Mel Frequency Cepstral Coefficients (MFCC) in C/C+

2019-04-29 04:43发布

问题:

Is there any implementation of MFCC available in C/C++? Any source codes or libraries?

I've already found http://code.google.com/p/libmfcc/ which seem to be good.

回答1:

A recap in 2016:

  • libmfcc is simple, MIT license, unsupported since 2010.
  • YAAFE provides MFCCs and other features, LGPLv3, unsupported since 2011.
  • Kaldi is overkill, but it can be used just for the MFCC. Apache License v2.0, and still supported.
  • PocketSphinx is the CMU toolkit for speech recognition, CMU license (BSD-style), and still supported.
  • SPTK is a research toolkit from Japan, Modified BSD license, and still supported.
  • OpenSMILE has two licenses: (1) a flexible one for research and personal use, and (2) a commercial one.

Note I mention PocketSphinx, instead of Sphinx, as the latter is written in Java.



回答2:

You could use the frontend from an open source speech recognition toolkit such as Kaldi or Sphinx.