Hi,
I want to use pocketsphinx and convert it to HTK to use it on julius too.
The default en-us pocketsphinx in 5prealpha is very good. In the conversion I need a continuous model however the pocketsphinx is PTM. I wanted to know it the continuous 5.2 sphinx model is trained with the same corpus as the pocket en-us since I need to do a fair comparison between pocketsphinx and julius at the end.
Thank you very much !
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thank you Nickolay ! This is because I have two toolkits on my raspberry pi 3 (sphinx and julius), and I need to do a fair comparison between them. So I either use voxforge julius acoustic model as it is and then find a way to port it to sphinx or use sphinx acoustic model and then port it to HTK and then julius. But as Sphinx is based on around 800 hours of speech (compared to 140 hours of voxforge), I need to keep the sphinx and try to get the HTK. I have no other choice ! Any suggestion ?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I want to use pocketsphinx and convert it to HTK to use it on julius too.
The default en-us pocketsphinx in 5prealpha is very good. In the conversion I need a continuous model however the pocketsphinx is PTM. I wanted to know it the continuous 5.2 sphinx model is trained with the same corpus as the pocket en-us since I need to do a fair comparison between pocketsphinx and julius at the end.
Thank you very much !
This is pretty senseless activity.
Yes.
Thank you Nickolay ! This is because I have two toolkits on my raspberry pi 3 (sphinx and julius), and I need to do a fair comparison between them. So I either use voxforge julius acoustic model as it is and then find a way to port it to sphinx or use sphinx acoustic model and then port it to HTK and then julius. But as Sphinx is based on around 800 hours of speech (compared to 140 hours of voxforge), I need to keep the sphinx and try to get the HTK. I have no other choice ! Any suggestion ?
You can simply read the paper http://suendermann.com/su/pdf/oasis2014.pdf