Research on 3D Hand Tracking Using Particle Filtering

Human hand gesture is one of the natural ways used for communication for people. It also can be used for human-computer communications. So, Human hand gesture tracking and analys by artificial vision is a topic of interest. However, the high dimensions problem is one of the main barries on the way towards realtime tracking. A novel human hand gesture particle sampling method is put forward in this paper to deal with it. Microstructure of a Variable(MV), is firstly brought forward. Then, the Probability Graph(PG) is extracted from MV. Lastly, a new sampling method according to PG is presented. Both sampling and predicting are built on the base of stochastic multiple models in order to improve the tracking precision. Our experimental results demonstrate that, just using a few particles may effectively depict posteriori probability distributions in the particle filter tracking.