Marker based human pose estimation using annealed particle swarm optimization with search space partitioning

In this paper, a marker based human pose estimation from multi-view video sequences is presented. The pose estimation problem is defined as optimization of the 45 parameters which define body pose model and is solved using particle swarm optimization (PSO). The objective of this optimization is to maximize a fitness function which formulates how much body model matches with 2D marker´s coordinate in video frames. In this algorithm a sampling covariance matrix is used in the first part of the velocity equation of PSO that is annealed with iterations. One of the major problems of this algorithm is the high number of parameters that define the pose of the body model. To tackle this problem, we divide the optimization into six stages that exploit the hierarchical structure of the model. The first stage optimizes the six parameters that define the global orientation and position of the body. Other stages are related to optimization of right and left hand, right and left leg and head orientation, respectively. Experimental results on PEAR¹ dataset [1] indicate that the proposed algorithm can achieve lower estimation error in tracking human motion compared with Annealed Particle Filter (APF) and Standard Particle Filter (PF).