Signal Processing

(CSCI472 Signal Processing)

Prof. Walid Al-Atabany

Professor at the ITCS School
Nile University
[email protected]
Systems
• A system takes a signal as an input and transforms it into
another signal

= S[X(t)]

• In a very broad sense, a system can be represented as the

ratio of the output signal over the input signal
– That way, when we “multiply” the system by the input
signal, we get the output signal
– This concept will be firmed up in the coming weeks
 Systems Classifications

• Linear and nonlinear systems;

• Constant-parameter and time-varying-parameter systems;
(note constant = time invariant)
• Instantaneous (memoryless) and dynamic (with memory)
systems;
• Causal and non-causal systems;
• Continuous-time and discrete-time systems;
 LTI Continuous-Time Systems

• A continuous-time system is a system in which the signals at

its input and output are continuous-time signals
 Linearity

• A linear system is a system in which the superposition holds

 Scaling
o T[k x(t)] = k T[x(t)]
o k is a scalar constant and x(t) is an input

 Additivity
o T[x1(t) + x2(t)] = T[x1(t)] + T[x2(t)]
o x1(t) and x2(t) are two inputs
o x1(t) + x2(t) is a superposition (addition) of inputs
 Linearity
Examples:
• y(x)= a x Linear
• y(x)= a x + b Nonlinear

Whenever the explicit relation between the input and the output of a
system is represented by a nonlinear expression the system is nonlinear
 Linearity

• Transcendental system y(t ) = cos(x(t ))

 Answer: Nonlinear (in fact, fails both tests)

• Squarer y(t ) = x² (t )
 Answer: Nonlinear (in fact, fails both tests)
• Differentiation is linear
 Homogeneity test:

 Additivity test:
 Linearity

• Integration
 Homogeneity test

 Additivity test

Answer: Linear
• Human hearing
 Responds to intensity on a logarithmic scale
Answer: Nonlinear (in fact, fails both tests)
 Time Invariance

• System S does not change with time

• A system is said to be time invariant if, for any input x(t) and
any time t1, the response to the shifted input x(t – t1) is equal
to y(t – t1) where y(t) is the response to x(t) with zero initial
energy

• A system that is not time invariant is said to be time varying

or time variant
 Time Invariance

• A shift in the input produces a shift in the output by the same

amount
If T[x(t)] = y(t), then
T[x(t -t)] = y(t -t) for all real-valued time shifts t

Time Varying Systems

• Amplifier with Time-Varying Gain y(t) = tx(t)
• First-Order System 𝑦(𝑡)+ a(t) y(t) = bx(t)
 Time Invariance

• Identity system y(t ) = x(t )

– Step 1: compute yshifted(t) = x(t – t0)
– Step 2: does yshifted(t) = y(t – t0) ? YES.
• Answer: Time-invariant

• Tapped Delay Line

• Answer: Time-invariant
 Memoryless and Causality
• A system is memoryless if the output at time n depends only
on the input at that same time
• A continuous-time system S is called causal if:
▫ Whenever the input x(t)=0 and there are no initial
conditions, the output is y(t)=0
▫ The output y(t) does not depend on future inputs
 Memoryless and Causality

• A causal system that is not memoryless is said to have

memory. A system has memory if the output at time t1
depends in general on the past values of the input x(t) for
some range of values of t up to t = t1

• y(t) = y(t - 1) + x(t) + 2 x(t -1) is causal

• y(t) = y(t +1) + x(t +1) - 2 x(t -100) is NOT causal
 Memoryless and Causality
• A non-causal system and its realization by a delayed
causal system.
y(t) = f(t − 2) + f(t + 2)

• Starts even before the input is applied.

– If we are operating the system in real time at t, we do not
know what the value of the input will be two seconds later. Thus
it is impossible to implement this system in real time.
• Must delay to make causal
Bounded-Input Bounded-Output Stability (BIBO)
• Informally, a stable system is one in which small input signals lead to
responses that do not diverge
• If an input signal is bounded, then the output signal must also be
bounded, if the system is stable
x : x  U  y  V
• To show a system is stable we have to do it for all input signals. To
show instability, we just have to find one counterexample
• E.g. Consider the DT system of the bank account
y[n] = x[n]  1.01y[n  1]
• when x[n] = d[n], y[0] = 1
• This grows without bound, due to 1.01 multiplier. This system is
unstable.
 Bounded-Input Bounded-Output Stability (BIBO)

• For a bounded (i.e., well-behaved) input x(t), the output of a

BIBO stable system y(t) is also bounded
• All inputs cause system not to diverge (Bounded)

Stable

Unstable
 Systems Classification : Ex
Are the following systems
1. Memoryless, 2 Time-invariant, 3. Linear, 4. Causal, 5. Stable?
• Problems from Oppenheim
- P1.27, P1.28 and P1.31
 Representation of Systems by Differential
Equations
• Given a dynamic system represented by a linear differential
equation with constant coefficients:

▫ N initial conditions:
▫ Input x(t)=0 for t < 0,
• Complete response y(t) for t>=0 has two parts:
▫ Zero-state response
▫ Zero-input response
 RLC Circuits
• Kirchhoff’s voltage law,

• Second-order differential equation with constant coefficients

▫ Input the voltage source v(t)
▫ Output the current i(t)
 Representation of Systems by Differential
Equations
• Linear Time-Invariant Systems
▫ System represented by linear differential equation with
constant coefficients
▫ Initial conditions are all zero
▫ Output depends exclusively on input only
• Nonlinear Systems
▫ Nonzero initial conditions means nonlinearity
▫ Can also be time-varying
 Application of Superposition and Time
Invariance
• The computation of the output of an LTI system is simplified
when the input can be represented as the combination of
signals for which we know their response.
▫ Using superposition and time invariance properties

Time-
linearity Invariance
 Application of Superposition and Time
Invariance: Example
• Example 1: Given the response of an RL circuit to a unit step
source u(t), find the response to a pulse
 Convolution Integral
• Basic idea: use an (infinite) set of discrete time impulses to
represent any signal.
• Consider any discrete input signal x[n]. This can be written
as the linear sum of a set of unit impulse signals:

• Therefore, the signal can be expressed as:

 Convolution Integral
• In general, any discrete signal can be generically represented
as:

The sifting property

• Ex.
 Convolution Integral
• Generic representation of a signal:
• A very important way to analyse a system is to study the
output signal when a unit impulse signal is used as an input

• Loosely speaking, this corresponds to giving the system a

kick at n = 0, and then seeing what happens
• This is so common, a specific notation, h[n], is used to denote
the output signal, rather than the more general y[n].
• The output signal can be used to infer properties about the
system’s structure and its parameters.
 Convolution Integral
• The response of an LTI system S represented by its impulse
response h(t) to any signal x(t) is given by:

• Convolution is an operator that takes an input signal and

returns an output signal, based on knowledge about the
system’s unit impulse response h[n].
 Convolution Integral
• This is possible for LTI systems because they possess the
superposition property:

• Another Way
 Convolution Integral
• Discrete-time convolution • Continuous-time convolution

• For every value of n, we • For every value of t, we

compute a new summation compute a new integral
 Convolution Integral: Observations
• Impulse response is fundamental in the characterization of
linear time-invariant systems
• Any system characterized by the convolution integral is
linear and time invariant by the above construction
• The convolution integral is a general representation of LTI
systems, given that it was obtained from a generic
representation of the input signal
• Given that a system represented by a linear differential
equation with constant coefficients and no initial conditions,
or input, before t=0 is LTI, one should be able to represent
that system by a convolution integral after finding its
impulse response h(t)
 Graphical Convolution Methods
• Consider h[k]. Time-flip it and obtain h[-k]. Then time-shift it
by n and obtain h[n - k]. Since n can be an arbitrary number in
(-∞,∞), that means slide h[n-k] across x[n] while calculating
the convolution sum for that particular value of n
Graphical Convolution Methods
• From the convolution integral, convolution is equivalent to
Graphical Computation of Convolution Integral
• Example 1: Graphically find the unit-step y(t) response of
an averager, with T=1 sec, which has an impulse response
h(t)= u(t)-u(t-1)
Graphical Computation of Convolution Integral
• Example 2: Consider the graphical computation of the
convolution integral of two pulses of the same duration
 Graphical Computation of Convolution Integral
• Convolve the following two functions:

• Replace t with t in f(t) and g(t)

• Choose to flip and slide g(t) since it is simpler and symmetric
• Functions overlap like this:
 Graphical Computation of Convolution Integral
Convolution can be divided into 5 parts
I. t < - 2
• Two functions do not overlap
• Area under the product of the
functions is zero

II. - 2 < t < 0

• Part of g(t) overlaps part of f(t)
• Area under the product of the
functions is
 Graphical Computation of Convolution Integral
III. 0 < t < 2
•Here, g(t) completely overlaps f(t)
• Area under the product is just

IV. 2 < t < 4

• Part of g(t) and f(t) overlap
• Calculated similarly to -2 < t < 0

V. t ≥ 4
• g(t) and f(t) do not overlap
• Area under their product is zero
 Graphical Computation of Convolution Integral
Result of convolution (5 intervals of interest):
Discrete-time Convolution
• See example 2.3
Convolution Properties
– Commutative: f1(t) * f2(t) = f2(t) * f1(t)
– Distributive: f1(t) * [f2(t) + f3(t)] = f1(t) * f2(t) + f1(t) * f3(t)
– Associative: f1(t) * [f2(t) * f3(t)] = [f1(t) * f2(t)] * f3(t)
– Shift: If f1(t) * f2(t) = c(t),
– then f1(t) * f2(t - T) = f1(t - T) * f2(t) = c(t - T).
– Convolution with impulse, f(t) * 𝛿(t) = f(t)
– Convolution with shifted impulse, f(t) * 𝛿(t - T) = f(t - T)
Task
Plot the autocorrelation sequence of a sinewave with frequency 1 Hz,
sampling frequency of 200 Hz using the Matlab Xcorr built in function and this
subroutine:
 Case Study
Time Delay Estimation Processing Radar Returned Signal
% Assuming there are N samples of x
Y=xcorr(r,x);
R=Y(1:N);
Rrx=fliplr(R);
 Thanks

Questions