Adaptive Filters and Applications - Digital Signal Processing: Fundamentals and Applications - page 467

0.1

0.05

0

0 500 1000 1500 2000 2500 3000 3500 4000

0.15

0.1

0.05

0

0 500 1000 1500 2000 2500 3000 3500 4000

0.15

0.1

0.05

0

0 500 1000 1500 2000 2500 3000 3500 4000

Frequency (Hz)

FIGURE 10.10B
Spectrum for original speech, corrupted speech, and clean speech.

t ¼ t/fs; % Convert indices to time instant

x ¼ randn(1,length(wen)); % Generate random noise

n ¼ filter([ 0 0 0 0 0 0.5 ],1,x); % Generate the corruption noise

d ¼ wen þ n; % Generate signal plus noise

mu ¼ 0.01; % Initialize step size

w ¼ zeros(1,21); % Initialize adaptive filter coefficients

y ¼ zeros(1,length(t)); % Initialize the adaptive filter output array

e y; % Initialize the output array

% Adaptive filtering using LMS algorithm
for m ¼ 22:1:length(t)-1
sum ¼ 0;
for i ¼ 1:1:21
sum ¼ sum þ w(i)*x(m-i);
end
y(m) ¼ sum;
e(m) ¼ d(m)-y(m);
for i ¼ 1:1:21
w(i) ¼ w(i) þ 2*mu*e(m)*x(m-i);
end
end
% Calculate the single-sided amplitude spectrum for the original signal
WEN ¼ 2*abs(fft(wen))/length(wen);WEN(1) ¼ WEN(1)/2;

¼