EC367 Advanced Digital Signal Processing

Assignment-1
National Institute of Technology, Warangal

Instructions:

• Make reasonable assumptions. You should show all steps.

• All parts of the same question should be attempted simultaneously.

Q.1. Consider an FIR adaptive filter with u(n) as the tap input vector represented as

u(n) = α(n)s(n) + ν(n) (1)

where, s(w) = [1, e−jw , · · · e−jwM −1 ]T and ν(n) = [ν(n), ν(n − 1), · · · ν(n − M + 1)]T The complex
amplitude of the sinusoidal vector s(w) is a random variable with zero mean and variance σα2 =
E[|α(n)|2 ]
(A) Determine the correlation matrix of the tap input vector u(n).
(B) Suppose that the desired response d(n) is uncorrelated with u(n). What is the value of the
tap-weight vector of the corresponding Wiener filter?
(C) Suppose that the variance σα2 is zero and the desired response is defined by d(n) = ν(n − k),
where 0 ≤ k ≤ M − 1. What is the new value of the tap-weight vector of the Wiener filter?
(D) Determine the tap-weight vector of the Wiener filter for a desired response defined by d(n) =
α(n)e−juτ , where τ is the prescribed delay.

Q.2. Evaluate the coefficients of the weight vector and minimum mean-square error of a Wiener filter with
the specified correlation matrix of the tap input vector R as
" #
1.1 0.5
R= (2)
0.5 1.1

The cross-correlation vector p between the tap-input vector x(n) and the desired response d(n) is
p = [0.5, 0.25]T

Q.3. Evaluate the coefficients of an adaptive equalizer for two successive adaptation cycles with some
samples of the transmitted sequence defined as x(n) = [0.2, 0.3, 0.4, ...]. The channel is characterized
by impulse response as h(n) = [0.26, 0.93, 0.26]T . Design an adaptive equalizer using LMS algorithm.

Q.4. The statistical characterization of a multiple linear regression model of order four is as follows:
The correlation matrix of the input vector u(n) is
 
1.1 0.5 0.1 −0.1
 0.5 1.1 0.5 0.1 
R= (3)
 
 0.1 0.5 1.1 0.5 


−0.1 0.1 0.5 1.1

The cross-correlation vector between the observable data and the input vector is p = [0.5, −0.4, −0.2, −0.1]T .
The variance of the observable data d(n) is σd2 = 1.
 The variance of the additive white noise is σv2 = 0.1
A Wiener filter of varying length M operates on the input vector u(n) as input and on the observable
data d(n) as the desired response. Compute the mean-square error produced by the Wiener filter for
M =1, 2, 3, 4.

Q.5. Differentiate between the open and closed loop systems referring to relevant examples and diagrams.

Q.6. Demonstrate that the LMS algorithm acts as a low-pass filter with a low cutoff frequency when the
step-size parameter µ is small.

Q.7. Consider the operation of an adaptive line enhancer using the LMS algorithm under a low signal-to-
noise ratio. The correlation matrix of the input vector is defined by R = σI2 , where I is the identity
matrix. Show that the steady-state value of the weight-error correlation matrix K(n) is given by
K(∞) = µ2 Jmin I, where µ is the step-size parameter and Jmin is the minimum mean-square error.

—END—

Page 2