I have done extensive research and build a few products to do this in the past.
They key is to do the speech recognition offline. That is capture all the speech and do the recognition as a batch job down the road. Real time recognition is normally limited to less then 1000 words/phrases and takes power and speech expert to do it. http://www.ejtalk.com/ can help with the recognition for dynamic and batch. They are familiar with asterisk. Ask for Emmett. In order to recognize and convert properly you are looking at 6 to 1 time ratio. It will take 6 seconds of CPU time to convert 1 second of speech into text. Same for doing the reverse, converting 1 second of text back to speech takes about 6 seconds. The ratio stays the same because as the converters get better they need bigger cpu's to do it right. The most important factor is the quality of the captured speech sounds/data. This is a silk purse from a sow's ear problem. Telephone g.711 is difficult to recognize, g.729 compression makes the job even harder. Make sure you are capturing at the highest level you can get. I designed created a system in the past that is used by www.orchestrate.com to do similar. What was done there was to modify the voicemail server to pass a copy of a voicemail recording, via email, to the recognizer, which converts and then email the results back. The voicemail email client, stay with me here it gets better, receives the email with the new text and attaches that to the original voicemail recording. Did I mention that we used an email server to store the voicemail recordings so that when you check your mail (dare I say universal inbox ) you got both the recording and the transcript? Let me know how it works out. And if you have more questions I am available to help. Race Vanderdecken asterisk near Vanderdecken D.O.T commie -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Cirelle Enterprises Sent: 01 November 2004 17:27 To: Asterisk Users Mailing List - Non-Commercial Discussion Subject: [Asterisk-Users] Speech to Text Conversion has anybody found anything which works for speech to text translation? Implementation being instead of (or as well as) vm wav file being sent in email, a text translation would accompany the wav file???? Regards Greg Cirino ___________________________________ Cirelle Enterprises Inc. 603-425-2221 www.cirelle.com Website Design www.cirelle.net ProSpeed High Speed Dial-up - 5 Times Faster www.cedata.com Web, FTP, Email Hosting Services www.mlsbot.com MLS IDX Services When You Want It Done Well, Just Call Cirelle It's not just a Rhyme... There's a Reason! _______________________________________________ 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
