Motion Capture of Articulated Chains by Applying Auxiliary Information to the Sequential Monte Carlo Algorithm

In recent years Sequential Monte Carlo (SMC) algorithms have been applied to capture the motion of humans. In this paper we apply a SMC algorithm to capture the motion of an articulated chain, e.g., a human arm, and show how the SMC algorithm can be improved in this context by applying auxiliary information. In parallel to a model-based approach we detect skin color blobs in the image as our auxiliary information and find the probabilities of each blob representing the hand. The probabilities of these blobs are used to control the drawing of particles in the SMC algorithm and to correct the predicted particles. The approach is tested against the standard SMC algorithm and we find that our approach improve the standard SMC algorithm.