Lip Reader Combines Audio, Video
From: Discovery Channel - 03/13/2007
By: Tracy Staedter

Adding the ability to read lips to computer voice recognition systems could
greatly improve their accuracy, especially in "noisy" situations. Since some
sounds that are easily confused in the audio domain are easy to identify in
the visual domain and some words look identical in the visual domain,
combining audio and visual recognition would help fill the gaps in each
technique. A group of researchers at the University of East Angia, in
England, is beginning a three-year project they hope will produce a camera
capable of recognizing simple words and phrases. The technology could be
applied to both law enforcement and voice-command systems. To iron out
ambiguities in lip-reading, the team will focus on extracting information
from the lips using two approaches - modeling the shape and color of the lips
as they move and measuring the size of the mouth opening. They will then
experiment with techniques for matching visual cues to the words spoken, in
order to differentiate between words that are produced by similar lip
movements. "The fact is that it works and gives good results," says
University of Cambridge technology professor Peter Robinson. 

Read the entire article at:
http://dsc.discovery.com/news/2007/03/13/lipreader_tec.html?category=technology&guid=20070313111530&dcitc=w19-502-ak-0000

Links:
Peter Robinson
http://www.cl.cam.ac.uk/~pr10/

Richard Harvey
http://www2.cmp.uea.ac.uk/~rwh/default.htm

University of East Angia
http://www1.uea.ac.uk/cm/Home