home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.ibm.pc.soundcard
- Path: sparky!uunet!mcsun!fuug!kiae!hqlgu!news-server
- From: shura@speech.lang.lgu.spb.su (Alexander L. Gorlovsky)
- Subject: Re: Speech recognition
- Reply-To: shura@speech.lang.lgu.spb.su
- Organization: University of St-Petersburg, Speech Recognition Lab.
- Date: Wed, 12 Aug 1992 11:54:16 GMT
- Message-ID: <ABelFYgy98@speech.lang.lgu.spb.su>
- Lines: 41
- Keywords: speech recognize
- Sender: news-server@lgu.spb.su
-
-
- We are happy to inform you that we are the group trying to recognize
- speech for years and have some experience and results in this field.
- At the moment we have:
-
- SPEAKER DEPENDENT RECOGNIZOR
-
- Features: * Active vocabulary up to 500 isolated words/phrases
- * Emulation of up to 100 keystrokes per spoken word
- * Response time < 0.7 sec for 12 Mhz IBM AT286
- * Recognition accuracy better than 97%
- * Memory requirements:
- 28 Kb of conventional RAM plus about 600 bytes per word
- of active vocabulary
- * DOS environment
- * One pass training procedure and verification session
-
- SPEAKER INDEPENDENT RECOGNIZOR
-
- Features: * Active vocabulary up to 100 isolated words/phrases
- * Emulation of up to 100 keystrokes per spoken word
- * Response time < 0.7 sec for 12 Mhz IBM AT286.
- * Recognition accuracy better than 97%
- * Memory requirements:
- 28 Kb of conventional RAM plus about 3000 bytes
- per word of active vocabulary.
- * DOS environment
- * Preliminary expert preparation of the vocabulary
-
- We've developed original PC ADC/DAC board for this purposes
- which is already for sale,
- but our software runs also with COVOX, ADLIB, SOUNDBLASTER cards.
- Just now we are looking for a possibility to have contracts
- for developing this kind of products. If anybody knows something
- please make us know.
-
- Thanks. Buy!
-
- Alex.
-
-
-