Hand gesture recognition for man-machine interaction

Addresses some basic problems of hand gesture recognition. Three steps important in building gestural vision-based interfaces are discussed. The first step is the hand location through detection of skin colored regions. Representative methods based on 2D color histograms are described and compared. The next step is the hand shape (posture) recognition. An approach that uses the morphological hit-miss operation is presented. Finally, application of hidden Markov models in recognition of dynamic gestures is explained. An experimental system that uses the discussed methods in real time is described and recognition results are presented