Spech Recognation using Mfcc for Security

Question

ANDI RiskiWijaya 2012-9-27

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/49227-spech-recognation-using-mfcc-for-security

|i have project. to make security sistem for home using mfcc method in MATLAB.. one voice (sample .wav) would become database...

please help.... spyjaya@gmail.com

this is my work.. please download and help to find the error.. thanks

http://www.mediafire.com/?mhy1fxh1o5mf2m7

7 个评论
显示 5更早的评论隐藏 5更早的评论

ANDI RiskiWijaya 2012-10-4

在 MATLAB Online 中打开

this is my work

clear;clc;close all; pc = 60; p=[0 0 0 1 0 1 1 1 0 1 1 1 0 1 1 1 0 0 0 1 0 0 0 1 1 0 0 0 0 0 0 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 0 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0 1 1 1 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 0 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 1 1 1 0 1 1 1 0 1 1 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 0 0 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 1 0 1 1 1 0 1 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 0 0 1 1 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 1 0 0 1 1 0 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 0 0 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 0 1 0 0 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 0 1 0 0 0 1 1 1 1 1 1 1 0 0 1 1 1 1 1 1 0 1 1 1 0 1 1 1 0 0 1 0 0 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1 0 0 1 1 1 0 0 1 0 0 0 0 0 0 0 0 1 1 0 1 1 1 0 0 1 1 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 1 1 0 0 1 1 0 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 0 1 0 0 0 1 1 0 0 1 1 0 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 0 1 0 0 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 0 1 1 1 0 1 1 1 0 0 0 1 0 0 0 0 0 0 0 1 1 0 1 1 1 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 0 1 0 1 0 1 0 0 0 0 1 1 0 0 1 0 0 1 1 0 1 1 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 1 0 0 1 1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 1 1 0 0 0 1 0 0 0 0 1 0 0 1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 0 1 1 0 0 1 1 0 0 0 0 1 0 0 0 1 0 1 1 1 0 1 1 1 0 1 1 1 0 0 0 1 0 1 0 1 0 0 0 0 1];

file1 = uigetfile ('*.wav'); % mengambil file wav [y1,x1] = wavread (file1); %sound(y1); % untuk membaca file wav

%Mengubah sinyal suara dalam sampling ke dalam domain waktu y11 = y1(1:12000,1) t1 = 0:length(y11)/x1:1; t1 = 0:1.5/length(y11)-1:1.5; t1 = 0:1.5/length(y11):1.5; t1 = t1(1,1:end-1);

%Perintah untuk desain filter BPF N=10;fs=24000;fp1=2000;fp2=10000; [b,a] =ellip(N,0.5,2,[fp1 fp2]/(fs/2)); [e,f]=freqz(b,a,1024);

%perintah untuk memfilter sinyal suara z1 = filter(b,a,y11);

% set parameter decimate n = 10; %hanya diambil 1/10 sampel

% Hitung decimate g1 = decimate(z1,n);

% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Untuk menampilkan magnitude tiap inputan display (y11); % input1 = amplitude %untuk menampilkan gambar sinyal suara figure subplot (2,1,1);plot(t1,y11); title('Bentuk Sinyal suara dengan Fs 9500Hz'); xlabel('Time (sec)'); ylabel('Amplitude');grid on subplot (2,1,2); plot(y11);title('Bentuk Sinyal suara dengan Fs 9500Hz'); xlabel('sample'); ylabel('Amplitude');grid on %perintah untuk menampilkan gambar filter figure plot(f/pi*(fs/2),abs(e)); title ('Band Pass Filter');grid %perintah menampilkan sinyal hsil filter figure plot (z1); title ('Sinyal suara setelah di filter');grid xlabel('sample');ylabel('Amplitude') %Tampilkan sinyal hasil BPF kontinu, diskrit, dan hasil decimate figure subplot (2,1,1); stem(g1); title('Decimated Signal1');grid xlabel('sample');ylabel('Amplitude'); subplot (2,1,2); plot(g1); title('Decimated Signal1');grid xlabel('sample');ylabel('Amplitude'); figure stem(z1);xlabel('sample');ylabel('Amplitude');

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% Ambil input dari hasil desimasi in1=g1; alpha=1/sqrt(2); z=1;i=1;m=1;k=1;l=1;

% Hitung ekstraksi ciri berdasarkan rumus MFCC for ii = 1:size(Letters,2); for jj = 1:size(Users,2); for kk = 1:NoOfSamples file_name = strcat(Users(jj),'_',Letters(ii),int2str(kk)); Samples = eval(char(file_name)); zz = find(Samples) < max(Samples/3);%Threshold speech regions Samples(zz) = 0; zz = find(Samples); Speech_Region = Samples(zz)/norm(Samples(zz)); WindowSize = floor((size(Speech_Region,1))/(NoOfWindows+1)); ww = 0; for ll = 0:OverlapSize:(NoOfWindows-1)/2 bb = Speech_Region(floor(ll*WindowSize)+1:floor(ll*WindowSize)+WindowSize).*hamming(WindowSize); fb = fft(bb); mb = 1000 * log10(1 + fb./8000); mfout = dct(log(abs(mb)),NoOfFilters); MFCC(ii,kk,ww*NoOfFilters+1:ww*NoOfFilters+NoOfFilters) = mfout; ww = ww + 1; end end end end

%% Perform Gaussian Modelling for MFCC Windows = size(mfcc,3); tempStorage = zeros(size(Users,2)*NoOfSamples,Windows); tempStorage(:,:) = mfcc(1,:,:); obj_A = gmdistribution.fit(tempStorage,1,'Regularize',0.01); tempStorage(:,:) = mfcc(2,:,:); obj_B = gmdistribution.fit(tempStorage,1,'Regularize',0.01); tempStorage(:,:) = mfcc(3,:,:); obj_C = gmdistribution.fit(tempStorage,1,'Regularize',0.01); tempStorage(:,:) = mfcc(4,:,:); obj_Five = gmdistribution.fit(tempStorage,1,'Regularize',0.01); tempStorage(:,:) = mfcc(5,:,:); obj_Point = gmdistribution.fit(tempStorage,1,'Regularize',0.01); tempStorage(:,:) = mfcc(6,:,:); obj_V = gmdistribution.fit(tempStorage,1,'Regularize',0.01);

for ii = 1:size(Letters,2); for jj = 1:size(Users,2); for kk = 1:NoOfTestSamples %% Extract MFCC for test data InputIn = strcat(Users(jj),'_',Letters(ii),int2str(kk)); Samples = eval(char(InputIn)); zz = find(Samples) < max(Samples/3);%Threshold speech regions Samples(zz) = 0; zz = find(Samples); Speech_Region = Samples(zz); mfcc_test = zeros(1,MFCCNo); WindowSize = floor((size(Speech_Region,1))/(NoOfWindows+1)); ww = 0; for ll = 0:OverlapSize:(NoOfWindows-1)/2 bb = Speech_Region(floor(ll*WindowSize)+1:floor(ll*WindowSize)+WindowSize).*hamming(WindowSize); fb = fft(bb); mb = 2595 * log10(1 + fb./700); mfout = dct(log(abs(mb)),NoOfFilters); mfcc_test(1,ww*NoOfFilters+1:ww*NoOfFilters+NoOfFilters) = mfout; ww = ww + 1; end %% Classify MFCC test data on Mahanalobis distance D1(1) = mahal(obj_A,mfcc_test); D1(2) = mahal(obj_B,mfcc_test); D1(3) = mahal(obj_C,mfcc_test); D1(4) = mahal(obj_Five,mfcc_test); D1(5) = mahal(obj_Point,mfcc_test); D1(6) = mahal(obj_V,mfcc_test); [m Ind] = min(D1); if(Ind == ii) MFCCCorrect_Test(jj,ii) = MFCCCorrect_Test(jj,ii) + 1; end end end end

% inisialisasi output sebagai vector ciri

vektorciri=outfil6;

% plot vector ciri figure plot (vektorciri); title ('Vektor ciri Suara'); hold off xlabel('vektor ciri ke-n');ylabel('ekstraksi ciri')

%program normalisasi dan pengkodean %normalisasi vektorciri v1=max(vektorciri); vektorciri=vektorciri./v1 for i=1:24 if vektorciri(i)>0.49 vektorciri(i)=1 else vektorciri(i)=0 end end disp(vektorciri);

% Nilai fitnes for i=1:25 nilaifitnes1=0; for j=1:24 if vektorciri(j)== p(i,j) nilaifitnes1=nilaifitnes1+1; end end nilaifitnes(i)=nilaifitnes1/24*100; end

bestfitnes=max(nilaifitnes);

generate=0;

%kriteria terministik disp(nilaifitnes); %probabilitas kumuatif totalfitnes=0;

for i=1:25 totalfitnes=nilaifitnes(i)+totalfitnes; end

probabilitaskumulatif(25)=0;

probabilitaskumulatif(1)=(nilaifitnes(1)*100/totalfitnes); for j=2:25 probabilitaskumulatif(j)=(nilaifitnes(j)*100/totalfitnes) + probabilitaskumulatif(j-1); end disp(probabilitaskumulatif);

while ( bestfitnes < 98 )&( generate < 10000 )

%Fungsi seleksi for o=1:30; putar=randperm(100); seleksi=putar(1); seleksi=seleksi+probabilitaskumulatif(1); t=0; for t=1:25; if probabilitaskumulatif(t) <= seleksi for k=1:29; induk(o,k)=p(t,k); end end end end

%susun induk u=0; for i=1:30 if induk(i,25)==1; u=u+1; for l=1:29 induk1(u,l)=induk(i,l); end end end for i=1:30 if induk(i,26)==1; u=u+1; for l=1:29 induk1(u,l)=induk(i,l); end end end

for i=1:30 if induk(i,27)==1; u=u+1; for l=1:29 induk1(u,l)=induk(i,l); end end end

for i=1:30 if induk(i,28)==1; u=u+1; for l=1:29 induk1(u,l)=induk(i,l); end end end

for i=1:30 if induk(i,29)==1; u=u+1; for l=1:29 induk1(u,l)=induk(i,l); end end end

%cross over for i=1:29 a=randperm(100); probabilitascrossover=a(50); if probabilitascrossover < pc b=randperm(29); titik=b(14); for k=1:titik offspring(i,k)=induk1(i+1,k); offspring(i+1,k)=induk1(i,k); end

    for k=titik:29
        offspring(i+1,k)=induk1(i+1,k);
        offspring(i,k)=induk1(i,k);
    end
    else
    for k=1:29
    offspring(i,k)=induk1(i,k);
    offspring(i+1,k)=induk1(i+1,k);
    end
    end
    i=i+2;
end

for i=1:30 nilaifitnes1=0; for j=1:24 if vektorciri(j)== offspring(i,j) nilaifitnes1=nilaifitnes1+1; end end nilaifitnes(i)=nilaifitnes1/24*100; end bestfitnes=max(nilaifitnes); generate=generate+1; end