how to use cmusphinx for creating a database for new language

Question

I need to develop an engine for punjabi language. i came across this link- http://cmusphinx.sourceforge.net/wiki/tutorial for tutorials of cmusphinx but it is not showing me anything. So can someone help me as how to use cmusphinx for this purpose or any other tutorial link.

Answer 1

The tutorial for acoustic model training is here:

http://cmusphinx.sourceforge.net/wiki/tutorialam

Basically you need to do the following steps:

1. Collect audio recordings
2. Transcribe them and split on utterances (short segments)
3. Dump the database in cmusphinx format
4. Prepare phoneset and phonetic dictionary
5. Run the training
6. Evaluate the model on held-out data

The steps are pretty straightforward.