home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.misc
- Path: sparky!uunet!decwrl!csus.edu!netcom.com!bskendig
- From: bskendig@netcom.com (Brian Kendig)
- Subject: Wanted: Poor man's speech recognition (something simple)
- Message-ID: <czhna=d.bskendig@netcom.com>
- Date: Mon, 31 Aug 92 19:06:17 GMT
- Organization: Starfleet Headquarters: San Francisco
- Lines: 30
-
- At the Franklin Institute in Philadelphia, there's a computer exhibit
- that features a system that asks you to repeat six words twice each,
- then lets you toggle a lamp, a windowshade, a fan, music, &c. by
- speaking one of the words.
-
- This system impressed me, but what impressed me even more was that it
- was running on an Apple IIe!
-
- If primitive speech recognition is so simple that even an Apple IIe
- can do it, then I'd be surprised if there doesn't exist some source
- code on the net that shows how to do this. All it entails, really, is
- generating a sound sample and coming up with a deviance tolerance for
- that sample during the training phase, then matching everything the
- computer "hears" against that sample to see if it falls within the
- tolerance. But I don't want to bother coming up with the algorithms
- myself (too much else to do :).
-
- Can anybody point me to code that I could use to play around with
- primitive speech recognition? I have access to a Mac with a
- microphone, and this looks like it could be a neat project. Thanks!
-
- << Brian >>
-
-
- --
- _/_/_/ Brian Kendig Macintosh Jedi Live never to be ashamed
- _/_/ Starfleet Captain Oracle Employee if anything you do or say
- _/ Intrepid Adventurer Saturn SL2 Owner is published around the world
- bskendig@netcom.com Twin Peaks Junkie -- even if what is published
- Princeton '92! BSE/CS Writer/Actor/Singer is not true.
-