[asterisk-dev] How to stream audio to external app forspeech recognition and recognize dtmf in parallel ?

[Tim Panton]

On 2 Oct 2006, at 12:03, Robert Rozman wrote:

> Hi,
>
> we're writting interface module for our speech recognition system.  
> We would
> like to export stream of audio samples to external app, but to  
> preserve dtmf
> recognition and dialplan progress.
>
> I wonder if recording application would be a good start for that  
> (recording
> application obviously streams audio and makes recording out of it in
> parallel)
>
>
>
> We're also interested in best way to report speech recognition results
> back....
>
> Best way would be to be able to call extension in dialplan, for  
> instance :
>
> 1, DTMF "1"
> 2, DTMF "2"
>
> support, spoken word "support"
> sales, spoken word "sales"
>
> ------
>
> but also putting results in variable would be probably fine....
>
>
> Any advice how to develope such scenario ? What is the best module  
> code to
> start with ? Any similar solutions ?
>

I think you should be able to do _all_ of this via the agi interface.
You can add agi's to the dialplan, pass them arguments and they can set
variables. An EAGI can also get the audio stream.

The only problem I see is with your DTMF handling. Asterisk will  
probably
already have recognized the DTMF long before it gets to your app.
The channel drivers all do DTMF in their own way - for some
input channels (like zap/PRI and IAX) the DTMF was never stored as
a tone, which means you would have to arrange to synthesize it for
those channels, just for your recogntion engine to decode them !


Tim Panton

www.mexuar.com