Real-time 3D full body motion gesture recognition

Gesture is one of the fundamental way of human machine natural interaction. To understand this, system should be able to interpret 3D movements of human. This paper presents a real-time 3D full body gesture recognition system by tracking joints of head, neck, shoulder, arms, hands and legs. This tracking is done by Kinect motion sensor with OpenNI API and 3D full-body motion gestures are recognized. In this system, User to Kinect distance is adapted using proposed COG(center of gravity) correction method. Joint Position is normalized using the proposed 3D Joint position normalization method. For gesture recognition data mining classification algorithms such as Naive Bayes and Back Propagation Neural Network is used. The system is trained to recognize 12 gestures used by the umpires in a cricket match. It is tested with about 20000 frames of 15 users using 5-fold cross validation method. It achieved 98.11% accuracy with Neural Network based classifier and 88.84% accuracy with Naive Bayes classifier.