A novel evolutionary approach for fitting and tracking articulated skeletons to 3D data of upper-body sequences typical of videoconferencing or newscasting is presented. A 24-dimensional skeleton fitting problem is solved with a novel particle swarm optimisation technique designed for simultaneous fitting and tracking. Experiments show good quality fits even in the presence of self-occlusions and partial 3D data.
|Number of pages||2|
|Publication status||Published - 8 Dec 2005|