Speech recognition with Asterisk

Hi,

I wants to implement the Speech Recognition IVR with Asterisk. Can any one help me out how to do with dialplan and how to create the grammar file for that.

Thanks in advance

Ketan

Hi,

See:

wiki.asterisk.org/wiki/display/ … nition+API

I am also very interested in speech recognition in Asterisk. Firstly to build a Speech Recognition IVR, and secondly to do some work on the code for a colledge project. Does anyone have any contacts of people that are working on the Asterisk Speech Recognition API? Because I think that this is the best way to tackle this problem. But a big question still remains - which speech recognition software to use.

I also found these two links:

spokentech.org/index.html
scribblej.com/svn/

The first link uses some MRPC servers (is there a plan/wish to support this protocol in Asterisk?), the second is quite old (a bit over then two years) …

Anyways, any help about this topic is greately appreciated.

Ok, not manj people interested in speech recognition here :smile:

Does anyone know who is the developer of the Asterisk Speech Recognition API? An e-mail contact would be wonderfull, does anyone have it?

If you got a reply at all, it will probably be to tell you to use the mailing lists and IRC channels. It is generally considered very bad form to go direct to a developer when there are other channels and there is a warning against it on the bug reporting site.

I would point out that our marketing department managed to get a prototype working just from the easily discoverable documentation, so it can’t be very difficult.

Like I said, I am not only interested in building a speech recognition IVR, but also getting into details about the speech recognition with Asterisk, prefebly to do some work on an open source project. And for that I need some more info from a person who is doing some work on this field. Apparently I will move on to IRC channels and mailing lists :wink:

For other people interested in this field, here are two interesting links:

issues.asterisk.org/view.php?id=14578
asterisk.org/astdocs/api/AstSpeech.html

If you intend to change the actual Asterisk source code, the correct communication channels are the asterisk developer mailing lists and IRC channels.

However, I would suggest that unless you are comfortable using the feature as an end user, using the existing documentation, you are unlikely to be ready to do that.

The time to start using those channels is when you have a general understanding of the source code, but need specific points clarifying.

For IVR type of application I recommend you to use UniMRCP integrated with asterisk with uni-ast package:

http://code.google.com/p/unimrcp/wiki/asteriskUniMRCP

Powered by CMUSphinx http://cmusphinx.sourceforge.net it will provide a good recognition accuracy with open source software components. It supports several languages at your choice.

If you intend to recognize voicemail and transcribe messages you can use CMUSphinx directly

http://nsh.nexiwave.com/2010/09/voicemail-transcription-with.html

@david_55
The plan that I had is to concentrate on the speech recognition part (for the main research) and hope to integrate it eazily with the existing Asterisk Speech Recognition API. That is why I wanted to get the info from the users that have already used speech recognition with Asterisk. So in this phase I do not need details about the code, but general guiding lines on how to go about this problem.
Did your marketing team used Sphinx as the speech recognition engine or anything else?

@nsh
Thank you very much for the very usefull tip. Can you explain why it is better to use the “UniMRCP” based solution for connecting with Asterisk as apposed ot use the existing Asterisk Speech Recognition API?

Ok, I made some progress on this part. It looks like PocketSphinx + UniMRCP + Asterisk is the way to go if you want speech recognition with Asterisk.

The problem that I have now is that even though I followed all te official install tutorials for UniMRCP, I still can not get the darn thing to work in Asterisk. Does anybody have a working implementation and is willing to share the .conf files and possible tricks he needed to do to get the combination of packages working?