Using HTK acoustic models with sphinx-4

Question

I have been using sphinx-4 as a speech recognition tool and I am still trying to learn the tool. I have ana Arabic acoustic model developed using HTK. I wanted to know how can I use this acoustic model in Sphinx-4.

Answer 1

There are 3 ways:

1. Use HTK models in sphinx4 decoder directly, as described in. http://nsh.nexiwave.com/2009/09/using-htk-models-in-sphinx4.html Make sure you adjusted frontend properly

2. Convert HTK models to Sphinx format using conversion scripts http://cmusphinx.sourceforge.net/2010/08/python-htk-converter/

3. Train models with Sphinxtrain from the same audio data

The last way is the easiest way to use and it will give you the biggest accuracy. For first two you need to modify the frontend configuration too. Most likely it will not be 100% compatible with HTK frontend.