Location of code for extracting MFCC from audio file?

[Colin Gilbert]

Hi,

I want to implement my own classifier, and I want to use the CMUSphinx MFCC-extraction implementation. I've seen many implementations, but for purposes of standardization/comparison it seems a better idea to use an existing open, tested, quality toolkit.

How would one use the code in Sphinxbase/Pocketsphinx to extract MFCC from an audio file (or live input) without passing it onto an HMM? Which functions, in which files? And, of course, I intend on open-sourcing it. :)

Thanks!
COlin

[Nickolay V. Shmyrev]

How would one use the code in Sphinxbase/Pocketsphinx to extract MFCC from an audio file (or live input) without passing it onto an HMM?

header is sphinxbase/include/sphinxbase/fe.h

implementation in sphinxbase/src/libsphinxbase/fe/*.[ch]

example is in sphinxbase/src/sphinx_fe

[Colin Gilbert]

Cool! Thanks for the fast response. I really appreciate it. I am currently experimenting with the convenient sphinx_fe tool that was compiled along with the rest of the program.

Now my big question is: Is there a way to run this tool on live microphone input? If so, how?

[Nickolay V. Shmyrev]

Now my big question is: Is there a way to run this tool on live microphone input? If so, how?

No, you can only process files.

[Colin Gilbert]

Would a unix pipe potentially work? And also, does the framerate refer to number of samples per second, or number of bytes per frame?

[Colin Gilbert]

Nevermind, I read the source code for the framerate question, and pipes are generally way too small to be reliable :P