Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!spool.mu.edu!uunet!mcsun!ukc!yorkohm!minster!russell
From: russell@minster.york.ac.uk
Newsgroups: comp.ai.neural-nets
Subject: Re: Hand Shape recognition (Sign Language)
Message-ID: <675681558.17304@minster.york.ac.uk>
Date: 31 May 91 09:19:20 GMT
Article-I.D.: minster.675681558.17304
References: <54953@nigel.ee.udel.edu>
Reply-To: russell@minster.york.ac.uk  (russell)
Organization: Department of Computer Science, University of York, England
Lines: 91

In article <54953@nigel.ee.udel.edu> peters@ee.udel.edu (Shirley Peters) writes:
>I'm looking for references and names of research done in the area of
>Sign language recognition.  This could be hand shape recognition using a
>glove sensing device, or video image recognition, or anything else that
>will ultimately end up dealing with sign language.
>
>Thanx in advance,
>Shirley

Try

Gestures and Neural Networks in Human-Computer Interaction.  IJCNN-91, Seattle.
To appear July 1991.  (R. Beale and A. D. N. Edwards)

Interpreting Gestural Input using Neural Networks.  IEE Colloquium on Neural
Nets in Human-Computer Interaction.  IEE Digest 1990/179.  (Russell Beale and
Alistair D. N.  Edwards)

(available from me if you have problems!)

In Brief
++++++++
This research used a Powerglove (a cheap DataGlove) and back-prop.  Simulated 
data showed excellent recognition results for a subset of American One-Handed
Spelling Language.  Error in signing gestures coped with perfectly
satisfactorily.  There is currently more work in progress.

Also note that other details of this work, and related stuff of potential
interest, can be found in

Pattern Recognition and Neural Networks in Human-Computer Interaction
Russell Beale and Janet Finlay, eds.  Ellis--Horwood.  Available late
1991.

Also look at Fels and Hinton, GloveTalk, 

Fels, S. S. \& Hinton, G. E. (1990) Building Adaptive Interfaces with
Neural Networks: The Glove-Talk Pilot Study, in: Diaper, D., Gilmore,
D., Cockton, G. \& Shackel B. (Eds.) {\em Human-Computer Interaction:
Interact U90}, Proceedings of the IFIP TC 13 Third International
Conference on Human-Computer Interaction, North-Holland, Oxford, pp.
683--687.
 
which uses a dataglove to produce speech via a customised language.

Gesture Recognition using Recurrent Neural Networks, Kouichi Murakami
and Hitomi Taguchi, CHI'91, pp 237--242. 

reports work on posture (static)
and gesture (dynamic) sign recognition.

Kramer, J. and Leifer, L. (1988) The talking glove: a communication aid
for deaf, deaf-blind, and non-vocal individuals, {\em Rehabilitation
Research and Development Center 1988 Progress Report}, Veterans
Administration, Palo Alto, California, pp.123--124.

use a dataglove as an aid for the disabled, while Pausch and Williams

Pausch R. and Williams, R. D. (1990) Tailor: Creating Custom User
Interfaces Based on Gesture, Computer Science Report No. TR-90-06,
Department of Computer Science, University of Virginia.

have taken a very different approach to using gestures to generate
speech. Instead of using sign language, their aim is that the
communicator to move his or her hand in motions which mimic the
movements of the tongue in natural speech.
Their objective is to use this technique as a means of communication
for people who cannot communicate vocally because of disabilities
such as cerebral palsy.

So far Pausch and Williams' results have been encouraging in that they
have managed to get subjects with cerebral palsy to consistently generate
suitable gesture curves.

The latter two use conventional pattern recognition techniques.

Hope this helps.

Russell.

____________________________________________________________
 Russell Beale, Advanced Computer Architecture Group,
 Dept. of Computer Science, University of York, Heslington,
 YORK. YO1 5DD.  UK.               Tel: [044] (904) 432771
				   Fax: [044] (904) 432767

 russell@uk.ac.york.minster			JANET
 russell%minster.york.ac.uk@nsfnet-relay.ac.uk  ARPA 
 ..!ukc!minster!russell				UUCP
 russell@minster.york.ac.uk			eab mail
____________________________________________________________