Hello all,

Has anyone tried using PD's FFT capabilities to extract MFCC (Mel-Frequency Cepstral Coefficients ) values from incoming (real-time) audio? Or indeed, from a bank of sampled audio?

I would be interested in hearing suggestions from anyone as to how to go about it, since looking at the math formulas is making my head hurt and I can't seem to formulate a plan of action to get it done.

BTW, the reason behind trying to get MFCCs from FFT analysis is so I can feed them to either the ANN object or a gridlflow/ruby FANN neural network, for classification. If anyone has already done that, I would be ecstatic to hear from them!

Cheers,

David