Machine Vision Recognition of American Sign Language Using Hidden Markov Models

Thad Starner, Joshua Weaver, and Alex Pentland

Read about our effort to put a computer vision based ASL translator in a baseball cap. Our current gloveless wearable system recognizes 40 word vocabulary ASL in 5 word sentences from the cap mounted camera at 97+% word accuracy, even with an unrestricted grammar! This result is better than our previous desk based system. For details and the latest information see Perceptual Computing TR #466 (compressed Postscript) or see the html version . This paper will come out eventually in PAMI in 1998 (submitted 4/26/96).

Our older technical report can be found atTR #375 (compressed Postscript) or see the html version .

For a tutorial on HMM's, see Thad Starner's Master's thesis TR #316 (compressed Postscript).