用MATLAB来实现3个通道的语音盲源分离 .rar_matlab通道分离_信号分离_盲分离_盲处理_语音盲分离

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Generate simulate data % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % data length N = 44000; % read data from wave file s1 = wavread('s0002.wav'); s1 = s1(1:N)'; s2 = wavread('S0033.wav'); s2 = s2(1:N)'; s3 = wavread('S0024.wav'); s3 = s3(1:N)'; s = [s1; s2; s3]; % mix the data A = [ 1 4 5 3.4 2.4 2 6 2 3.5]; x = A * s; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Now we'll recover s1, s2 & s3 from x % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% len = size(x, 2); % number of data samples IC_num = size(x, 1); % number of ICs %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Center the data with 0 mean % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% for (cnt = 1: IC_num) x(cnt, :) = x(cnt, :) - mean(x(cnt, :)')'; end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Whiten the data % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Rx = x*x'/len; [Ux Sx] = eig(Rx); % assert that Ux*Sx*Ux'-Rx = 0 Vx = inv(Sx).^0.5*Ux'; % Vx is the whiten matrix z = Vx * x; % z is whiten data % assert that z*z'/len = I %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Recover ICs % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % What kind of G we will be used ? % G(y) = (1/alpha) * ln(cosh(alpha*y)) % g(y) = tanh(alpha*y) % or % G(y) = -exp(-y^2/2) % g(y) = y * exp(-y^2/2) % g'(y) = (1-y^2)*exp(-y^2/2) w = eye(IC_num); w_old = w; eps = 1.0e-5; flag = 1; while (flag == 1) for (cnt = 1: IC_num) y = w(:, cnt)'*z; Gy = -exp(-y.^2/2); gy = y.*(-Gy); gy1 = (1-y.^2).*(-Gy); E1 = z*gy'/len; E2 = mean(gy1')'; w(:, cnt) = E1-E2*w(:, cnt); end wt = w'; [Uw Sw] = svd(wt*wt'); wt = inv(Uw * sqrt(Sw)) * wt; w = wt'; change = ((sum(sum((abs(w'*w_old) - eye(IC_num)).^2)))/IC_num) if ( change < eps) flag = 0; end w_old = w; end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Recover s % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% s_n = w'*z; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Play all the music % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Play mix sound 1 Fs=16000; x(1,:)=x(1,:)/(max(s_n(1,:))-min(x(1,:)))*2; wavwrite(x(1, :),Fs,'mixsound1') %wavplay(x(1, :), Fs); % Play mix sound 2 x(2,:)=x(2,:)/(max(x(2,:))-min(x(2,:)))*2; wavwrite(x(2, :),Fs,'mixsound2') %wavplay(x(2, :), Fs); % Play mix sound 3 x(3,:)=x(3,:)/(max(x(3,:))-min(x(3,:)))*2; wavwrite(x(3, :),Fs,'mixsound3') %wavplay(x(3, :), Fs); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Play sound 1 s_n(1,:)=s_n(1,:)-mean(s_n(1,:)); s_n(1,:)=s_n(1,:)/(max(s_n(1,:))-min(s_n(1,:)))*2; wavwrite(s_n(1, :),Fs,'sound1') wavplay(s_n(1, :), Fs); % Play sound 2 s_n(2,:)=s_n(2,:)-mean(s_n(2,:)); s_n(2,:)=s_n(2,:)/(max(s_n(2,:))-min(s_n(2,:)))*2; wavwrite(s_n(2, :),Fs,'sound2') wavplay(s_n(2, :),Fs); % Play sound 3 s_n(3,:)=s_n(3,:)-mean(s_n(3,:)); s_n(3,:)=s_n(3,:)/(max(s_n(3,:))-min(s_n(3,:)))*2; wavwrite(s_n(3, :),Fs,'sound3') wavplay(s_n(3, :), Fs); figure; subplot(3,3,1); plot(s1); title(strcat('source1')); subplot(3,3,4); plot(s2); title(strcat('source2')); subplot(3,3,7); plot(s3); title(strcat('source3')); subplot(3,3,2); plot(x(1,:)); title(strcat('Mixsound1')); subplot(3,3,5); plot(x(2,:)); title(strcat('Mixsound2')); subplot(3,3,8); plot(x(3,:)); title(strcat('Mixsound3')); subplot(3,3,3); plot(-0.2.*s_n(3,:)); title(strcat('seperation1')); subplot(3,3,6); plot(-0.1.*s_n(1,:)); title(strcat('seperation2')); subplot(3,3,9); plot(-0.2.*s_n(2,:)); title(strcat('seperation3'));