Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!mjfg
From: mjfg@eng.cam.ac.uk (M. J. F. Gales)
Subject: Technical report available
Message-ID: <1994Mar21.130751.6851@infodev.cam.ac.uk>
Sender: news@infodev.cam.ac.uk (USENET news)
Nntp-Posting-Host: fear.eng.cam.ac.uk
Organization: Cambridge University Engineering Department, UK
Date: Mon, 21 Mar 1994 13:07:51 GMT
Lines: 71

The following technical report is available by anonymous ftp from the
archive of the Speech, Vision and Robotics Group at the Cambridge
University Engineering Department.

              ROBUST CONTINUOUS SPEECH RECOGNITION
                USING PARALLEL MODEL COMBINATION

                   Mark Gales and Steve Young

              Technical Report CUED/F-INFENG/TR172

	    Cambridge University Engineering Department 
		        Trumpington Street 
		        Cambridge CB2 1PZ 
			     England 


                             Abstract

This paper addresses the problem of automatic speech recognition in
the presence of interfering noise. It focuses on the Parallel Model
Combination (PMC) scheme, which has been shown to be a powerful
technique for achieving noise robustness. However, most experiments
reported on PMC to date have been on small, 10-50 word vocabulary
systems. In this paper, PMC is applied to the Resource Management (RM)
1000 word continuous speech recognition task. This reveals
compensation requirements not highlighted by the smaller vocabulary
tasks, in particular, it is necessary to compensate the differential
as well as the static parameters to achieve good recognition
performance.

The database used for these experiments was the RM speaker independent
task with Lynx helicopter noise from the NOISEX-92 database added.
The experiments reported here used the HTK RM recogniser developed at
CUED modified to include PMC based compensation for the static, delta
and delta-delta parameters. After training on clean speech data,
adding noise at 18-20dB signal to noise ratio was found to seriously
degrade the performance of the recogniser.  However, using PMC the
performance was restored to a level comparable with that obtained when
training directly in the noise corrupted environment. Additionally,
PMC is shown to be robust to convolutional noise for this task.

************************ How to obtain a copy ************************

a) Via FTP:

unix> ftp svr-ftp.eng.cam.ac.uk
Name: anonymous
Password: (type your email address)
ftp> cd reports
ftp> binary
ftp> get gales_tr172.ps.Z
ftp> quit
unix> uncompress gales_tr172.ps.Z
unix> lpr gales_tr172.ps (or however you print PostScript)

b) Via postal mail:

Request a hardcopy from

Mark Gales,
Cambridge University Engineering Department, 
Trumpington Street, 
Cambridge CB2 1PZ,
England.

or email me: mjfg@eng.cam.ac.uk




