Hand gestures recognition using dynamic Bayesian networks

In this study a method for hand gesture recognition using dynamic Bayesian networks was presented. This study includes two main subdivisions namely: hand posture recognition and dynamic hand gesture recognition (without hand posture recognition). In the first session, after hand segmentation using a method based on histogram of direction and fuzzy SVM classifier, we train the posture recognition system. In the second session, after skin detection and face and hands segmentation, their tracing were carried out by means of Kalman filter. Then, by tracing the obtained data, the positions of hand was achieved. For combining the achieved data and output of hand posture recognition unit we utilize Bayesian dynamic network. For recognition of 12 hand gestures in this study, 12 Bayesian dynamic networks with two distinct designs were used. The difference between these two models is in the utilizing features and their relations with each other. Therefore, one of these models was used based on each gesture feature. The results of implementation show the about 90% average accuracy for all gestures.