Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!doc.ic.ac.uk!agate!usenet.ins.cwru.edu!howland.reston.ans.net!darwin.sura.net!sgiblab!munnari.oz.au!cs.mu.OZ.AU!karenj
From: karenj@ee.mu.OZ.AU (Karen Jenkin)
Subject: endpoint detection
Message-ID: <karenj.736735653@mullian.ee.Mu.OZ.AU>
Keywords: endpoint detection
Sender: news@cs.mu.OZ.AU
Organization: Computer Science, University of Melbourne, Australia
Date: Fri, 7 May 1993 00:47:33 GMT
Lines: 12

I am attempting to find the beginning of an utterance using the algorithm by 
Rabiner and Sambur (Bell Sys. Tech. Journal, V.54 #2, 1975). Their algorithm 
is tailored for telephone speech, but my utterances aren't from this medium
(TIMIT). OK, so with their algorithm, the start is indicated *way* too early.

Has anyone else adapted their algorithm to "normal" speech (not on telephone).
In particular, I think the equations for the thresholds need to be adjusted
somehow for speech NOT high pass filtered at 100Hz, nor low pass filtered at
4kHz.

Thanks, Karen
(karenj@mullian.ee.mu.oz.au)
