Scribd
Upload
34 views6 pages

Assignment Report

This document summarizes research on using a Wiener filter for speech enhancement by reducing noise from ocean waves. It includes spectrograms of original speech, noise, and noisy speech at different signal-to-noise ratios. The results show that the Wiener filter is able to remove noise at different time frames and SNR levels. It also provides MATLAB code implementing the Wiener filter for speech enhancement.

Uploaded by

Bhaskar Chakradhar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
34 views6 pages

Assignment Report

This document summarizes research on using a Wiener filter for speech enhancement by reducing noise from ocean waves. It includes spectrograms of original speech, noise, and noisy speech at different signal-to-noise ratios. The results show that the Wiener filter is able to remove noise at different time frames and SNR levels. It also provides MATLAB code implementing the Wiener filter for speech enhancement.

Uploaded by

Bhaskar Chakradhar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

SPEECH ENHANCEMENT USING WIENER FILTER

Bhaskar C(12EC28), Rachith P(12EC76)

Original Speech Spectrogram

Noise Spectrogram: Ocean Waves

Noisy Speech (SNR = -5dB)

De-noised(-5dB) Speech Spectrogram

Noisy Speech (SNR = 0dB)

De-noised(0dB) Speech Spectrogram

Noisy Speech (SNR = 10dB)

De-noised(10dB) Speech Spectrogram

Short Term Wiener Filter Estimate

Results:
1. As seen in the de-noised(-5dB) spectrogram, wiener filter removes noise in
regions other than the time around 1.5 sec.
2. As seen in the de-noised(0dB) spectrogram, wiener filter removes noise in initial
time frames around 1.5sec but not during initial time frames.
3. As seen in the de-noised(10dB) spectrogram, wiener filter removes noise in both
the time frames( around 1.5 sec and initial frames).
4. As seen in short term Wiener filter estimate plot, the filter removes noise
significantly in regions other than 1.2-1.8sec time frame bins.

SNR Comparison(in dB):


Input SNR
-5
0
10

Output SNR
10.12
4.56
19.56

CODE:
[x, fs] = audioread(['n-5.wav']);
%[x, fs] = audioread(['n0.wav']);
%[x, fs] = audioread(['n10.wav']);
len = floor(20*fs/1000); % To find frame size
if rem(len,2)==1, len=len+1;
end;

len1 = floor(len*0.5); % 50% overlap


len2=len-len1;
window

= hanning(len);

m = len;
k = 1;
img = sqrt(-1);
x_old = zeros(len1,1);
frames = floor(length(x)/len1)-1;
mean_n = zeros(m,1); %mean of noise
xfinal = zeros(frames*len2,1);
j = 1;
for k = 1:25
mean_n = mean_n+abs(fft(window.*x(j:j+len-1),m)).^2;
j = j+len;
end
noise_mu2 = mean_n/25;
k=1;
alpha = 0.85;
for n = 1:frames-1
insign = window.*x(k:k+len-1);
spec = fft(insign,m);
s
= abs(spec);
s2 = s.^2;
theta = angle(spec);
PSD_1=s2-noise_mu2;
PSD(:,n) = PSD_1;
if(n>1)
PSD_1 = alpha * PSD(:,n-1) + (1-alpha) * PSD(:,n);
end
for i=1:length(PSD_1)
if PSD_1(i)<0
PSD_1(i)=0;
end
end
H_temp = PSD_1./(PSD_1+noise_mu2);
H(:,n) = H_temp;
new=H_temp.*s;
new=new.*exp(1i*theta);
xi_w = ifft(new);
xfinal(k:k+len1-1) = x_old(1:len1) + xi_w(1:len1);
x_old = xi_w(len1+1:len);
k = k + len1;
end
%sound(real(xfinal),fs)
%wavwrite(xfinal,fs,16,['Denoised.wav']);
return;

You might also like