Sequential Monte Carlo Tracking of Body Parameters in a Sub-Space

In recent years Sequential Monte Carlo (SMC) methods have been applied to handle some of the problems inherent to model-based tracking. In this paper two issues regarding SMC are investigated in the context of estimating the 3D pose of the human arm. Firstly, we investigate how to apply a sub-space to representing the pose of a human arm more efficiently, i.e., reducing the dimensionality. Secondly, we investigate how to apply a local method to estimated the maximum a posteriori (MAP). The former issue is based on combining a screw axis representation with the position of the hand in the image. The latter issue is handled by applying a method based on maximising a proximity function, to estimate the MAP. We find that both the sub-space and the proximity function are sound strategies and that they are an improvement over the current SMC-methods.