语音信号处理实验报告.doc

语音信号处理实验班级：学号：姓名：实验一基于MATLAB的语音信号时域特征分析（2 学时）1）短时能量（ 1）加矩形窗a=wavread('mike.wav');a=a(:,1);subplot(6,1,1),plot(a);N=32;for i=2:6h=linspace(1,1,2.^(i-2)*N);%形成一个矩形窗，长度为 2.^(i-2)*NEn=conv(h,a.*a);% 求短时能量函数Ensubplot(6,1,i),plot(En);if(i==2) ,legend('N=32' );elseif(i==3), legend('N=64' );elseif(i==4) ,legend('N=128' );elseif(i==5) ,legend('N=256' );elseif(i==6) ,legend('N=512' );endend1-10.5 1 1.5 2 2.54 x 102N=32 0 0.5 1 1.5 2 2.55 x 10N=64 0 0.5 1 1.5 2 2.510 x 1050.5 1 1.5 2 2.5 N=128 020 x 10100.5 1 1.5 2 2.5 N=256 040 x 1020N=5120 0.5 1 1.5 2 2.5x 10（ 2）加汉明窗a=wavread('mike.wav');a=a(:,1);subplot(6,1,1),plot(a);N=32; 3 43 43 43 43 43 4for i=2:6h=hanning(2.^(i-2)*N);% 形成一个汉明窗，长度为 2.^(i-2)*NEn=conv(h,a.*a);% 求短时能量函数Ensubplot(6,1,i),plot(En);if(i==2), legend('N=32' );elseif(i==3), legend('N=64' );elseif(i==4) ,legend('N=128' );elseif(i==5) ,legend('N=256' );elseif(i==6) ,legend('N=512' );endend1-10.5 1 1.5 2 2.52 x 101N=32 0 0.5 1 1.5 2 2.54 x 102N=64 0 0.5 1 1.5 2 2.54 x 102N=128 0 0.5 1 1.5 2 2.510 x 105N=256 0 0.5 1 1.5 2 2.520 x 1010N=5120 0.5 1 1.5 2 2.5x 102）短时平均过零率a=wavread('mike.wav');a=a(:,1);n=length(a);N=320;subplot(3,1,1),plot(a);h=linspace(1,1,N);En=conv(h,a.*a); %求卷积得其短时能量函数Ensubplot(3,1,2),plot(En);for i=1:n-1if a(i)>=0 3 43 43 43 43 43 4elseb(i) = -1;endif a(i+1)>=0b(i+1)=1;elseb(i+1)= -1;endw(i)=abs(b(i+1)-b(i)); %求出每相邻两点符号的差值的绝对值endk=1;j=0;while (k+N-1)<nZm(k)=0;for i=0:N-1;Zm(k)=Zm(k)+w(k+i);endj=j+1;k=k+N/2; % 每次移动半个窗endfor w=1:jQ(w)=Zm(160*(w-1)+1)/(2*N); %短时平均过零率endsubplot(3,1,3),plot(Q),grid;1-100.51 1.52 2.5 34x 10201000.51 1.52 2.5 34x 100.50204060801001201401601803）自相关函数N=240y=wavread('mike.wav');y=y(:,1);x=y(13271:13510);x=x.*rectwin(240);R=zeros(1,240);for k=1:240for n=1:240-kR(k)=R(k)+x(n)*x(n+k);endendj=1:240;plot(j,R);grid;2.521.510.5-0.5-1-1.5050100150200250实验二基于 MATLAB 分析语音信号频域特征1）短时谱cleara=wavread('mike.wav');a=a(:,1);subplot(2,1,1),plot(a);title('original signal');gridN=256;h=hamming(N);for m=1:Nb(m)=a(m)*h(m)endy=20*log(abs(fft(b)))subplot(2,1,2)plot(y);title('短时谱 ');gridoriginal signal10.5-0.5-100.51 1.52 2.5 34x 10短时谱10.500.20.40.60.81 1.2 1.4 1.6 1.8 22）语谱图[x,fs,nbits]=wavread('mike.wav')x=x(:,1);specgram(x,512,fs,100);xlabel('时间 (s)');ylabel('频率 (Hz)' );title('语谱图 ');语谱图50004000)zH3000(率频200010000.51 1.5 2时间 (s) 3）倒谱和复倒谱（1）加矩形窗时的倒谱和复倒谱cleara=wavread('mike.wav',[4000,4350]);a=a(:,1);N=300;h=linspace(1,1,N);for m=1:Nb(m)=a(m)*h(m);endc=cceps(b);c=fftshift(c);d=rceps(b);d=fftshift(d);subplot(2,1,1)plot(d);title( '加矩形窗时的倒谱')subplot(2,1,2)plot(c);title( '加矩形窗时的复倒谱')加矩形窗时的倒谱1-1-2050100150200250300加矩形窗时的复倒谱105-5-10050100150200250300（2）加汉明窗时的倒谱和复倒谱 cleara=wavread('mike.wav',[4000,4350]);a=a(;,1);N=300;h=hamming(N);for m=1:Nb(m)=a(m)*h(m);endc=cceps(b);c=fftshift(c);d=rceps(b);d=fftshift(d);subplot(2,1,1)plot(d);title( '加汉明窗时的倒谱')subplot(2,1,2)plot(c);title( '加汉明窗时的复倒谱')加汉明窗时的倒谱1-1-2-3050100150200250300加汉明窗时的复倒谱105-5-10050100150200250300实验三基于 MATLAB 的 LPC 分析MusicSource = wavread('mike.wav');MusicSource=MusicSource(:,1);Music_source = MusicSource';N = 256; % window length， N = 100 -- 1000;Hamm = hamming(N); % create Hamming windowframe = input( '请键入想要处理的帧位置= ' );% origin is current frameorigin = Music_source(((frame - 1) * (N / 2) + 1):((frame - 1) * (N / 2) + N));Frame = origin .* Hamm';%%Short Time Fourier Transform%[s1,f1,t1] = specgram(MusicSource,N,N/2,N);[Xs1,Ys1] = size(s1);for i = 1:Xs1FTframe1(i) = s1(i,frame);endN1 = input( '请键入预测器阶数= ' ); % N1 is predictor's order[coef,gain] = lpc(Frame,N1); % LPC analysis using Levinson-Durbin recursionest_Frame = filter([0 -coef(2:end)],1,Frame); % estimate frame(LP)FFT_est = fft(est_Frame);err = Frame - est_Frame; % error% FFT_err = fft(err);subplot(2,1,1),plot(1:N,Frame,1:N,est_Frame,'-r');grid;title('原始语音帧 vs.预测后语音帧 ')subplot(2,1,2),plot(err);grid;title('误差 ');pause%subplot(2,1,2),plot(f',20*log(abs(FTframe2)));grid;title('短时谱 ')%%Gain solution using G^2 = Rn(0) - sum(ai*Rn(i)),i = 1,2,...,P%fLength(1 : 2 * N) = [origin,zeros(1,N)];Xm = fft(fLength,2 * N);X = Xm .* conj(Xm);Y = fft(X , 2 * N);Rk = Y(1 : N);PART = sum(coef(2 : N1 + 1) .* Rk(1 : N1));G = sqrt(sum(Frame.^2) - PART);A = (FTframe1 - FFT_est(1 : length(f1'))) ./ FTframe1 ; % inverse filter A(Z)subplot(2,1,1),plot(f1',20*log(abs(FTframe1)),f1',(20*log(abs(1 ./ A))),'-r');grid;title('短时谱 ');subplot(2,1,2),plot(f1',(20*log(abs(G ./ A))));grid;title( 'LPC谱 ');pause%plot(abs(ifft(FTframe1 ./ (G ./ A))));grid;title('excited')%plot(f1',20*log(abs(FFT_est(1 : length(f1')) .* A / G )));grid;%pause%%find_pitch%temp = FTframe1 - FFT_est(1 : length(f1'));%not move higher frequncepitch1 = log(abs(temp));pLength = length(pitch1);result1 = ifft(pitch1,N);% move higher frequncepitch1((pLength - 32) : pLength) = 0;result2 = ifft(pitch1,N);%direct do real cepstrum with errpitch = fftshift(rceps(err));origin_pitch = fftshift(rceps(Frame));subplot(211),plot(origin_pitch);grid;title( '原始语音帧倒谱 (直接调用函数 )');subplot(212),plot(pitch);grid;title( '预测误差倒谱 (直接调用函数 )');pausesubplot(211),plot(1:length(result1),fftshift(real(result1)));grid;title('预测误差倒谱 (根据定义编写，没有去除高频分量)');subplot(212),plot(1:length(result2),fftshift(real(result2)));grid;title('预测误差倒谱 (根据定义编写，去除高频分量 )');原始语音帧 vs. 预测后语音帧0.40.2-0.2-0.4050100150200250300误差0.20.1-0.1-0.2050100150200250300短时谱50-50-100010203040506070LPC 谱100806040010203040506070原始语音帧倒谱(直接调用函数)0.5-0.5-1050100150200250300预测误差倒谱(直接调用函数)0.5-0.5-1050100150200250300预测误差倒谱(根据定义编写，没有去除高频分量)0.2-0.2-0.4-0.6050100150200250300预测误差倒谱(根据定义编写，去除高频分量)0.1-0.1-0.2-0.3050100150200250300预测误差倒谱(根据定义编写，没有去除高频分量)0.2-0.2-0.4-0.6050100150200250300预测误差倒谱(根据定义编写，去除高频分量)0.1-0.1-0.2-0.3050100150200250300预测误差倒谱(根据定义编写，没有去除高频分量)0.2-0.2-0.4-0.6050100150200250300预测误差倒谱(根据定义编写，去除高频分量)0.1-0.1-0.2-0.3050100150200250300实验四基于 VQ 的特定人孤立词语音识别研究1、mfcc.mfunction ccc = mfcc(x)bank=melbankm(24,256,8000,0,0.5,'m' );bank=full(bank);bank=bank/max(bank(:));for k=1:12n=0:23;dctcoef(k,:)=cos((2*n+1)*k*pi/(2*24));endw = 1 + 6 * sin(pi * [1:12] ./ 12);w = w/max(w);xx=double(x);xx=filter([1 -0.9375],1,xx);xx=enframe(xx,256,80);for i=1:size(xx,1)y = xx(i,:);s = y' .* hamming(256);t = abs(fft(s));t = t.^2;c1=dctcoef * log(bank * t(1:129));c2 = c1.*w';m(i,:)=c2';enddtm = zeros(size(m));for i=3:size(m,1)-2dtm(i,:) = -2*m(i-2,:) - m(i-1,:) + m(i+1,:) + 2*m(i+2,:);enddtm = dtm / 3;ccc = [m dtm];ccc = ccc(3:size(m,1)-2,:);2、vad.mfunction [x1,x2] = vad(x)x = double(x);x = x / max(abs(x));FrameLen = 240;FrameInc = 80;amp1 = 10;amp2 = 2;zcr1 = 10;zcr2 = 5;maxsilence = 8; % 6*10ms = 30msminlen = 15; % 15*10ms = 150msstatus = 0;count = 0;silence = 0;tmp1 = enframe(x(1:end-1), FrameLen, FrameInc);tmp2 = enframe(x(2:end) , FrameLen, FrameInc);signs = (tmp1.*tmp2)<0;diffs = (tmp1 -tmp2)>0.02;zcr = sum(signs.*diffs, 2);amp = sum(abs(enframe(filter([1 -0.9375], 1, x), FrameLen, FrameInc)), 2); amp1 = min(amp1, max(amp)/4);amp2 = min(amp2, max(amp)/8);x1 = 0;x2 = 0;for n=1:length(zcr)goto = 0;switch statuscase {0,1}if amp(n) > amp1x1 = max(n-count-1,1);status = 2;silence = 0;count= count + 1;elseif amp(n) > amp2 | ...zcr(n) > zcr2status = 1;count = count + 1;elsestatus = 0;count= 0;endcase 2,if amp(n) > amp2 | ...zcr(n) > zcr2count = count + 1;elsesilence = silence+1;if silence < maxsilence count = count + 1;elseif count < minlenstatus = 0;silence = 0;count= 0;elsestatus = 3;endendcase 3,break;endendcount = count-silence/2;x2 = x1 + count -1;3、codebook.m%clear;function xchushi= codebook(m) [a,b]=size(m);[m1,m2]=szhixin(m);[m3,m4]=szhixin(m2);[m1,m2]=szhixin(m1);[m7,m8]=szhixin(m4);[m5,m6]=szhixin(m3);[m3,m4]=szhixin(m2);[m1,m2]=szhixin(m1);[m15,m16]=szhixin(m8);[m13,m14]=szhixin(m7);[m11,m12]=szhixin(m6);[m9,m10]=szhixin(m5);[m7,m8]=szhixin(m4);[m5,m6]=szhixin(m3);[m3,m4]=szhixin(m2);[m1,m2]=szhixin(m1);chushi(1,:)=zhixinf(m1);chushi(2,:)=zhixinf(m2);chushi(3,:)=zhixinf(m3);chushi(4,:)=zhixinf(m4);chushi(5,:)=zhixinf(m5);chushi(6,:)=zhixinf(m6);chushi(7,:)=zhixinf(m7);chushi(8,:)=zhixinf(m8);chushi(9,:)=zhixinf(m9);chushi(10,:)=zhixinf(m10);chushi(11,:)=zhixinf(m11);chushi(12,:)=zhixinf(m12);chushi(13,:)=zhixinf(m13);chushi(14,:)=zhixinf(m14);chushi(15,:)=zhixinf(m15);chushi(16,:)=zhixinf(m16);sumd=zeros(1,1000);k=1;dela=1;xchushi=chushi;while(k<=1000)sum=ones(1,16);for p=1:afor i=1:16d(i)=odistan(m(p,:),chushi(i,:));enddmin=min(d);sumd(k)=sumd(k)+dmin;for i=1:16if d(i)==dminxchushi(i,:)=xchushi(i,:)+m(p,:);sum(i)=sum(i)+1;endendendfor i=1:16xchushi(i,:)=xchushi(i,:)/sum(i); endif k>1dela=abs(sumd(k)-sumd(k-1))/sumd(k); endk=k+1;chushi=xchushi;endreturn4、 testvq.mclear;disp('这是一个简易语音识别系统，请保证已经将您的语音保存在相应文件夹中') disp('正在训练您的语音模版指令，请稍后...')for i=1:10fname = sprintf(海儿的声音\\%da.wav' ,i-1);x = wavread(fname);[x1 x2] = vad(x);m = mfcc(x);m = m(x1:x2-5,:);ref(i).code=codebook(m);enddisp('语音指令训练成功，恭喜！)?'disp('正在测试您的测试语音指令，请稍后...')for i=1:10fname = sprintf(海儿的声音\\%db.wav' ,i-1);x = wavread(fname);[x1 x2] = vad(x);mn = mfcc(x);mn = mn(x1:x2-5,:);%mn = mn(x1:x2,:)test(i).mfcc = mn;endsumsumdmax=0;sumsumdmin=0;disp('对训练过的语音进行测试')for w=1:10sumd=zeros(1,10);[a,b]=size(test(w).mfcc);for i=1:10for p=1:afor j=1:16d(j)=odistan(test(w).mfcc(p,:),ref(i).code(j,:));enddmin=min(d);sumd(i)=sumd(i)+dmin; %×üê§??endendsumdmin=min(sumd)/a;sumdmin1=min(sumd);sumdmax(w)=max(sumd)/a;sumsumdmin=sumdmin+sumsumdmax;sumsumdmax=sumdmax(w)+sumsumdmax;disp('正在匹配您的语音指令，请稍后...')for i=1:10if (sumd(i)==sumdmin1)switch (i)case 1fprintf( '您输入的语音指令为:%s; 识别结果为 %s\n' ,'前 ', '前 ');case2fprintf( '您输入的语音指令为:%s; 识别结果为 %s\n' ,'后 ', '后 ');case3fprintf( '您输入的语音指令为:%s; 识别结果为 %s\n' ,'左 ', '左 ');case4fprintf( '您输入的语音指令为a:%s;识别结果为 %s\n' ,'右 ', '右 ');case5fprintf( '您输入的语音指令为:%s; 识别结果为 %s\n' ,'东 ', '东 ');case6fprintf( '您输入的语音指令为:%s; 识别结果为 %s\n' ,'南 ', '南 ');case7fprintf( '您输入的语音指令为:%s; 识别结果为 %s\n' ,'西 ', '西 ');case8fprintf( '您输入的语音指令为:%s; 识别结果为 %s\n' ,'北 ', '北 ');case9fprintf( '您输入的语音指令为a:%s;识别结果为 %s\n' ,'上 ', '上 ');case10fprintf( '您输入的语音指令为a:%s;识别结果为 %s\n' ,'下 ', '下 ');otherwisefprintf( 'error');endendendenddelamin=sumsumdmin/10;delamax=sumsumdmax/10;disp('对没有训练过的语音进行测试')disp('正在测试你的语音，请稍后...')for i=1:10fname = sprintf(o￡ ?ùμ ?éùò?\\%db.wav' ,i-1);x = wavread(fname);[x1 x2] = vad(x);mn = mfcc(x);mn = mn(x1:x2-5,:);%mn = mn(x1:x2,:)test(i).mfcc = mn;endfor w=1:10sumd=zeros(1,10);[a,b]=size(test(w).mfcc);for i=1:10for p=1:afor j=1:16d(j)=odistan(test(w).mfcc(p,:),ref(i).code(j,:));enddmin=min(d);sumd(i)=sumd(i)+dmin; %×üê§??endendsumdmin=min(sumd);z=0;for i=1:10if (((sumd(i))/a)>delamax)||z=z+1;endenddisp('正在匹配您的语音指令，请稍后...')if z<=3for i=1:10if (sumd(i)==sumdmin)switch (i)case 1fprintf( '您输入的语音指令为:%s; 识别结果为%s\n' ,'前 ', '前 ');case2fprintf( '您输入的语音指令为 :%s; 识别结果为 %s\n' ,'后 ', '后 ');case3fprintf( '您输入的语音指令为 :%s; 识别结果为 %s\n' ,'左 ', '左 ');case4fprintf( '您输入的语音指令为a:%s;识别结果为%s\n' ,'右 ', '右 ');case5fprintf( '您输入的语音指令为:%s; 识别结果为%s\n' ,'东 ', '东 ');case6fprintf( '您输入的语音指令为:%s; 识别结果为%s\n' ,'南 ', '南 ');case7fprintf( '您输入的语音指令为:%s; 识别结果为%s\n' ,'西 ', '西 ');case8fprintf( '您输入的语音指令为 :%s; 识别结果为 %s\n' ,'北 ', '北 ');case9fprintf( '您输入的语音指令为a:%s;识别结果为%s\n' ,'上 ', '上 ');case10fprintf( '您输入的语音指令为a:%s;识别结果为%s\n' ,'下 ', '下 ');otherwisefprintf( 'error');endendendelsefprintf( '您输入的语音无效?\n'￡)?endend（此文档部分内容来源于网络，如有侵权请告知删除，文档可自行编辑修改内容，供参考，感谢您的配合和支持）。