【含仿真录像】基于光谱特征的语音信号元音起始点检测matlab仿真

clc; clear; close all; warning off; addpath(genpath(pwd)); [b, Fs]=audioread('test.wav'); %Audio File, Fs is sampling rate sound(b,Fs); len=length(b)/Fs; %We get time of audio in seconds sam_len = floor(((len*1000)- 10)/(10)); max = zeros(10); getans=zeros(sam_len,1); %calculating the spectral peaks corresponding to a signal with interval of %20 ms. for i=2:sam_len b_test=b(Fs*(i-1)*0.01 : Fs*(i+1)*0.01); b_test_fft = fft(b_test,256); b_test_fft2=b_test_fft(1:128); [sortedX,sortingIndices] = sort(b_test_fft2,'descend'); maxValues = sortedX(1:10); for temp=1:10 getans(i,1)=getans(i)+abs(maxValues(temp)); end end %Enhancing the signal by using the normalisation technique k=length(getans); reference = getans(2); reference_index = 2; getans=smooth(getans); temp=40; down=0; getans_new=zeros(k,1); i=0; while(i<k) if (down==0 && temp<(k-1)) while(getans(temp)< getans(temp+1)) temp=temp+1; if(temp==k) break; end end if((getans(temp)-getans(reference_index)) > 10) for now=reference_index:(temp) if (now < k) getans_new(now,1) = (getans(now)-getans(reference_index))/(getans(temp)-getans(reference_index)); end end else for now=reference_index:(temp) if (now < k) getans_new(now,1) = (getans_new(reference_index-1)); end end end reference_index=temp+1; i = temp+1; down=1; end if(down==1 && temp<(k-1)) while(getans(temp) > getans(temp+1)) temp=temp+1; if(temp==k) break; end end if((getans(temp)-getans(reference_index)) < -10) for now=(reference_index):temp getans_new(now,1)=(getans(now)-getans(temp))/(getans(reference_index)-getans(temp)); end else for now=reference_index:(temp) if (now < k) getans_new(now,1) = (getans_new(reference_index-1)); end end end reference_index=temp+1; i = temp+1; down=0; end end f_new = smooth(getans_new); windowWidth = 10; halfWidth = windowWidth / 2 gaussFilter = gausswin(5) gaussFilter = gaussFilter / sum(gaussFilter); y = conv(gaussFilter,f_new); %Axis Alignment x_axis=0:1/Fs:len; if(length(x_axis)~=length(b)) q=length(x_axis)-length(b); if(q>0) x_axis=0:1/Fs:len-q/Fs; else x_axis=0:1/Fs:len+q/Fs; end end x_axis2=0:len/length(f_new):len; if(length(x_axis2)~=length(f_new)) q=length(x_axis2)-length(f_new); if(q>0) x_axis2=0:len/length(f_new):len-(q*len/length(f_new)); else x_axis2=0:len/length(f_new):len+(q*len/length(f_new)); end end x_axis3=0:len/length(y):len; if(length(x_axis3)~=length(y)) q=length(x_axis3)-length(y); if(q>0) x_axis3=0:len/length(y):len-(q*len/length(y)); else x_axis3=0:len/length(y):len+(q*len/length(y)); end end f1=smooth(getans); %Plots- subplot(4,1,1); plot(x_axis,b); title('Time Domain Signal'); subplot(4,1,2); plot(x_axis2, f1); title('Sum of 10 lagest peaks in DFT Spectrum'); subplot(4,1,3); plot(x_axis2,f_new); title('Normalized Waveform'); subplot(4,1,4); plot(x_axis3,y); title('VOP Evidence');