|
Dear list Another lead: Malcom Slaney implemented mfcc calculations in the Matlab Auditory Toolbox using a method that is different from the formulae previously cited on the list. Once I compared his method to the formula 2595*log10(1+hz/700) and found little difference. Malcom may remember where he devised the method from (I would bet from Rabiner's book?). By the way, I found the log10(1+hz/700) formula in the following conference paper: @InProceedings{mol01, author = "Sirko Molau and Michael Pitz and Ralf Schlüter and Hermann Ney", title = "Computing Mel-Frequency Cepstral Coefficients on the Power Spectrum", booktitle = {International Conference on Acoustic, Speech and Signal Processing}, year = {2001}, address = {Salt Lake City, UT}, month = {June}, } I don't currently have the paper at hand, but if someone could check, they may cite their source. Hope that it helps. Best regards Guillaume --
----------------------------------------------------------------------------
Equipe Perception et Design Sonores / Sound Perception and Design Team
STMS-IRCAM-CNRS UMR 9912 1, place Igor Stravinsky F-75004 Paris - FRANCE tel : (+33 1) 44.78.48.38 fax : (+33 1) 44.78.15.40 e-mail : lemaitre@xxxxxxxx -------------------------------------- -------------------------------------- |