using asterisk-unimrcp for adaptation

[hiyassat]

Hi Arsen,
we installed asterisk and UniMRCP with Pocketsphinx following
http://code.google.com/p/unimrcp/wiki/PocketSphinxPlugin
on same centos OS machine , i am using peers as sip client
http://peers.sourceforge.net/
in order to get best of pocketsphinx you should adapt the HMM used in
recognition, for specific user, in order to do this i need to
implement the following scenario
the user is prompted to read certain group of sentences, then each
sentence is mapped to its wave file and both the wave and
transcription is used to adapt the HMM to certain user, my questions
is :
1-is it possible to map the audio file that resulted from certain
session with the its transcription ? i.e how can i store the audio
file at server with reference to certain user and certain text and its
recognition result?
2- is it possible to have more than one HMM and directed user channel
to his adapted HMM?
thank you

[Dr. Hussein Hiyassat]

any help please

[Arsen Chaloyan]

Hi Hiyassat,

On Tue, May 22, 2012 at 5:57 AM, hiyassat <hiya...@gmail.com> wrote:
> Hi Arsen,
> we installed asterisk and UniMRCP with Pocketsphinx  following
> http://code.google.com/p/unimrcp/wiki/PocketSphinxPlugin
>  on same  centos OS machine , i am using peers as sip client
> http://peers.sourceforge.net/

Well, your setup is clear.

> in order to get best of pocketsphinx you should adapt the HMM used in
> recognition, for specific user,  in order to do this i need to
> implement the following scenario
> the user is prompted to read certain group of sentences, then each
> sentence is mapped to its wave file and both the wave and
> transcription is used to adapt the HMM to certain user,

And the goal is also clear and reasonable to me. Before going into
further details, I assume you realize that the PocketSphinx plugin
currently doesn't support the described functionality. Another
question is how this can be achieved using the MRCP framework in
general. You may need to look for voice enrolled grammars or so called
speaker-dependent recognition.

http://tools.ietf.org/html/draft-ietf-speechsc-mrcpv2-27#section-9.9

>my questions
> is :
> 1-is it possible to map the audio file that resulted from certain
> session with the its  transcription ? i.e how can i store the audio
> file at server with reference to certain user and certain text and its
> recognition result?

Use the corresponding header fields such as Personal-Grammar-URI,
Phrase-Id in an enrollment session.

> 2- is it possible to have more than one HMM and directed user channel
> to his adapted HMM?

I'm not sure I understood your question right. Perhaps you may need to
associate HMM with Personal-Grammar-URI, which is the handle between
the client and server.


> thank you

HTH

>
> --
> You received this message because you are subscribed to the Google Groups "UniMRCP" group.
> To post to this group, send email to uni...@googlegroups.com.
> To unsubscribe from this group, send email to unimrcp+u...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/unimrcp?hl=en.
>



--
Arsen Chaloyan
Author of UniMRCP
http://www.unimrcp.org