Three-Dimensional Gesture Tracking Algorithm Based on Microstructure of Hand Motion

Aiming at describing human hand 3D tracking from a new perspective and reducing the time cost of hand motion tracking, a novel model-based approach to 3D hand tracking from single camera is presented. First of all this paper introduces the hand model and the virtual platform based on data glove as the basic conditions of microstructure research. Then, the basic concept of microstructure of hand motion is proposed. By applying microstructure of hand motion to exploring hand state prediction, the microstructure in the process of grasping an object is emphatically described. Finally, a human hand 3D tracking algorithm based on microstructure of hand motion is put forward. The main innovation of the algorithm is to avoid our algorithm excessively relying on the dynamic model forecast. By predicting the current state on the whole grasping process with microstructure, rather than predicting the current state on a few frames before, the proposed algorithm can sample on the most likely particles. This method avoids impossible or less likely particles as far as possible. In order to validate validity and performance of the algorithm, a great deal of experiments are completed. The results show that, compared with the particle filter (PF) gesture tracking algorithm, the proposed algorithm has low time cost.