mel_endpoint2.m

语音识别程序。有需要的抓紧啦。用MATLAB实现的。

function out=mel_endpoint2(data) 
data=double(data); 
%data=filter([1 -0.9375], 1, data); 
data=data/max(data); 
 
xe=enframe(data,128,64); 
L=128; 
w1=window(@hamming,L);  
[m,n]=size(xe); 
xw=zeros(m,n); 
for i=1:m 
    xw(i,:)=xe(i,:).*w1'; 
    xf(i,:)=fft(xw(i,:),128); 
end 
y=melfilterbankm(20,128,8000); 
xs(1:m,1:20)=abs(xf(1:m,1:64))*y(1:20,:)'; 
 
%smooth by using a three=point median filter 
xsm=med3(xs); 
%%%%%%%%%%%%%%noise_freq 
noise_freq=sum(xsm(1:5,:))*1/5; 
 
%%%%%%%%%%%%%%%%X 
for i=1:20 
    X(:,i)=xsm(:,i)-noise_freq(i); 
end 
 
%%%%%%%%%%%%%%%%%%sum(x) 
E=sum(X); 
[P,I]=sort(E); 
I=seqreverse(I); 
%%%%%%%%%%%%%%%%% 
MIMSB(:,1)=sum(X(:,I(18:19))')'; 
%med3(MIMSB) 
VAR=sum(MIMSB)/m; 
 
 
%%%%%%%%%%%%%%%%F(m) 
Na=6 
F=sum(X(:,I(1:Na))'); 
%%%%%%%%%%%%%%%T(m) 
xms=sum(abs(xw(1:m,1:n))'); 
xmsm=med3(xms'); 
noise_time=sum(xmsm(1:5))*1/5 
T=xmsm-noise_time; 
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%EH 
 
EHL=sum(X(:,10:20)'); 
theh=2.5*sum(EHL(2:11)')/10; 
EH=med3(EHL); 
for i=1:m 
    if EH(i)>theh 
        mm(i)=1; 
    else 
        mm(i)=0; 
    end 
end 
 
EHH=med3((EH.*mm)')'; 
 
 
 
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%ETF 
c=0.9; 
d=max(F)/max(EHH); 
ETFl=T+c*F'+d*EHH'; 
 
ETF=med3(ETFl); 
 
 
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%ZCR 
%tmp1=zeros(m,128); 
%tmp2=zeros(m,128); 
data2=[data;0]; 
tmp1= enframe(data2(1:end-1), 128, 64); 
tmp2  = enframe(data2(2:end)  , 128, 64); 
signs = (tmp1.*tmp2)<0; 
diffs = (tmp1 -tmp2)>0.02; 
ZCRL  = sum(signs.*diffs, 2); 
 
 
    
ZCR=med3(ZCRL); 
 
 
 
 
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%word boundary detection 
th1=5;th4=15; 
coef1=0.7; 
coef2=0.8*0.5; 
coef3=0.25*0.75; 
coef4=1*0.5; 
thzcr=sum(ZCR(1:10))/10; 
th5=max([4,1.2*thzcr]); 
%th5=4; 
%%%%%%%%%%%%set the throld 
if VARth2(i) 
        m1=i; 
        for j=m1:m 
            if ETF(j)th4 
          
        break 
         
     else 
        i=m2; 
        m1=m2; 
     end 
    end 
end 
%%%%%%%%%find the beginning 
index=m1; 
if m1>15 
for i=index-1:-1:index-th4-1 
    if ETF(i)0 
  %      if ETF(m1-1)ETF(m2+1) 
          m2=m2+1; 
           flag=1; 
           if m2-index>5 
               flag=0; 
           end 
        else 
            flag=0 
            m2=m2-1; 
        end 
    else 
        flag=0; 
    end 
end 
   
%%%%%%%%%%%%%%%%%%%%%% 
count=0; 
if m2>0 
    for i=m2:m 
        if ETF(i)>th3(i)||ZCR(i)>th5 
            m3=i; 
            count=count+1; 
        elseif count>6 
               if m3-count-m2<10 
                      m2=i; 
                      break   
               end 
        end 
    end 
end 
     
                   
                
                    
                  
 
 
figure,plot(data) 
axis([1 length(data) -1 1]) 
ylabel('Speech'); 
line([m1*64 m1*64], [-1 1], 'Color', 'red'); 
line([m2*64 m2*64], [-1 1], 'Color', 'red'); 
         
 out=[m1 m2]    
        
% figure,plot(F) 
%figure,plot(E) 
 
 
%figure,plot(T) 
%figure,plot(ZCR) 
%figure,plot(ETF)