Intel has released software which enables a computer to
perform a similar task to human lip-reading, as an aid to existing
sound-based voice recognition systems.
The Audio Visual Speech Recognition (AVSR) software should
improve the accuracy of speech-recognition software under difficult
conditions, especially those involving background noise, Intel
said.
The aim of AVSR is to enable computers to synchronise the video
data captured on camera with the sound data to produce more
accurate speech recognition.
AVSR is part of Intel's OpenCV computer vision library, a
toolbox of imaging functions, which contains a number of face
detection algorithms.
With the speed of today's microprocessors, falling camera prices
and much greater video capture bandwidth from technologies like
Universal Serial Bus 2, mainstream PCs are capable of running
real-time computer vision algorithms, the company said.
OpenCV is an open-source code library which has seen more than
500,000 code downloads to date, Intel said.
Information about AVSR can be found at
http://www.intel.com/research/mrl/research/avcsr.htm.