Pocketspinx and transcripting example

[Robert M. Münch]

Not sure if I can ask a pocketsphinx question here, but here it goes:

I would like to do a transcript of some audio files. My goal is not to get a perfect transcript but a good enough one. So some errors are no problem at all.

I think that's something pocketsphinx can do. But I don't have a clue what parameters I should give. So far I have used:

pocketsphinx_batch -hmm ./pocketsphinx-0.8/model/hmm/en_US/hub4wsj_sc_8k/ -dict ./pocketsphinx-0.8/model/lm/en_US/cmu07a.dic -ctl ./ -cepdir ./Dropbox/Apps/Rrecordings/ -cepext .mp4

The idea was to transcript the .mp4 files located on the Dropbox folder. I get a bunch of INFO messages and especially this one:

INFO: batch.c(774): TOTAL 0.00 seconds speech, 0.00 seconds CPU, 0.00 seconds wall

So, looks like either no input file is read or something else is wrong, which I can't figure out.

How can I use the command line to just try to transcript a .mp4 (or other format) audio file?

Thanks. Robert

[Nickolay V. Shmyrev]

1) Build latest sphinxbase and pocketsphinx from subversion or github

http://github.com/cmusphinx

2) Download en-us generic acoustic and language model and unpack them:

http://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/US%20English%20Generic%20Acoustic%20Model/en-us.tar.gz/download

http://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/US%20English%20Generic%20Language%20Model/cmusphinx-5.0-en-us.lm.dmp/download

3) Convert file to 16khz 16bit mono

 ffmpeg -i file.mp3 -ar 16000 -ac 1 file.wav

4) Run the transcription

 pocketsphinx_continuous -infile file.wav -hmm en-us -lm cmusphinx-5.0-en-us.lm.dmp -logfn /dev/null