VoxForge
VoxForge
I use stereo audio files to extract MFCC features with HTK.
I don't know how HTK processes stereo audio files.
I want to use my features which are produced with Matlab,and then compare the speech recognition rate with MFCC.
So I need to know the details how HTK works.
Does HTK works like this in Matlab code?
a=wavread(test.wav);
a=(a(:,1)+a(:,2))/2;
Does it?
My English is not very good. I hope you understand my question.
Thanks for your help.
>I use stereo audio files to extract MFCC features with HTK.
from section 5.11 of the HTKBook:
For interleaved πlaw as used in Switchboard, the default is to add the two channels together. The left channel only can be obtained by setting the environment variableSTEREOMODE to LEFT and the right channel only can be obtained by setting the environment variable STEREOMODE to RIGHT