Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!doc.ic.ac.uk!warwick!zaphod.crihan.fr!jussieu.fr!univ-lyon1.fr!swidir.switch.ch!scsing.switch.ch!xlink.net!howland.reston.ans.net!cs.utexas.edu!geraldo.cc.utexas.edu!portal.austin.ibm.com!awdprime.austin.ibm.com!pablo
From: pablo@austin.ibm.com (Paul Greenwood)
Subject: ICSS voice training
Originator: pablo@greenwood.austin.ibm.com
Sender: news@austin.ibm.com (News id)
Message-ID: <CJzwIr.qJD@austin.ibm.com>
Date: Fri, 21 Jan 1994 19:51:55 GMT
Organization: IBM Advanced Workstation Division
Keywords: ICSS voice training
Lines: 32


I have been following the discussion here but I can't remember if this has
been asked.  I know that ICSS is speaker-independent but is it possible to
train it so that you get better recognition with a particular person?
If so, then what % recognition do you think could be achieved?

A recent post stated:

>   < stuff deleted >

>IMHO the recognition after getting the microphone sensitivity set and things
>like that is still topping out at about 80% and tops out at about 95% once
>you learn the rate and flexation that the programs are looking for.

>   < other stuff deleted >

>Michael Schievelbein

Basically, could we approach 100% more closely if ICSS was "trained" and by
how much?


I hope this wasn't a perfectly stupid question.....




-- 
            -- Paul Greenwood --  (pablo@netmail.austin.ibm.com)

	Take the folks at Coca-Cola.  For many years, they were content
to sit back and make the same old carbonated beverage.  It was a good
