Third Problem Assignment: EE603 - DSP and Its Applications
Third Problem Assignment: EE603 - DSP and Its Applications
Assigned on: August 24, 2018 Due on: August 31, 2018
Notes:
(1) Copying will be dealt with strictly. Institute disciplinary procedures will be invoked if
any form of cheating is detected.
(2) The submissions must include the comments, plots AND the code in a SINGLE PDF.
Without the code, the submission will not be evaluated. If you submit a zip file containing
the code, plots etc., this PDF must STILL be included in the zip file.
(3) The computer assignments should be solved using GNU Octave or any other free/open
source software kit approved by the instructor. Solutions that work only on Matlab will
not be accepted.
PRᴏBᴌᴇᴍ 1
(10 points)
Let ℎ[n] and H(z) denote the impulse response and system function of stable all-pass LTI
system. Let ℎi [n] denote the impulse response of the (stable) LTI inverse system. Assume that
ℎ[n] is real. Show that ℎi [n] = ℎ[−n].
PRᴏBᴌᴇᴍ 2
(10 points) System S1 has real impulse response ℎ1 [n] and real-valued frequency response
H1 (ejω ).
(a) Does the impulse response ℎ1 [n] have any symmetry? Explain.
(b) System S2 is a linear phase system with the same magnitude reponse as system S1 . What
is the relationship between ℎ2 [n], the impulse response of the system S2 , and ℎ1 [n]?
(c) Can a casual IIR filter have linear phase? Explain. If your answer is yes, provide an
example. If not, prove that it is not possible for an IIR filter to have linear phase.
PRᴏBᴌᴇᴍ 3
(10 points) An LTI system has generalized linear phase and system function H(z) = a+bz−1 +
cz−2 . The impulse response has unit energy, a ≥ 0, and H(ejπ ) = H(ej0 ) = 0.
(a) Determine the impulse response ℎ[n].
(b) Plot |H(ejω )|.
PRᴏBᴌᴇᴍ 4
(10 points) A causal LTI has system has the system function
(1 − 1.5z−1 − z−2 )(1 + 0.9z−1 )
H(z) =
(1 − z−1 )(1 + 0.7jz−1 )(1 − 0.7jz−1 )
(a) Write down the difference equation that is satisfied by the input and output of the system.
(b) Plot the pole-zero diagram and indicate the ROC for the system function.
(c) Sketch |H(ejω )|.
(d) State whether the following are true or false about the system:
(i) The system is stable.
(ii) The impulse response approaches a constant for large n.
(iii) The magnitude of the frequency response has a peak at approximately ω = ±π.
(iv) The system has a stable and causal inverse.
PRᴏBᴌᴇᴍ 5
(10 points) Consider a discrete-time LTI filter whose impulse response ℎ[n] is nonzero only
over five consecutive time samples; the filter’s frequency response is H(ejω ). Let signals x[n]
and y[n] denote the filter’s input and output, respectively.
Moreover, you are given the following information about the filter.
(a)
π
(b) There exists a signal a[n] that has a real and even DTFT A(ejω ) given by
Statement: If the group delay of the filter is constant for 0 < ω < π, then the impulse response
must have the property that either
ℎ[n] = ℎ[M − n]
or
where M is an integer.
If the statement is true, show why it is true. If it is false, provide a counter example.
PRᴏBᴌᴇᴍ 7
(30 points) For this problem, you will use the rhino.wav file provided with the homework. You
may need to add the command pkg load signal on the top of your script for some commands
to work.
(a) Load the wave file into Octave. Plot the samples. What is the sampling frequency? Play
the sound using the soundsc command.
(b) Plot the frequency content of the waveform versus the frequency (in Hertz) for any one
channel and comment on the frequency content.
(c) Add Gaussian noise of variance 0.01 to each channel. You can generate Gaussian noise
with variance 0.01 by using 0.01 * randn(N, 2), where N is the sequence length.
(d) Use the soundsc function to play the noisy waveform. What do you observe?
(e) Now, let’s try filtering the noisy signal with a Butterworth filter. You can create a But-
terworth filter using the butter function. Create a sixth order Butterworth low-pass filter
with cut-off corresponding to 1 kHz.
(f) Filter the noisy waveform with the Butterworth filter using the filter command. Play
the sound with soundsc. What do you observe? Why?