A particle filter based algorithm for robust tracking of hands and face under occlusion

This paper presents a particle filter based algorithm for tracking face and hands of a signer. During signing, the hands and the face occlude each other frequently and a proper multiple object tracking algorithm is needed for accurate results. We use separate particle filters for the two hands and the face, where each filter effects the particle weights of others. This methodology ensures that the particles of each object will not jump to other objects and the tracking continues without problem during occlusion.