Signals & Systems

MT - 252
Lecture # 10: Discrete-Time LTI System
Impulse Response & Convolution Sum

Dr. Hafiz Zia Ur Rehman

[email protected]

Department of Mechatronics Engineering

Air University, Islamabad
 Discrete-Time LTI Systems: the Convolution Sum

◼ The Representation of Discrete-Time Signals in Terms of Impulses

◼ Any discrete-time signal 𝒙[𝒏] can be represented as a function of shifted unit
impulses 𝜹[𝒏 − 𝒌].
◼ Pictorially, when a signal 𝒙[𝒏] is multiplied with 𝜹[𝒏], the output is a unit
impulse with amplitude 𝒙[0].
 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
◼
◼
Time Invariance
Sampling Property of Impulse om
Chapter 10
◼ ◼ How to check/test?
Any signal 𝒙[𝒏], can always
Imagebe Segmentation
expressed in sum of impulses:

This is called sampling property

of an impulse

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
◼ Sampling Property of Impulse
www.ImageProcessingPlace.c
om
◼ Any signal 𝒙[𝒏], can always be Chapter 10
Image Segmentation
expressed in sum of impulses:

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
◼ Sampling Property of Impulse
www.ImageProcessingPlace.c
om
◼ Any signal 𝒙[𝒏], can always beChapter 10in
expressed sum of impulses:
Image Segmentation

x  n = + x  −2   n + 2 + x  −1   n + 1 + x  0  n  + x 1  n − 1 + x  2  n − 2 +

+
x  n =  x k   n − k 
k =−

◼ The key idea in visualizing how the discrete-time unit impulse can be used
to construct any discrete-time signal as a sequence of individual impulses.
◼ In other words, for any signal 𝒙[𝒏], we can always express it as a sum of
impulses.

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
◼ Sampling Property of Impulse
www.ImageProcessingPlace.c
om
◼ Any signal 𝒙[𝒏], can always beChapter 10in
expressed sum of impulses:
Image Segmentation

+
x  n =  x k   n − k  Scaled Impulses
k =−

◼ This corresponds to the representation of an arbitrary sequence as a linear

combination of shifted unit impulses 𝜹[𝒏 − 𝒌], where the weights in this
linear combination are 𝒙[𝒌].

◼ Example: Consider 𝒙[𝒏] = 𝒖[𝒏] , the unit step.

1 n  0 
u  n =  u  n =  u  k    n − k 
0 n  0 k =0


u  n =    n − k 
k =0

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Unit Impulse Response
◼ Response of DT-LTI system
◼ The impulse response, as it is named, is the response of the system to a unit impulse
input.
◼ Recall the definition of a unit impulse:

x  n =   n System y  n = h  n

Unit Impulse Response to a Unit Impulse

◼ In another way, the “impulse response” of a system, i.e., 𝒉[𝒏], is the output that it
produces in response to an impulse input.

Definition: if and only if 𝒙[𝒏] = 𝜹[𝒏] then 𝒚[𝒏] = 𝒉[𝒏]

7
 Digital Image Processing, 3rd ed.
Gonzalez & Woods
◼ Response of DT-LTI system (Unit Impulse Response)
www.ImageProcessingPlace.c
om
◼ Given the system equation, youChapter
can find10
the
“impulse response” just by
Image
feeding 𝒙[𝒏] = 𝜹[𝒏] into Segmentation
the system.
3

2 4

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
Linear Time Invariance (LTI) om
Chapter 10
Image Segmentation
𝑥1 𝑛 LTI 𝑦1 𝑛

– Time invariance property

𝑥1 𝑛 ± 1 LTI 𝑦1 𝑛 ± 1

– Scaling Property

𝑎𝑥1 𝑛 LTI 𝑎𝑦1 𝑛

– Time invariance and Scaling Property

𝑏𝑥1 𝑛 ± 1 LTI 𝑏𝑦1 𝑛 ± 1

– Time invariance, Scaling, and Additive Property or Linear Time Invariance property

𝑎𝑥1 𝑛 + 𝑏𝑥1 𝑛 ± 1 LTI 𝑎𝑦1 𝑛 + 𝑏𝑦1 𝑛 ± 1

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
Linear Time Invariance (LTI) om
Chapter 10
Image
LTISegmentation
𝛿𝑛 ℎ𝑛

– Time invariance property:

𝛿 𝑛−𝑘 LTI ℎ 𝑛−𝑘

– Scaling Property:
𝑎𝛿 𝑛−1 LTI 𝑎ℎ 𝑛−1

– Time invariance and Scaling Property:

𝑥 1 𝛿 𝑛−1 LTI 𝑥 1 ℎ 𝑛−1

– Time invariance, Scaling, and Additive Property or LTI property:

𝑥 1 𝛿 𝑛−1 +𝑥 2 𝛿 𝑛−2 LTI 𝑥 1 ℎ 𝑛−1 +𝑥 2 ℎ 𝑛−2

𝑦𝑛
𝑥𝑛 +∞ +∞

෍ 𝑥 𝑘 𝛿 𝑛−𝑘 LTI ෍ 𝑥 𝑘 ℎ 𝑛−𝑘

𝑘=−∞ 𝑘=−∞

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Convolution-sum Representation (DT-LTI)
◼ Convolution-sum
◼ Suppose that we know that the impulse response of an LTI system is 𝒉[𝒏] and
want to determine the output 𝒚 𝒏 .
◼ Then, from the superposition property, the response 𝒚[𝒏] of the system to the
input 𝒙[𝒏] can be expressed as

y  n =  x  k  h  n − k 
k =

◼ Let 𝒉𝒌 [𝒏] denote the response of the linear system to the shifted unit impulse
𝜹[𝒏 − 𝒌]. Then the above equation can be written as

y  n  =  x  k  hk  n 
k =

If we know the response 𝒉 𝒏 of a linear system to the set of shifted unit impulses
𝜹 𝒏 − 𝒌 , we can construct the response to an arbitrary input 𝒙 𝒏 . 11
 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
◼ Convolution-sum om
Chapter 10
Image Segmentation

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
◼ Convolution-sum om
Chapter 10
Image Segmentation

𝒚𝟏 [𝒏]

𝒚𝟐 [𝒏]

𝒚𝟑 [𝒏]

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
◼ Convolution-sum om
Chapter 10
Image Segmentation

𝒚𝟏 [𝒏]

LTI System
𝒚𝟐 [𝒏]

𝒚𝟑 [𝒏]

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Convolution-sum Representation (DT-LTI)

◼ Convolution-sum (Summary)
◼ Convolution is the process by which an input interacts with an LTI system to
produce an output.
◼ Let 𝒉𝒌 [𝒏] denote the response of the linear system to the shifted unit impulse
𝜹[𝒏 − 𝒌]. Then, from the superposition property, the response 𝒚[𝒏] of the
system to the input 𝒙[𝒏] can be expressed as

y  n =  x k  h n − k 
k =−

◼ This result is known as the convolution sum or superposition

sum. It is symbolically written as

y  n = x  n  h  n 15
 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
◼ Example-01(Method-01): Consider om an LTI system with impulse response
Chapter 10
𝒉[𝒏] and input 𝒙[𝒏], as illustrated in Figure.
Image Segmentation

◼ Determine 𝒚[𝒏]?
◼ Solution:
y  n = x  n  h  n

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution:- www.ImageProcessingPlace.c
om
Chapter 10
y  n = x  n  h n , Image Segmentation

y  n =  x k  h n − k ,
k =−

y  n =  x  k  h  −k + n 
k =−

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution:- www.ImageProcessingPlace.c
om
Chapter 10
Image Segmentation

0, n0
 
  x  k  h  −k + 0 , n=0
 k =−
 
  x  k  h  −k + 1 , n =1
 k =−
y  n =   
 x  k  h  −k + 2 , y  −, , −1 =  x  k  h ( −, , −1) − k   0
 k
n=2
k =
=−
 
 x  k  h  −k + 2 ,
 k
n=3
=−

0, n3

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution:- www.ImageProcessingPlace.c
om
Chapter 10
Image Segmentation

0, n0
 
  x  k  h  −k + 0 , n=0
 k =−
 
  x  k  h  −k + 1 , n =1
 k =− 
y  n =   y  0 =  x  k  h  0 − k  ,
 x  k  h  −k + 2 ,
 k
n=2 k =

y  0 = x  0 h  0  ( 0.5 )  (1) = 0.5

=−
 
 x  k  h  −k + 2 ,
 k
n=3
=−

0, n3

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution:- www.ImageProcessingPlace.c
om
Chapter 10
Image Segmentation

0, n0
 
  x  k  h  −k + 0 , n=0
 k =−
 
  x  k  h  −k + 1 , n =1
 k =− 
y  n =   y 1 =  x  k  h 1 − k  ,
 x  k  h  −k + 2 ,
 k
n=2 k =

 
=−
y 1 = x  0 h  0 + x 1 h 1
 x  k  h  −k + 2 ,
 k y 1 = ( 0.5 )  (1) + ( 2 )  (1) = 2.5
n=3
=−

0, n3

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution:- www.ImageProcessingPlace.c
om
Chapter 10
Image Segmentation

0, n0
 
  x  k  h  −k + 0 , n=0
 k =− 

  y  2 =  x  k  h 1 − k  ,
  x  k  h  −k + 1 , n =1 k =
 k =−
y  n =   y  2 = x  0 h  0 + x 1 h 1
 x  k  h  −k + 2 , y  2 = ( 0.5 )  (1) + ( 2 )  (1) = 2.5
 k
n=2
=−
 
 x  k  h  −k + 2 ,
 k
n=3
=−

0, n3

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution:- www.ImageProcessingPlace.c
om
Chapter 10
Image Segmentation

0, n0
 
  x  k  h  −k + 0 , n=0 
 k =−
  y 3 =  x  k  h 1 − k  ,
  x  k  h  −k + 1 , n =1 k =
 k =−
y  n =   y 3 = x 1 h 1
 x  k  h  −k + 2 ,
 k
n=2
=− y 3 = ( 2 )  (1) = 2
 
 x  k  h  −k + 2 ,
 k
n=3
=−

0, n3

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution:- www.ImageProcessingPlace.c
om
Chapter 10
Image Segmentation

0, n0
 
  x  k  h  −k + 0 , n=0
 k =−
  
  x  k  h  −k + 1 , n =1 y  4, ,   =  x  k  h 1 − k  = 0
 k =−
y  n =   k =
 x  k  h  −k + 2 ,
 k
n=2
=−
 
 x  k  h  −k + 2 ,
 k
n=3
=−

0, n3

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution:- www.ImageProcessingPlace.c
om
Chapter 10
0, n0
Image Segmentation
 
  x  k  h  −k + 0 , 0, n0
n=0
 k =− 0.5, n=0
  
  x  k  h  −k + 1 , n =1 2.5, n =1
 k =−
y  n =  
y  n = 
2.5, n=2
 x  k  h  −k + 2 ,
 k
n=2 2.0,
=−
n=3
  
 x  k  h  −k + 2 , 0, n3
 k
n=3
=−

0, n3

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution:- Summary of all the steps
www.ImageProcessingPlace.c
om
Chapter 10
Image Segmentation

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
Solution (second Method): Consideroman LTI system with impulse response
Chapter 10
𝒉[𝒏] and input 𝒙[𝒏], as illustrated in Figure.
Image Segmentation


y  n =  x k  h n − k ,
k =−

y  n  = x  0 h  n − 0 +x 1 h  n − 1 ,
y  n  = 0.5h  n − 0 +2h  n − 1

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
Example-02: Find the convolution (𝑦om𝑛 ) of the two sequences 𝑥[𝑛] and
Chapter 10
ℎ[𝑛] represented by,
Image Segmentation

x  n  = 0, 0, 3, 1, 2,0  ;
h  n  = 0, 3, 2, 1
Solution:-

0, n0
9,
y  n = x  n  h n ,  n=0
9, n =1
 
y  n = y  n  = 11,
 x k  h n − k , 5,
n=2
k =− n=3


2, n=4
y  n =  x  k  h  −k + n  0,
k =−  n4

y  n  = 0, 9,9,11,5, 2, 0 

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution (Third Method): www.ImageProcessingPlace.c
om
Chapter 10
k: -2 -1 Image
0 Segmentation
1 2 3 4 5

x(k): 3 1 2

h(0-k): 1 2 3

h(1-k): 1 2 3

h(2-k): 1 2 3

h(3-k): 1 2 3

h(4-k): 1 2 3

h(5-k): 1 2 3

Final Output →→→ y  n  =  0, 9,9,11,5, 2, 0 

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
Example-03: Find the convolution (𝑦om𝑛 ) of the two sequences 𝑥[𝑛] and
Chapter 10
ℎ[𝑛] represented by,
Image Segmentation

x  n  = 0, 0, 1, 2, 4,0  ;
h  n  = 0, 0, 1, 1, 1,1,0  ;
Solution:-

~Try Yourself ~

Example-04: Find the convolution (𝑦 𝑛 ) of the two sequences 𝑥[𝑛] and

ℎ[𝑛] represented by,

x  n  = 0, 2, 1, -2, 3,-4  ;

h  n  = 0,3, 1, 2, 1,4,0  ;

~Try Yourself ~

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
Example-05 (Matrix Method~Forth Method):
om Find the convolution (𝑦 𝑛 ) of the two
Chapter 10
sequences 𝑥[𝑛] and ℎ[𝑛] represented by,
Image Segmentation
x  n  =  1, 2, 3 ;

h  n  =  1, 2, 3 ;

y  n  = 1 4 10 12 9

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
Example-05: Find the convolution (𝑦 𝑛om) of the two sequences 𝑥[𝑛] and ℎ[𝑛]
represented by, Chapter 10
Image Segmentation
x  n  =  1, 2, 3 ; h  n  =  1, 2, 3 ;

x  n

h  n 1 2 3
Length of 𝒚[𝒏]
1 1 2 3
m = 3,
n = 3,

2 2 4 6 L = ( m + n) −1

3 3 6 9

y  n  = 1 4 10 12 9

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
Example-06: Find the convolution (𝑦 𝑛om) of the two sequences 𝑥[𝑛] and ℎ[𝑛]
represented by, Chapter 10
Image Segmentation
x  n =  u  n ,
n
with 0    1.
h  n = u  n;

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
www.ImageProcessingPlace.c
Example-06: Find the convolution (𝑦 𝑛om) of the two sequences 𝑥[𝑛] and ℎ[𝑛]
represented by, Chapter 10
Image Segmentation
x  n =  u  n ,
n
with 0    1.
x k  =  ku k 
h  n = u  n;

Solution:

y  n = x  n  h n ,
 h  −k − n  = u  −k − n 
y  n =  x k  h n − k , n0
k =−

y  n =  x  k  h  −k + n 
k =−
h  −k + n  = u  −k + n 

n0

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution: www.ImageProcessingPlace.c
om
Chapter 10


Image Segmentation
 0
y  n =  x k  h n − k ,   ( u  k  ) ( u  − k − n  ) , n  0 n0
k
 0,
 k =− 
y  n =   y  n =   k
k =−

  , n0

 ( k u  k  ) ( u  − k + n  ) ,
 
y  n =  x  k  h  −k + n . n0  k =0
k =− k =0

x k 

h  −k − n 
n0

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
x k 
www.ImageProcessingPlace.c
om
Chapter 10
Image Segmentation

See that

𝒚[𝒏] = 0 for 𝒏 < 0

How about 𝑛 ≥ 0?

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
x k 
www.ImageProcessingPlace.c
om
Chapter 10
Image Segmentation

𝒏 → +∞

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution: www.ImageProcessingPlace.c
om
Chapter 10


Image Segmentation
 0
y  n =  x k  h n − k ,   ( u  k  ) ( u  − k − n  ) , n  0 n0
k
 0,
 k =− 
y  n =   y  n =   k
k =−

  , n0

 ( k u  k  ) ( u  − k + n  ) ,
 
y  n =  x  k  h  −k + n . n0  k =0
k =− k =0

 k
=  0 +1 + 2 +
x k  k =0

 0, n0

y  n  = 1 −  n +1
 , n0
h  −k + n   1−

n0

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Digital Image Processing, 3rd ed.
Gonzalez & Woods
Solution: www.ImageProcessingPlace.c
om

Chapter 10
 0, n0  0, n0  k
=  0 +1 + 2 +
 Image Segmentation
y  n  = 1 −  n +1
k =0
y  n =   k
  , n0  , n0
 k =0  1−
1 −  n +1
n  1 −  ,  1

 1 −  n +1 
  =  n +1
k

y  n =   − 1 ,
 u  n
k =0
 1
  − 1
 1− 

1

k =0
k =
1−
 1 −  n +1 
y  n =   u  n
 1− 

1
1−

© 1992–2008 R. C. Gonzalez & R. E. Woods

 Convolution-sum Representation (DT-LTI)

~Try Yourself ~

From book

Example 2.4
Example: 2.5
 Acknowledgement
◼ The slides are prepared based on the following textbook:
◼ Chapter #2 (Art. 2.1), signals & systems by Alan V. Oppenheim, S. Hamid
Nawab, 2nd edition, Pearson education - prentice hall
◼ Reading Assignment :2.1.1,2.1.2.
◼ Special thanks to
◼ Prof. Dr. Muhammad Shehzad Hanif
◼ Prof. Dr. Mark Hasegawa-Johnson