Research on Fast 3D Hand Motion Tracking System

3D hand tracking is one of the major research topics in the field of human-computer interaction. We present a novel model-based hand tracking method in this paper, which embeds hierarchical optimization method into the particle-filter-based tracking frames to improve the efficiency of particles sampling from the hidden state space. Firstly, the low dimension hidden state space is introduced to approximately describe the hand configuration state in the original high dimension configuration space, and the dynamic hand model in the hidden state space is presented according to the physiological constraints of hand motion. Secondly, to obtain more efficient particles during tracking, hierarchical genetic optimization method is regarded as the importance sampling function to modify the sampling algorithm of particle-filter. Experiments demonstrate that our approach can have fast tracking performance even under the clutter background when hand part self-occlusion exists.