Using JSGF grammar instead of DMP language model (use -jsgf instead of -lm) in CMU Sphinx (pocketsphinx)

Question

I need to train CMU Sphinx offline in my android app. I downloaded this project that use this parameter but when I start it I have this error: Fatal signal 11 (SIGSEGV) at 0x0000001c (code=1). This error I also have when Sphinx can not find the accoustic model. I found answer on my question here, but it is does not work for me. What I am doing wrong, can anybody can provide me any tutorial, or a good answer?

So the question is how to use .jsgf files istead of .DMP files as language model for Pocketsphinx like it used in Recognizer task in Inimesed project?

Milena

Provide the file pocketsphinx.log created on your device – Psychomancy

Here is the Log: NFO: acmod.c(242): Parsed model-specific feature parameters from /mnt/sdcard/Android/data/ee.ioc.phon.android.inimesed/files/hmm/et_ee/16000/feat.params INFO: feat.c(684): Initializing feature stream to type: 's2_4x', ceplen=13, CMN='current', VARNORM='no', AGC='none' INFO: cmn.c(142): mean[0]= 12.00, mean[1..12]= 0.0 ERROR: "acmod.c", line 84: Acoustic model definition is not specified neither with -mdef option nor with -hmm I can provide whole file if needed – Milena

This error means that you didn't put the data files into proper location. Check what folder should contain the model in sources (-hmm option) – Psychomancy

Answer 1

If CMU Sphinx works similarly on Android and on Windows, the following may help:

http://www.aiaioo.com/cms/index.php?id=28 http://homepages.abdn.ac.uk/k.vdeemter/pages/teaching/NLP/practicals/JSGFGrammar.html http://www.w3.org/TR/jsgf/

Example JSGF file:

#JSGF V1.0;
grammar hello;
public <command> = ( open | close ) ( computer | window | music | note );

Save the above grammar in a text file called "hello.gram", then run sphinx with:

pocketsphinx_continuous.exe -hmm hub4wsj_sc_8k -jsgf hello.gram -dict hub4.5000.dic

Mosasaur