Gesture recognition aids automatic sign language learning

Researchers at the University of Oxford and the University of Leeds have developed software that automatically learns sign language by viewing TV programs that are both subtitled and signed. The algorithm uses gesture recognition to associate gestures made by the signer with subtitle text on the TV. The research is published in a CVPR 2009 paper, titled “Learning sign language by watching TV (using weakly aligned subtitles)” by Patrick Buehler, Mark Everingham, and Andrew Zisserman. More information is available in a New Scientist article.