| << Chapter < Page | Chapter >> Page > |
Churchill, Winston. "A History of the English Speaking Peoples." Skyhorse Publishing, 2011. Audiobook.
Collins, Suzanne, and McCormick, Carolyn. "Mockingjay (The Hunger Games, Book 3)." Scholastic Press, 2010. Audiobook.
Hutson, Michael. "Acoustic Echo Cancellation Using Digital Signal Processing." Diss. University of Queensland, 2003. Abstract. Web.
Sayed, Ali H. Fundamentals of Adaptive Filtering . Hoboken, NJ: J. Wiley&Sons, 2003. Print.
Stan, Embrechts Guy-Bart, and Archambeau Dominique Jean-Jacque. "Comparison of Different Impulse Response Measurement Techniques." Diss. University of Liege, Belgium, 2002. Abstract. Web.
Blms.m example
%%%Andres Cedeno Dec. 14,2012%%%
%%%Apply BLMS Algorithm to an echoed .wav file%%%[desired,fs] = wavread('No Echo Male.wav'); %read in non-echoed signal[input,fs] = wavread('Echo Male.wav'); %read in echoed signalt = 1/fs:1/fs:length(desired)/fs; %create time vector for plot
mu = 0.0006; % Step sizen = 128; % Block length
hb = adaptfilt.blmsfft(64,mu,1,n); %create filter object[mumax,mumaxmse] = maxstep(hb,desired) %see if filter is stable with choice of step size[y,e] = filter(hb,input,desired); %filter the input signal with adaptive filterout = [y(n:end); y(1:n-1)]; %account for shift caused by block convolutionerror = desired-out; %error is desired minus output
% wavwrite(y,fs,32,'DeechoedMale')%Plot each signal.
subplot(4,1,1); plot(input);title('Input Signal with Echo');
ylabel('Signal Value'); grid on;subplot(4,1,2); plot(desired);
title('Desired Output Signal');ylabel('Signal Value'); grid on;
subplot(4,1,3); plot(out);title('Filter Output Signal');
ylabel('Signal Value'); grid on;subplot(4,1,4); plot(error);
title('Error');ylabel('Signal Value'); grid on;
%calculate root mean square of error signalrms = sqrt(mean(error.^2))
%Other filters for stem plot comparison% hl = adaptfilt.lms(64,mu);
% hn = adaptfilt.nlms(64,mu);% [mumax,mumaxmse] = maxstep(hl,desired)% [mumax,mumaxmse] = maxstep(hn,desired)% [y,e] = filter(hn,input,desired);% [y,e] = filter(hl,input,desired);% figure(2)
% stem(hl.coefficients,'g')% hold on
% stem(hb.coefficients,'r')% stem(hn.coefficients)
Lms.m example
clear all;
order = 1000; % Filter sizestep = 0.007; % Step size
readfile = 'Wiessmale.wav';iterations = length(readfile);
Fs = 44100; % Sampling ratesinput = wavread(readfile);
sdesired = wavread('Hanmale.wav');current = zeros(order,1);
in = zeros(order, 1); % Initialize input vectorfor i=1:iterations
in(1) = sinput(i);out(i) = dot(current, in); % Adaptive filter output
error = sdesired(i)-out(i); % Estimate errorcurrent = current + 2*step*error*in; % Update filter coefficients
for n=order:-1:2 % Iterate through input vectorin(n)=in(n-1);
endserror(i)=error;
cost(i)=error*error;end
subplot(3,1,1); plot(input_signal);title('Input Signal with Echo');
ylabel('Signal Value'); grid on;subplot(3,1,2); plot(desired_signal);
title('Desired Output Signal');ylabel('Signal Value'); grid on;
subplot(3,1,3); plot(filter_output);title('Filter Output Signal');
ylabel('Signal Value'); grid on;figure
subplot(2,1,1); plot(impulse);title('Impulse Response');
ylabel('Signal Value'); grid on;subplot(2,1,2); plot(error_signal);
title('Error');ylabel('Signal Value'); grid on;
Nlms.m example
clear all;
order = 3000; % Filter sizestep = 0.007; % Step size
readfile = 'Weissimp.wav';iterations = length(readfile);
Fs = 44100; % Sampling ratesinput = wavread(readfile);
sdesired = wavread('Hanimp.wav');current = zeros(order,1);
in = zeros(order, 1); % Initialize input vectorfor i=1:iterations
in(1)=sinput(i);out(i)=dot(current, in); % Adaptive filter output
error = sdesired(i)-out(i); % Estimate errorstep = 1/(dot(in, in));
current = current + step*error*in; % Update filter coefficientsfor n=order:-1:2 % Iterate through input vector
in(n)=in(n-1);end
serror(i)=error;cost(i)=error*error;
endsubplot(3,1,1); plot(sinput);
title('Input Signal with Echo');ylabel('Signal Value'); grid on;
subplot(3,1,2); plot(sdesired);title('Desired Output Signal');
ylabel('Signal Value'); grid on;subplot(3,1,3); plot(out);
title('Filter Output Signal');ylabel('Signal Value'); grid on;
figuresubplot(2,1,1); plot(impulse);
title('Impulse Response');ylabel('Signal Value'); grid on;
subplot(2,1,2); plot(serror);title('Error');
ylabel('Signal Value'); grid on;
Code for deconvolution
%%%%%%%%%%%%%%%%%%
% Used to reconstruct the original signal x(t) given an impulse% response of a room h(t) and the echoed audio recording y(t). Does so% by deconvolving y(t) by h(t). Also attempts to remove the frequency
% content of any room/equipment noise%
% @param ht -- impulse response of the room (column vector)% @param yt -- noisy echoed signal (column vector)
% @param nt -- the background noise introduced by recording% equipment and room (column vector)
% @return xt -- the original signal, reconstructed (column vector)%
% @author Andrew Martin%%%%%%%%%%%%%%%%%%
function [ xt ]= find_xt(ht,yt,nt)
%zero pad the shorter two signals to the length of the longestmax_length = max(length(ht),max(length(yt),length(nt)));
ht = [ht ; zeros(max_length - length(ht),1)];
yt = [yt ; zeros(max_length - length(yt),1)];
nt = [nt ; zeros(max_length - length(nt),1)];%fft to frequency domain
Yf = fft(yt);Hf = fft(ht);
Nf = fft(nt);%subtract noise freq content out and divide by impulse response%added epsilon to Hf to avoid divisions by 0
Xf = (Yf-Nf)./(Hf+eps);%ifft to the time domainxt = ifft(Xf);
end
Notification Switch
Would you like to follow the 'Characterization and application of echo cancellation methods' conversation and receive update notifications?
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|