ICA资料分享_盲源分离_ICA_语音信号处理_语音信号叠加_

% ica_test.m: % This file is used for testing the % ICA(Independent Component Analysis) algorithm; % FastICA(Fast Independent Component Analysis) algorithm; % Whitening algorithm. % % Copyright (c) 2018 CHEN Tianyang % more info contact: tychen@whu.edu.cn clear;close; %% ICA test % -------------------在读入信号是纯净的非混叠信号时----- [s1,fs] = audioread('voice\sound1.wav'); [s2,~] = audioread('voice\sound2.wav'); s1 = s1(:)'; s2 = s2(:)'; s1 = s1/max(s1); s2 = s2/max(s2); L = min(length(s1),length(s2)); s1 = s1(1:L); s2 = s2(1:L); % 语音混叠 H = rand(5,2)*[s1;s2]; mx = mean(H,2); % c = cov(H'); H = H - mx*ones(1,L); % subtract means from mixes 字段减去均值，中心化 % wz=2*inv(sqrtm(c)); % get decorrelating matrix % X = wz*H; % decorrelate mixes so cov(X')=4*eye(N) X=H; % -------------------在直接读入已混叠的信号时------- % [s1,fs] = audioread('c1_mixed1.wav'); % [s2,~] = audioread('c1_mixed2.wav'); % s1 = s1/max(s1); % s2 = s2/max(s2); % X=[s1';s2']; % X=X-repmat(mean(X,2),1,size(X,2)); % figure;plot(s1); % figure;plot(s2); % -------------------over-------------------------- % 信号时序扰乱 % [N,P]=size(X); % P=48000, N=2, for example % permute=randperm(P); % generate a permutation vector % X_mixed=X(:,permute); % time-scrambled inputs for stationarity 时序扰乱 %------------------------------- tic; S = myICA(X(1:2,:)); % S = myICA2(X,2); % S_mixed = myICA(X_mixed); toc % 如输入信号时序扰乱，则输出需将序列理顺 % S_unmixed = zeros(N,P); % for i = 1:P % sample = permute(1,i); % S_unmixed(:,sample) = S_mixed(:,i); % end %------------------------------- sound(X(1,:),fs); sound(X(2,:),fs); sound(S(1,:),fs); sound(S(2,:),fs); sound(S_unmixed(1,:),fs); sound(S_unmixed(2,:),fs); sound(sum(S),fs); %% FastICA test close;clear; % [s1,fs] = audioread('c1_mixed1.wav'); % s2 = audioread('c1_mixed2.wav'); [s1,fs] = audioread('voice\sound1.wav'); s2 = audioread('voice\sound2.wav'); s3 = audioread('voice\sound3.wav'); s4 = audioread('voice\sound4.wav'); s1 = s1/max(s1); s2 = s2/max(s2); s3 = s3/max(s3); s4 = s4/max(s4); figure;plot(s1); figure;plot(s2); figure;plot(s3); figure;plot(s4); S=[s1,s2,s3,s4]'; %信号组成4*N [N,P] = size(S); t = 1/fs:1/fs:P/fs; A=rand(4); % A=[0.8 0.9;0.9 0.8]; X=A*S; %观察信号 %源信号波形图 figure(1); subplot(4,1,1);plot(t,s1);axis([0 P/fs -1 1]);title('源信号'); subplot(4,1,2);plot(t,s2);axis([0 P/fs -1 1]); subplot(4,1,3);plot(t,s3);axis([0 P/fs -1 1]); subplot(4,1,4);plot(t,s4);axis([0 P/fs -1 1]); xlabel('Time/s'); %观察信号(混合信号)波形图 figure(2); subplot(4,1,1);plot(X(1,:));title('观察信号(混合信号)'); subplot(4,1,2);plot(X(2,:)); subplot(4,1,3);plot(X(3,:)); subplot(4,1,4);plot(X(4,:)); tic; % Z=FastICA(S); Z = lihao_myfastica(S); toc sound(S(1,:)); sound(S(2,:)); sound(S(3,:)); sound(S(4,:)); sound(X(1,:)); sound(X(2,:)); sound(X(3,:)); sound(X(4,:)); sound(Z(1,:)); % 旅行 sound(Z(2,:)); % 考生 sound(Z(3,:)); % 马克思 sound(Z(4,:)); % 张爱玲 %% white test mu=[0,2]; %数学期望 sigma=[1 2;2,6]; %协方差矩阵 points=mvnrnd(mu,sigma,1000); %生成50个样本 points = points'; figure;scatter(points(1,:),points(2,:),'r*'); axis equal; Y = myWhite(points,1); figure;scatter(Y.PCAW(1,:),Y.PCAW(2,:),'r*');hold on; scatter(Y.ZCAW(1,:),Y.ZCAW(2,:),'y*'); axis equal; cov(points') cov(Y.PCAW') cov(Y.ZCAW') %%