I’ve used htk2s3conv to convert my models from HTK to SPHINX. The conversion is correct but when I tried the converted models on PocketSPHINX decoder, it always gives “Final State not reached”. So the question is “is HTK and SPHINX acoustic models compatible?”
--- (Edited on 10/27/2011 6:13 am [GMT-0500] by ) ---
> So the question is “is HTK and SPHINX acoustic models compatible?”
The thing is that you also need to adjust feature extraction parameters to match the one from HTK. See for details
alternatively you can retrain HTK model from CMUSphinx features.
--- (Edited on 10/28/2011 16:59 [GMT+0400] by nsh) ---