Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!doc.ic.ac.uk!uknet!pipex!uunet!europa.eng.gtefsd.com!howland.reston.ans.net!math.ohio-state.edu!cs.utexas.edu!utnut!torn!nott!cunews!dominic
From: dominic@oeg.carleton.ca (Dominic Richens)
Subject: DIY speaker _independent_ recognition
Message-ID: <CB71pH.EEv@cunews.carleton.ca>
Followup-To: poster
Originator: dominic@
Keywords: speaker independent recognition
Sender: news@cunews.carleton.ca (News Administrator)
Organization: OEG - Ontario Telepresence Project
Date: Tue, 3 Aug 1993 17:24:53 GMT
Lines: 42


Hi all,

Could someone offer some opinion as what sort of modifications would be
needed to make Doug Danforth's "QUICKY RECOGNIZER" more reliable in a
speaker _independent_ context?

My assumption is that it one would simply use a larger training set, and an
appropriate classifier (i.e. neural net) instead of euclidean distance.

I'm only interested in recognising "yes", "no" and the digits.

thanx in advance,

P.S. I've appended the first part of Doug Danforth's posting:

! QUICKY RECOGNIZER sketch:
! 
[...]
! 
! Overview:
! (1) Find the begining and end of the utterance.
! (2) Filter the raw signal into frequency bands.
! (3) Cut the utterance into a fixed number of segments.
! (4) Average data for each band in each segment.
! (5) Store this pattern with its name.
! (6) Collect training set of about 3 repetitions of each pattern (word).
! (7) Recognize unknown by comparing its pattern against all patterns
!    in the training set and returning the name of the pattern closest
!    to the unknown.
!
[...]
!
!Good luck,
!
!Doug Danforth

-- 
!@$^&*)($#$@!@#$^&*()_+_)(*&^$#@!$^&*()_+)(*&^$#@$^&*+_(*&^$#@#^&*()&*$#@(*&+_$
     Dominic Richens  :  dominic@oeg.carleton.ca  :  Tel. (613) 820 0764 
Ontario Telepresence Project, 2670 Queensview Dr., Ottawa, ON, K1N 6N5, CANADA
      "There are 70 billion people in the world, where are they hiding?"
