why the Real Cepstrum code doesn't give an accurate figure?

2 次查看（过去 30 天）

显示更早的评论

Enez Furkan Cihan 2017-8-7

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/351783-why-the-real-cepstrum-code-doesn-t-give-an-accurate-figure

编辑： Enez Furkan Cihan 2017-8-7

在 MATLAB Online 中打开

Hi, implementing a basic cepstral analysis, but there's a problem that we haven't got it; please look at ;

this's the saved voice of vowel 'e' and the basic code for real cepstrum is that;

eplot=find(abs(e)>0.01,1):length(e);
y=ifft(log(abs(fft(eplot))));
plot(y)

and getting this;

I think, we should have got a figure like that ;

What we miss? Thanks in advance.

4 个评论
显示 2更早的评论隐藏 2更早的评论

Adam 2017-8-7

在 MATLAB Online 中打开

If you aren't taking the fft of the indices rather than the data it would help if you show tha actual code you are using because what you posted is doing an fft on index values which is equivalent to something like this:

y = ifft( log( abs( fft( 200:20000 ) ) ) );
figure; plot( y )

which gives a plot similar to yours.

Enez Furkan Cihan 2017-8-7

编辑：Enez Furkan Cihan 2017-8-7

在 MATLAB Online 中打开

[Y,FS]=audioread('eee.wav');
N=length(Y);
T=(0:N-1)/FS;
eplot=find(abs(T)>0.3,1):length(T);
b=ifft(log(abs(fft(Y))));
figure;subplot(2,1,1);plot(T,Y);subplot(2,1,2);plot(b)