ECE569 16S hw5
ECE569 16S hw5
Homework Assignment #5
Due: 12 April 2016
x̂(n)
- +l
?
- H1 (z) - ↓2 - ↑2 - F1 (z) -
For this structure, design FIR analysis and synthesis filters of length N = 8, following
the design procedure discussed in class, so as to achieve perfect reconstruction. The
details of this procedure are provided in the document “A Procedure for Designing
Two-channel PR Filter Banks,” which is available at the course web site. This design
methodology begins by applying the Parks-McClellan algorithm to obtain an appro-
priate length half-band filter. In the design of this filter, use passband and stopband
edges of ωp = π/4 and ωs = 3π/4. Report the following for your results.
(a) The coefficients of the filters H0 (z), H1 (z), F0 (z), and F1 (z).
(b) A plot of the magnitude response of H0 (z) and H1 (z), demonstrating the splitting
of the input signal into low and high frequency bands.
(c) The transfer functions
associated with the regular and alias components of the filter bank response.
2. An ARMA process has an autocorrelation function rxx (m) whose z-transform is given
by
4 − 17z −1 + 4z −2
Pxx (z) =
1 − 52 z −1 + z −2
1
with a region of convergence of 2
< |z| < 2.
(a) Determine the filter H(z) for generating {x(n)} from a white noise sequence. Is
such an H(z) unique?
(b) Determine a stable linear whitening filter for the sequence {x(n)}.
4. The first several values of the autocorrelation sequence for a random signal {x(n)} are
given by
Using the Yule-Walker equations, find AR(p) models for {x(n)} for p = 1, 2, . . . , 7.
What do you think is the best choice for p? Compute rxx (`) for ` = 0, 1, . . . , 7 for this
model and compare them to those for {x(k)}. Is an AR model appropriate for this
{x(n)}?