[asterisk-speech-rec] DTMF

[David Brazier]

Hi

I find that I'm not getting reliable results trying to detect DTMF using
Lumenvox SRE via Asterisk.  In fact, I don't know if it really recognises
DTMF at all - sometimes it just happens to match against the voice grammar -
e.g. a beep is matched against spoken "yes".  My application uses code like:

SpeechCreate
SpeechLoadGrammar(yesno, builtin:grammar/boolean)
SpeechLoadGrammar(yesno-dtmf, builtin:dtmf/boolean)
SpeechActivateGrammar(yesno)
SpeechActivateGrammar(yesno-dtmf)
SpeechStart
SpeechBackground(prompt, 5)
SpeechDeactivateGrammar(yesno-dtmf)
SpeechDeactivateGrammar(yesno)

So users can say yes or no or press 1 or 2.  However, it recognises speech
OK and but rarely DTMF.  In the Lumenvox Engine documentation, I see
references to the built-in Boolean grammar being:

builtin:dtmf/boolean?y=1;n=2

but that doesn't make a difference.  

In Call Browser, the call details don't show any DTMF events.  Calls are
originated by Asterisk via a SIP/RTP provider, and we're getting DTMF as
out-of-band RFC2833.  In the Asterisk recording, I can hear the tones have
been suppressed from the audio stream as expected, but in the Lumenvox Call
Browser, there are full-length tones, so I guess the Asterisk/Lumenvox
interface may be adding the audio tones back in, and the SRE should be
recognising the tones.  Is that how it should work?  Is there any special
configuration required to get DTMF passed through and recognised correctly?

Any help appreciated.

David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.digium.com/pipermail/asterisk-speech-rec/attachments/20070222/6f29c3fb/attachment.htm