I dumped about 2 weeks of my life into doing just batch speech-to-text using Sphinx2. After doing all sorts of custom configurations to the Sphinx batch run-time parameters and using a very limited vocabulary I was able to recognize about 95% of the phrases that were uttered in 4000 random snippets of conversations that I recorded from Asterisk.
I was never able to get real-time conversion working in any reliable form over phone quality audio, and due to the processor and memory requirements it would be rather limiting to try using it on a busy IVR system. Sphinx runs best the more RAM it has(suggested minimum 256MB of RAM) and it is very much a processor hog. I would be very interested to hear if anyone has any experiences with IBM's viavoice product in a real-time capacity(even though I understand it is quite expensive for a multi-stream license). In the end Sphinx2 worked well enough for what I needed to do(batch processing of phrases in a limited vocabulary), but nowhere near well enough to try using it real-time in any way. Sphinx4 promises to be much better at conversion, but it is very much still beta at this time. Hope this helps. MATT--- -----Original Message----- From: Mike Meyer [mailto:[EMAIL PROTECTED] Sent: Friday, August 27, 2004 1:27 PM To: Asterisk Users Group Subject: [Asterisk-Users] Speech Recognition and Asterisk All; Since I have interest in providing the capability for callers to speak the department, person or number they wish to call, as well as other IVR scenarios, I have been reviewing much of this lists email archives and searching the web for open source voice recognition that will work with the Asterisk PBX. What I am trying to determine, is what will it take to get it working on Asterisk? How much effort and cost? So far I have uncovered references to the following: 1) VoiceXML standards, and forums 2) OpenVXI - which supports VoiceXML, simulated speech, telephony 3) PublicVoiceXML 4) Sphinx - a Carnegie Mellon University Speech recognition project funded by DARPA >From what I can tell, I feel I am uncovering the tip of the ice berg and this may not be trivial. But it seems that the Voice recognition application, once developed, would have to be linked via an AGI to the asterisk dial plan. Has anyone gotten Voice recognition working with Asterisk? Last I saw, a few were attempting to apply Sphinx back in the December and April time frame. Any shared successes, progress or direction on Sphinx or any other VR app would be appreciated before I start down this road. Thanks, Mike Meyer _______________________________________________ Asterisk-Users mailing list [EMAIL PROTECTED] http://lists.digium.com/mailman/listinfo/asterisk-users To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users _______________________________________________ Asterisk-Users mailing list [EMAIL PROTECTED] http://lists.digium.com/mailman/listinfo/asterisk-users To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
