Hi,
I want to know where I can find the sphinx 3 acoustic model which was trained
with same data that pocketsphinx default acoustic model (hub4wsj_sc_8k) was
trained.
I am currently using pocketsphinx and its default acoustic model and it did a
great job on my app. Now I need the same model in Sphinx 3 because my app.
also has an option to use Sphinx 3 decoder.
In /usr/local/share/sphinx3/model/hmm/hub4_cd.. file there is a default sphinx
3 acoustic model but accuracy is bad. I guess its not same with pocketsphinx
default acoustic model.
I looked to this address: http://www.speech.cs.cmu.edu/sphinx/models/
but didn't decide which sphinx 3 model in there will work as good as
pocketsphinx default acoustic model?
One last thing, I want to know where I can find the mixture weights of the
pocketsphinx default acoustic model?
Thanks a lot,
Berker.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am currently using pocketsphinx and its default acoustic model and it did
a great job on my app. Now I need the same model in Sphinx 3 because my app.
also has an option to use Sphinx 3 decoder.
Sphinx3 can decode using the sam esemi-continuous model as pocketsphinx. There
is no public continuous model of the same purpose available.
Sphinx3 can decode using the sam esemi-continuous model as pocketsphinx.
There is no public continuous model of the same purpose available.
Actually I need that continuous model both in using Sphinx 3 and also using in
srmc to convert HTK model.
Is there any way to convert existing semi-continuous model to continuous model
?
2-) What changes do I have to do while using those semi-cont models in Sphinx
3 ?
Thanks a lot for your help,
Berker.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I want to know where I can find the sphinx 3 acoustic model which was trained
with same data that pocketsphinx default acoustic model (hub4wsj_sc_8k) was
trained.
I am currently using pocketsphinx and its default acoustic model and it did a
great job on my app. Now I need the same model in Sphinx 3 because my app.
also has an option to use Sphinx 3 decoder.
In /usr/local/share/sphinx3/model/hmm/hub4_cd.. file there is a default sphinx
3 acoustic model but accuracy is bad. I guess its not same with pocketsphinx
default acoustic model.
I looked to this address:
http://www.speech.cs.cmu.edu/sphinx/models/
but didn't decide which sphinx 3 model in there will work as good as
pocketsphinx default acoustic model?
One last thing, I want to know where I can find the mixture weights of the
pocketsphinx default acoustic model?
Thanks a lot,
Berker.
Hi again,
I found a file in : http://www.cs.cmu.edu/~dhuggins/Projects/pocketsphinx/wsj
1/mixture_weights.bz2
is this the right file that I was asking for pocketsphinx default acoustic
model mixture weights ?
Thanks a lot,
Berker.
Sphinx3 can decode using the sam esemi-continuous model as pocketsphinx. There
is no public continuous model of the same purpose available.
Mixtures are here
http://cmusphinx.svn.sourceforge.net/viewvc/cmusphinx/trunk/pocketsphinx-
extra/
Thanks,
Actually I need that continuous model both in using Sphinx 3 and also using in
srmc to convert HTK model.
Is there any way to convert existing semi-continuous model to continuous model
?
2-) What changes do I have to do while using those semi-cont models in Sphinx
3 ?
Thanks a lot for your help,
Berker.
No
It just works I think