testpnn.asv

基于GMM的概率神经网络PNN具有良好的泛化能力

%程序是正确的,但泛化能力差
clc;
clear;
t1=cputime;
num=10;
disp('正在计算数字识别PNN模型...')
for i=1:num
	fname = sprintf('..\\digit\\%da.wav',i-1);
	x = wavread(fname);
	[x1 x2] = vad(x);
	m = mfcc(x);
	m = m(x1-2:x2-2,:);
    l=length(m);
	ref(i).mfcc = m';
    ref(i).tc=i*ones(1,l);
end

p=ref(1).mfcc;
tc=ref(1).tc;
for i=1:num-1
    p=[p,ref(i+1).mfcc];
    tc=[tc,ref(i+1).tc];
end
t=ind2vec(tc');

spread=1;
net=newpnn(p,t,spread);


disp('正在计算测试数字识别的结果...')
s=0;
for i=1:num
	fname = sprintf('..\\digit\\%db.wav',i-1);
	x = wavread(fname);
	[x1 x2] = vad(x);
	m = mfcc(x);
	m = m(x1-2:x2-2,:);
	y=sim(net, m');
    test(i).yc=vec2ind(y);
     test(i).num=0;
     for j=1:length(test(i).yc)
          if  test(i).yc(j)==i
               test(i).num= test(i).num+1;
         end
     end
    t=test(i).num/length(test(i).yc)
     s=s+t;
%     figure 
%     hist(test(i).yc);
end
s=s/10

t2=cputime-t1
% close all;
% disp('正在进行模板匹配...')
% dist = zeros(10,10);
% for i=1:10
% for j=1:10
% 	dist(i,j) = dtw(test(i).mfcc, ref(j).mfcc);
% end
% end
% 
% disp('正在计算匹配结果...')
% for i=1:10
% 	[d,j] = min(dist(i,:));
% 	fprintf('测试模板 %d 的识别结果为：%d\n', i, j);
% end