Intel teaches computers to lip-read

Intel has released open source speech-recognition software that teaches computers to "read" users lips.

Combined with face detection algorithms from Intel's OpenCV computer vision library, the company's Audio Visual Speech Recognition software allows computers to track mouth movements and enhance speech applications in noisy environments.

The accuracy of current speech recognition software is limited by background noise, with the result that users often have to rely on specialist headsets or find a suitable quiet place, the company said.

"Human recognition is seldom based on a single type of information," said Justin Rattner, director of the company's Microprocessor Research Labs, in a statement.

"Intel wants to develop technology that allows computers to interact with the world the way humans do."

The software is available as part of OpenCV, Intel's open source computervision library.