bprengong.m

用matlab编程,采用神经网络是别的方法,

clear;
num=15;

disp('正在计算识别模型...')
for i=1:num
   fname = sprintf('%ss%d.wav', 'data\train\', i); 
   [s, fs] = wavread(fname);
    v =mfcc(s, fs); 
    [n,m]=size(v);
    a=reshape(v,1,n*m);
    b=a(1:15);
	ref(i).p = b;
    
end 
p=ref(1).p;
for i=1:num-1
    p=[p;ref(i+1).p];
end
T=[1 0 0 0 0 0 0 0 0 0 0 0 0 0 0;
   0 1 0 0 0 0 0 0 0 0 0 0 0 0 0;
   0 0 1 0 0 0 0 0 0 0 0 0 0 0 0;
   0 0 0 1 0 0 0 0 0 0 0 0 0 0 0;
   0 0 0 0 1 0 0 0 0 0 0 0 0 0 0;
   0 0 0 0 0 1 0 0 0 0 0 0 0 0 0;
   0 0 0 0 0 0 1 0 0 0 0 0 0 0 0;
   0 0 0 0 0 0 0 1 0 0 0 0 0 0 0;
   0 0 0 0 0 0 0 0 1 0 0 0 0 0 0;
   0 0 0 0 0 0 0 0 0 1 0 0 0 0 0;
   0 0 0 0 0 0 0 0 0 0 1 0 0 0 0;
   0 0 0 0 0 0 0 0 0 0 0 1 0 0 0;
   0 0 0 0 0 0 0 0 0 0 0 0 1 0 0;
   0 0 0 0 0 0 0 0 0 0 0 0 0 1 0;
   0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
];

Pr=[min(p(1,:)),max(p(1,:))];
for k=2:num
Pr=[Pr;min(p(k,:)),max(p(k,:))];
end
p=p';
T=T';

%net=newff(Pr,[16,25,15],{'tansig','logsig','purelin'},'traingdm');
%net.trainParam.show = 50;
%net.trainParam.lr = 0.05;
%net.trainParam.epochs = 10000;
%net.trainParam.goal = 1e-5;
%net =train(net,p,T);



%Y=sim(net,p);

%yc=vec2ind(Y);
maxvalue=0;k=0;
disp('正在计算测试识别的结果...')
for i=1:num
   fname = sprintf('%ss%d.wav', 'data\test\', i); 
   [s, fs] = wavread(fname);
    v =mfcc(s, fs); 
    [n,m]=size(v);
    a_test=reshape(v,1,n*m);
    b_test=a_test(1:15);
    test(i).p_test = b_test;
    
end 
p_test=test(1).p_test;
for i=1:num-1
    p_test=[p_test;test(i+1).p_test];
end
p_test=p_test';
%Y_test=sim(net,p_test);

[nb,minb,maxb,na,mina,maxa]=premnmx(p_test,p);

net=newff(Pr,[25,15],{'tansig','logsig'},'traingdm');
net=init(net);
net.trainParam.show = 50;
net.trainParam.lr = 0.2;
net.trainParam.epochs = 10000;
net.trainParam.goal = 1e-5;

 %[n,m]=size(na);
  %na=reshape(na,1,n*m)
net=train(net,nb,na);
nc= sim(net,nb);
c=postmnmx(nc,mina,maxa);

[ntest,mintest,maxtest]=premnmx(p_test);
nnetout= sim(net,ntest);
netout=postmnmx(nnetout,mintest,maxtest);


real=p;


me=mean(real-netout)
st=std(netout-real)
save e:\mybiye\mymatlabsuoshi\rengongshenjingnetwork\wzy.txt c -ASCII;

%net=newff(minmax(p),[10,15],{'tansig','logsig'},'traingdm');
%net.trainParam.show = 50;
%net.trainParam.lr = 0.05;
%net.trainParam.epochs = 100;
%net.trainParam.goal = 0.1;
%net=train(net,p,T);
%y= sim(net,p);

%y_test= sim(net,p_test);


%maxvalue=0;k=0;
%for m=1;num
    %for n=1:num
        %if (Y_test(n,m)>maxvalue)
           % maxvalue=Y_test(n,m);
           % k=n;
           %end
           %end
    %msg = sprintf('测试说话人 %d 和模板中的说话人 %d 匹配', m, k);
    %disp(msg);
   %maxvalue=0;
   % k=0;
   %end