🚀 go-pugleaf

In article <1895@hoptoad.uucp> gnu@hoptoad.uucp (John Gilmore) writes:
>A recent article in Newsbytes Japan mentions:
>
>	Toshiba's Voice Recognition LSI -- Toshiba (Tokyo) has developed
>	a powerful LSI for recognizing human speech.  This new product
>	recognizes a variety of spoken sounds with 95% accuracy.
>	Toshiba plans to use this LSI for a voice input system for its
>	word processors.

This is a rather meaningless statement, even for a press release.  How many
sounds?  What kind of sounds (individual phones (~=letters), words, phrases,
whistles etc.)?  If it's talking about speech sounds (as opposed to any sounds
the human vocal tract can make), what is the size of the
vocabulary one can build with it?  Do words have to be separated by silence?
Does it work in real time?  Is it even trainable?  (I can imagine having to
talk to my computer with a Japanese accent :-)  If anyone knows more about
this...

There are lots of voice recognition boards out there.  Most are fairly
primitive, which is part of the reason we haven't them used more.

Need I say that my employer doesn't necessarily share my opinion?
-- 
Mike Maxwell
Boeing Advanced Technology Center
	arpa: michaelm@boeing.com
	uucp: uw-beaver!uw-june!bcsaic!michaelm