Gesture Recognition for Traffic Hand-Signals Training Simulator Using Kinect

Human gesture recognition is a way to interpret human movement and/or posture automatically. In this paper, it is used as the main interaction for the developed traffic hand-signals training simulator, based on the Kinect skeleton tracking system. Therefore, the gestures defined in this work are traffic hand-signals used in Thailand. They consist of both static postures and dynamic movements. The recognition is trained and constructed using the rule-based system. The rules must be trained to distinguish these traffic hand-signals, based on both movement information and depth-map information of hands. Then, in a part of the simulator, the artificial intelligent techniques are applied to make it realistic and challenge. The techniques include finite state machine, pathfinding, and path following. They are implemented and used for individual vehicles in the scene. Then, the performances of these two key components of the developed system, the hand-signals recognition and the traffic simulator, are evaluated. It is shown that the system can achieve a very promising performance in both aspects of the recognition accuracy and the user satisfaction.