At 06:19 PM 11/13/00 +0000, Ed Sleight wrote:
>Would it not be a logical extension that incoming audio could also be
>compared to the .wav files?
ummmm... no. You can't compare sound files anywhere near as easily as
text; speech recognition is a whole separate family of software. The basic
problems are voices coming in with many accents, at many speeds, competing
high noise levels, compression (processor) on pretty high, off frequency,
and the limitations of sound over SSB. DragonSpeaking has been trying for
5 years to get it right with comparatively pure audio and still can't
decode my voice properly. They spend a few million a year these days on
R&D, well over $30 million to date.
Anybody with DragonSpeaking want to pipe their incoming audio into DS and
see what results you get? Work let my license expire when DS failed for
three of our top 6 exec-types.
I'd advise Wayne to wait til M$ comes out with recognition-enabled OS (2000
according to the timetable released in 1998, ha ha) and give it a very
quick try then, just for giggles.
Mike N2VR
--
WWW: http://www.writelog.com/
Submissions: writelog@contesting.com
Administrative requests: writelog-REQUEST@contesting.com
Problems: owner-writelog@contesting.com
|