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.Keywords-human hand 3D tracking; particle filter; microstructure of hand motion 2013 Fourth International Conference on Intelligent Systems Design and Engineering Applications /13 $26.00

No abstract

No abstract

Selection and manipulation are one of the fundamental aspects of gesture interaction. Selection is the precondition of gesture interaction. Collision detection is a key step in the gesture operation, which determines the reality and immersion of gesture interaction. However, the current related approach lack of a unified interaction model, thus it is difficult to fundamentally improve the user experience. In this paper, it is the application of navigation information to guide the object selection and collision detection. This paper obtains the navigation information of object selection and collision detection by cognitive information including the historical data of 3D hand, the hand and the object state information and the relative relationship between the hand and object, and uses the navigation information to guide the operator to carry out the next step operation. The main innovation is the establishment of a unified interaction model which is based cognitive information, and the correlation algorithm is proposed and implemented. The experimental results show that, the algorithm described in this paper not only reduce the time cost of object selection and collision detection, but also enhance the user's sense of immersion and allow users to operate more comfortable and enjoyable.

No abstract