  A typical boxcar signal is shown below on
signals the left
•  The boxcar signal where the amplitude of
the signal is maintained at a constant level
between two consecutive instants of
clocking Boxcar signal Digital signal
•  Amplitude of a boxcar signal is defined at
all values of time and can take any value •  The digital signal is shown above on the
right hand side
9 10
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  The amplitude of the digital signal at any One- and Multi-Dimensional Signals
particular instant of time is represented by a •  The one-dimensional (1-D) signal is a
finite number of digits, usually binary bits function of a single independent variable
•  It can take a value within a finite set of •  The multi-dimensional signal is a function
discrete values of more than one independent variable
•  In this course, we shall refer to a discrete- •  For example, the two-dimensional (2-D)
time signal as a digital signal if there is no signal is a function of two independent
ambiguity variables
11 12
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

2
 •  The three-dimensional (3-D) signal is a Deterministic Signals
function of three independent variables •  A signal generated by a precisely defined
•  Example of a 2-D signal is an image or a process is known as a deterministic signal
photograph •  Such signals are uniquely defined by a well-
•  Example of a 3-D signal is the video signal defined rule and can be repeatedly
generated in the same form

13 14
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

Random Signals
•  A signal that cannot be uniquely determined •  The plot of the variation of the amplitude of
a signal as a function of the independent
by any well-defined process is called a
variable is called its waveform
random signal
•  The objective of modeling a signal is to
•  A random signal is usually a member of an
ensure that waveform of the model of the
ensemble of signals and its form cannot be
signal is as close as possible to the original
predicted in advance
waveform
•  Statistical techniques are used for the
analysis of such signals
15 16
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

Analog Signals •  The independent variable of a one-

•  A one-dimensional analog signal is a dimensional analog signal will be denoted
function of the independent real variable as t and show its functional dependence
which is assumed to take all values in the inside parentheses as (t)
range from −∞ to +∞ •  For example, x(t) represents a one-
•  We shall refer to the independent variable dimensional analog signal
as time independent of the nature of the •  The value x(t o ) of the analog signal at a
variable
€ € particular instant of time t = t o is called its
17 18 amplitude
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

€
€

3
 •  The two continuous independent variables •  In practice, the amplitude of an analog
of a 2-D analog signal will be denoted as signal is limited between a range of two
u and v values
•  For example, x(u,v) is a 2-D analog signal •  The range of the largest and smallest
•  The two independent variables of a black- possible values is called the dynamic range
and-white image are the two spatial •  Their ratio is usually specified in decibel
coordinates

19 20
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

Digital Signals •  A discrete-time signal is thus a sequence of

•  The independent variable of a discrete-time numbers
1-D signal usually is denoted by n •  The digitized image is an example of a 2-D
•  For example, {x[n]} denotes a discrete-time discrete-time signal whose independent
1-D signal variables are discrete spatial variables,
•  The n-th sample of the 1-D discrete-time denoted by m and n
signal {x[n]} is x[n] whose numerical value •  For example, {x[m,n]} represents a digitized
is called the amplitude of the signal image
21 22
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  In most applications, the independent •  For example, the discrete-time signal x[n]
discrete variable n of a discrete-time 1-D or generated by periodically sampling the
the variables m and n of a discrete-time 2-D continuous-time signal xa (t) at uniform
signal, are assumed to take integer values, time intervals is given by
where −∞ < n < +∞ and −∞ < m < +∞ x[n] = xa (t) t=nT = xa (nT )
•  In some applications, a continuous-time € … , − 2, −1, 0, 1, 2,…
signal is converted into a discrete-time
€ signal for processing
€ using a discrete-time
€
23
system 24
Copyright © 2015, S. K. Mitra € Copyright © 2015, S. K. Mitra

4
 •  Figure below shows the discrete-time signal •  The sampling frequency or the sampling
generated from a continuous-time signal rate, denoted as FT , is given by the number
obtained by uniform sampling of samples per unit time
•  The reciprocal of the sampling frequency
€ T = 1/ FT
is known as the sampling period or
sampling interval, which is the spacing T
between two consecutive samples x[n] and
€
x[n +1]
25 26
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  The amplitude of a digital signal at any Power Line Signal

discrete time instant, in practice, is bounded •  The household ac electrical current shown
between two values below is an example
•  The difference between its largest and
smallest possible values is its dynamic
range, and the ratio of these two values is
specified in decibel

27 28
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  It is modeled mathematically as •  One full cycle of the signal, called the

x(t) = 170sin(120πt) period, is of length 1/ 60 ≅ 16.7 ms
where t is in seconds and the amplitude is in •  Inverse of the period is called the cyclic
volts frequency or simply the frequency, which is
•  The here 60 Hz€
€ signal is seen to be a periodic signal as
the shape of the signal repeats periodically

29 30
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

5
 •  The sinusoidal signal is a basic continuous- Musical Sound Signal
time signal •  The signal generated by a musical
•  Almost all continuous-time signals can be instrument is another example of an analog
modeled as a weighted sum of the signal
sinusoidal signals of different cyclic •  The signal created by hitting the key of a
frequencies piano is a tonal signal or a pure tone which
is a sinusoidal signal

31 32
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  The frequency of the tonal signal depends

on the note that has been played
•  Figure in the next slide shows the 12 pure
tonal signals along with their respective
cyclic frequencies generated by playing the Note B
piano keys
•  The tonal signal represented by the note B is
shown in the next slide
33 34
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  The signals generated by many musical

instruments playing the pure tones are not
pure sinusoidal signals
•  They are more complicated in form as can
be seen in the waveform of a trumpet
•  It is a pure sinusoidal signal with a cyclic playing the note B shown in the next slide
frequency of 494 Hz
•  Its period is of length 1/ 494 ≅ 2 ms
35 36
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

6
 •  These musical signals usually are
represented by a weighted sum of the tonal
signal and sinusoidal signals of cyclic
frequencies that are multiples of the cyclic
frequency of the tonal signal being played
•  Note that the actual trumpet signal is also a
periodic signal with a period that is same as
that of the pure tonal signal of the note B
37 38
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  The cyclic frequency of the tonal signal is Time Series

called the pitch or fundamental frequency •  Often digital signals encountered in practice
•  The cyclic frequencies that are integer are observed for a finite duration of time
multiples of the pitch frequency are called •  Examples of such signals are daily stock
the overtones or harmonics prices, the value of total monthly exports of
a country, the annual yields per acre of
crops in a country, etc.

39 40
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  This type of finite extent signal, usually •  Seasonally adjusted quarterly Gross
called a time series, occurs in business, National Product of the United States in
economics, physical sciences, social 1982 dollars from 1976 to 1986
sciences, engineering, medicine, and many
other fields
•  Plots of some typical time series are shown
in the next slide

41 42
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

7
 •  Monthly mean St. Louis, Missouri, Telegraph Signals
temperature in degrees Celsius for the years •  One of the earliest example developed to
1975 to 1978 represent textual messages for transmission
over wires for long distances
•  It is formed by a sequence of pulses of two
different widths

43 44
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  The short and long width pulses are called Images Captured by Digital Cameras
dot and dash, respectively •  Inexpensive cameras use a single sensor
array with each pixel originally capturing
one of the three primary colors, Red (R),
•  Morse code representation of the word Green (G), and Blue (B), of the RGB color
SIGNAL is shown above coordinate system

45 46
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  A full color image is then generated by •  Figure below shows a 24-bit color image
some type of color interpolation “peppers”
•  A black-and-white (gray-level) image is
usually formed by a color conversion from a
color image in the RGB space

47 48
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

8
 •  Figure below shows the black-and-white Cruise Control System
version of “peppers” obtained by a color •  Here, the difference between the actual
conversion speed of the car, as measured by a sensor,
and the desired speed set by the driver is
used to adjust the throttle to maintain a
constant speed

49 50
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  The free-body diagram of the simplified •  u(t) is the force generated at the tire/road
model of pertinent system is shown below interface which is applied to the car with a
mass m
•  v(t) is the velocity
•  The force bv(t) due to the wind drag and
rolling resistance acts in the opposite
direction to the car's forward motion

51 52
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  u(t) is the input signal Automobile Shock Absorber System

•  v(t) is the output signal •  The system consists of a shock absorber and
•  The system equation modeling the a coil spring attached to the car body next to
simplified free-body diagram is given by the wheel and the wheel assembly whose
free-body diagram is shown in the next slide
m dv(t) + bv(t) = u(t)
dt
•  Output equation governing the speed of the
car is y(t) = v(t)
53 € 54
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

9
 •  Forces acting on the system are
Fc (t) is the upward inertial force
Fs (t) is the spring force
Fd (t) is the damper force
•  m is the mass of the car € •  Input-output relations
•  b is the damping coefficient of the damper € d 2 y(t) b dy(t) k
+m + m y(t) = x(t)
•  k is the spring constant € dt 2 dt input
55 56 output
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

Square Root Computation Bank Account

•  The square root of a number α can be •  Consider the yearly growth of balance in a
calculated using bank account paying compound interest at
1 ⎛ α ⎞ the end of each year
y[n] = ⎜ y[n −1] + ⎟
2 ⎝ y[n −1]⎠ •  Denote the balance in the account at the end
where 0 ≤ n < ∞ of the n-th year as x[n]
•  As n → ∞ , y[n] → α •  Let the interest being paid by the bank be
5% annually
€
57 58
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra
€
€ €

•  Assume that the initial amount in the x[1] = x[0] ×1.05 = 100 ×1.05 = 105.00
account is $100, that is, x[0] = 100 x[2] = x[1] ×1.05 = x[2] × (1.05)2 = 110.25
•  The algorithm to determine the balance at x[3] = x[2] ×1.05 = x[0] × (1.05) 3 = 115.76
the end of the n-th year is € 
x[n] = x[n −1] ×1.05 € •  It follows from the above set of equations
•  The balances at the end of the following € that
several years are € x[n] = x[0] × (1.05)n
€
59 60
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

10
 Data Interpolation •  The value of the function needed at any
•  In some applications, certain well-defined other values of the independent variable
functions of an independent variable are inside the specified range is determined by
pre-computed at equally-spaced values of applying an interpolation formula to a
the independent variable in a specific range subset of the stored data
and stored in a look-up table •  Two such functions that are often used in
signal processing applications are the sine
and cosine of an angle
61 62
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  To illustrate the basic idea, assume that the •  Suppose we need to determine the value of
values of the function sin(θ ) are given at 6 sin(0.15π ) by interpolation
equally-spaced values of θ in the range •  The simplest formula is to take the average
0 ≤ θ ≤ 0.5π as given below of the value of sin(0.1π ) and sin(0.2π ) :
€ € sin(0.1π ) + sin(0.2π )
sin(0.15π ) =

θ 0 0.1π € 0.2π 0.3π 0.4π 0.5π 2
€ sin(θ) 0 0.3090 0.5878 0.8090 0.9511 1.0 € 0.3090
€ + 0.5878
= = 0.4484
2
63 € 64
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

Noise Masking
•  The value of sin(0.15π ) computed using
MATLAB is 0.4540 •  Digital signals often get corrupted by
additive random noise
•  The simple averaging of two consecutive
data€samples is more commonly known as •  A mathematical representation of such
bilinear interpolation noisy signals is given by
x[n] = s[n] + η[n]
•  Two interesting applications of the bilinear
interpolation are presented later in this where x[n] is the noisy signal, s[n] is the
course uncorrupted signal and η[n] is the random
65 66 noise
€
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

11
 •  In some applications, the data obtained by •  Figures below and the in the next slide show
measurement is repeatable the Raman spectra of ethanol from a single
•  If the samples of the measured data are one-second collection, from 10 one-second
corrupted by additive noise, the amount of collections, and 100 one-second collections
noise can be significantly reduced by
averaging the data from a very large number
Single one-second collections
of measurements
•  This approach to noise masking is known as
67 ensemble averaging 68
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  In the bottom figure shown in the previous

slide, the peaks near relative wave numbers
of 1000 and 1500 cm −1 are clearly visible

€
10 one-second collections 100 one-second collections

69 70
Copyright © 2015, S. K. Mitra Copyright © 2015, S. K. Mitra

•  Simplest examples are the analog-to-digital

(A/D) and digital-to-analog (D/A)
converters
•  The A/D converter is a single-input,
multiple-output systems
•  The D/A converter is a multi-input, single-
output system

71
Copyright © 2015, S. K. Mitra

12