# State University of New York at Buffalo

CSE 473/573 Fall 2011 Homework Set #1

Date: Monday September 12, 2011; Due: Monday September 26, 2011 at the start of Class

Name: ____________________________ Student Number: ______________________

Problem (1) (Convolution) 20%
Let the input signal be x|n]and let the system impulse response be b|n], which are defined as
follows:
x|n] = o|n] +2o|n - 1] - o|n - S]
and
b|n] = 2o|n + 1] + 2o|n - 1].

in which the unit impulse function is defined as:
o|n] = ]
u,
1,

n = u
n = u

Compute and hand plot each of the following convolutions:
(a) y
1
|n] = x|n] ∗ b|n]
(b) y
2
|n] = x|n + 2] ∗ b|n]
(c) y
3
|n] = x|n] ∗ b|n + 2]

Problem (2) (Fourier Series) 20%
Consider the following three continuous-time signals with a fundamental period of I =
1
2
, :
x(t) = cos(4nt),
y(t) = sin(4nt),
z(t) = x(t)y(t)

(a) Determine the Fourier Series coefficients of x(t).
(b) Determine the Fourier Series coefficients of y(t).
(c) Use results of parts (a) and (b), along with the multiplication property of the Fourier
Series, to determine the Fourier Series coefficients of z(t) = x(t)y(t).
(d) Determine the Fourier Series coefficients of z(t) through direct expansion of z(t) in
trigonometric form, and compare your result with that of part (c).

Problem (3) (MATLAB) 60%
In this problem, we will consider a technique known as block convolution, which is often
used in real-time implementation of digital filters for speech processing and image
processing in which short processing delay may be desired. This technique is particularly
useful when processing long input sequence with a relatively short duration filter.

The input sequence is broken into short blocks, each of which can be independently
processed with relatively little delay. The linearity of convolution guarantees that the
superposition of the outputs from all individual blocks will be equal to the convolution of the
entire sequence with the impulse response of the filter. For this problem, you will perform
each of the small convolutions using the MATLAB function conv.

To illustrate the procedure, assume that you have a filter with a finite-length impulse
response b|n] which is nonzero only on the interval u ¸ n ¸ P - 1. Also assume that the
input sequence x|n] is u for n < u and that the length of x|n] is significantly greater than P.
You can break the signal x|n] into segments of length I,
x|n] = x
¡
|n - rI],
«
¡=0

where I > P, and
x
¡
|n] = ]
x|n + rI],
u
u ¸ n ¸ I - 1,
otheiwise,

For this problem, we define the unit step function as:
u|n] = ]
u,
1,

n < u
n ¸ u

(a) For b|n] = (u.9)
n
(u|n] -u|n - 1u]) and x|n] = cos(n
2
) sin(2nn S ⁄ ), compute
y|n] = b|n] ∗ x|n] for u ¸ n ¸ 99 directly using conv. Make a plot of y|n] over this
range using stem.
(b) For I = Su, break the sequence x|n] into two sequences, each of length 50. Compute
y
0
|n] = b|n] ∗ x
0
|n], and y
1
|n] = b|n] ∗ x
1
|n], where x
0
|n] contains the first 50
samples of x|n] and x
1
|n] contains the second 50 samples of x|n]. The form of the
output y|n] is given by:
y|n] = b|n] ∗ x|n] = y
0
|n] + y
1
|n - k]
Determine the appropriate value of k to use, and note that y
0
|n] and y
1
|n] will both be of
length I + P - 1. When y
0
|n] and y
1
|n] are added together, there will generally be a
region where both are nonzero. It is for this reason that this method of block convolution
is called the overlap-add method. Compute y|n] in this manner, and plot y|n] over the
range u ¸ n ¸ 99. Is your result the same as what you found in Part (a)?

break the sequence into two sequences. and ∗ . each of length 50. Compute in this manner. where contains the first 50 samples of and contains the second 50 samples of . 0 (a) For 0. we define the unit step function as: 0. The input sequence is broken into short blocks. 0 1. For this problem. .Problem (3) (MATLAB) 60% In this problem. we will consider a technique known as block convolution. each of which can be independently processed with relatively little delay. which is often used in real-time implementation of digital filters for speech processing and image processing in which short processing delay may be desired. you will perform each of the small convolutions using the MATLAB function conv. The form of the output is given by: ∗ and will both be of Determine the appropriate value of to use. Make a plot of over this range using stem. assume that you have a filter with a finite-length impulse response which is nonzero only on the interval 0 1. When and are added together. To illustrate the procedure. and 0 1. You can break the signal into segments of length . 0 otherwise. The linearity of convolution guarantees that the superposition of the outputs from all individual blocks will be equal to the convolution of the entire sequence with the impulse response of the filter. For this problem. This technique is particularly useful when processing long input sequence with a relatively short duration filter. Compute ∗ . (b) For 50. Is your result the same as what you found in Part (a)? .9 10 and cos sin 2 ⁄5 . . where . there will generally be a region where both are nonzero. It is for this reason that this method of block convolution is called the overlap-add method. Also assume that the input sequence is 0 for 0 and that the length of is significantly greater than . and plot over the range 0 99. compute ∗ for 0 99 directly using conv. and note that length 1.