0
$\begingroup$

I am working on speech databases that have a stereo format. I want to extract spectral information (Mel-filterbanks, MFCC, LPCC) and also some other prosody features like the fundamental Frequency F0.

Is there a standard way to handle the computations of speech features for Stereo data?

This is for speech recognition but also spotting some irregular speech audio events, so it is different from MIR.

The maximum amplitude difference between the two channels is 7% and the mean is 0.3%. Does not tell much about the spectral features, will investigate the difference in the frequency domain now.

$\endgroup$
1
  • $\begingroup$ what i did about pitch detection on a stereo file was, rather than add the two channels and do autocorrelation on that mono audio signal, i did autocorrelation on both individually and added the two autocorrelation functions. then i did the same pitch detection on that what i normally did with a mono signal. the two channels still should be from a common monophonic source. $\endgroup$ – robert bristow-johnson Apr 4 at 3:16
0
$\begingroup$
Is there a standard way to handle the computations of speech features for Stereo data?

Not really. It really depends on why your recording is stereo in the first place and how exactly the recordings are made.

The preferred method would be to have a single close up microphone recording. Two or more channel can be helpful, if it's far field recording. Google uses two channels, Amazon uses seven. If it's stereo from produced media (movie, song, etc.) then the best extraction method would depend on the way its produced, but typically summing the two channels would be the best.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.