Digital Signal Processing Reference
In-Depth Information
Compare this result with the theoretical result, which is 1 exactly.
Find the mean and variance of one realization and compare with the theoretical
values. As in Task 1, consider different SNRs.
Task 3
Find the autocorrelation function of the signals x(t), n(t), and y(t). Find the cross-
correlation R
xn
(s) between the signal x(t) and the noise process n(t).
Experiment # 4: Filter Design with Application to Noise Reduction
Introduction
MATLAB provides built-in codes for designing analog and digital filters. In this
experiment we will consider Butterworth and Chebychev-I filters. You can type on
the command line [
help butter
and [
help cheby1
to know the design
parameters of these filters. As an application of filtering, we will design a filter for
the purpose of reducing noise that corrupts a narrowband signal of known
frequency (if the signal frequency is unknown, then this approach fails and we
need an adaptive filter for noise reduction).
White Gaussian noise n(t) is a broadband signal since it is uncorrelated (i.e. its
autocorrelation function is a weighted delta function, hence its power spectral
density is constant for all frequencies as shown in Fig.
D.5
.
Normally, we are interested in a specific frequency band (-B \ f \ B) for
practical applications. For example, in speech signal processing, the important
frequency band is about 4 kHz, and the whole audible spectrum is less than
20 kHz. Since noise power is given by
p
n
¼
Z
1
G
n
ð
f
Þ
df
;
1
x
(
t
) =
δ
(
t
)
X
(
f
) = 1
FT
1
t
, sec
f
, Hz
0
0
Fig. D.5
Spectrum of a time delta function
Search WWH ::
Custom Search