Newsgroups: comp.speech
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!hookup!swrinde!howland.reston.ans.net!agate!hpg30a.csc.cuhk.hk!hkpu01.hkp.hk!encserver!stchoy
From: stchoy@encserver (Choy Sze Tsan)
Subject: About Baum-Welch algorithm -- scaling
Message-ID: <D4GJw6.9yG@hkpu01.hkp.hk>
Sender: usenet@hkpu01.hkp.hk (Usenet Account)
Nntp-Posting-Host: 158.132.51.15
Organization: Hong Kong Polytechnic University
X-Newsreader: TIN [version 1.2 PL2]
Date: Thu, 23 Feb 1995 14:50:29 GMT
Lines: 20


Recently I've been implementing the Baum-Welch 
learning algorithm for HMM parameters re-estimation, and 
found out that scalings are required. I have implemented
the scaling method from Rabiner in "A Tutorial on Hidden
Markov Models and Selected Applications in Speech 
Recognition", but still fail. I don't know what's wrong ,
may be bugs :(  . Can anyone tell me where I can get 
source codes of an implementation of this re-estimation
algorithm with scaling? Thanks in advance.
Any suggestions for any relevant references on 
the implementation aspects are also welcome. 

Best Regards
Clifford Choy
**********
*	email address: 	stchoy@encserver.en.polyu.edu.hk
*			clifford@encomp.en.polyu.edu.hk
**********

