VoxForge
Hi,
I want to convert sphinx model to julius.
I builded Speech Recognition Model Coverter (http://sourceforge.net/projects/srmc/ ) used to transform Sphinx acoustic model to htk acoustic model.
But I have problem with running julius with converted model (Reading in HMM definition). I think it's because of format mismatching.
Is there anyone who has managed to convert model from sphinx to julius and run julius with that model?
Could you help me with that?
Thanks in advance
--- (Edited on 12/17/2013 6:32 am [GMT-0600] by grisza50) ---
It is possible in theory but for some cases it might be not.
Julius expects hmms for every triphone to be present. Sphinx has some fallbacks in case triphones are missing. Escpecially in the end of a word.
But it's hard to say without some error logs.
--- (Edited on 2/3/2014 5:02 am [GMT-0600] by Binh) ---