A new hierarchical particle filtering for markerless human motion capture

Abstract

Particle ltering (also known as the condensation algorithm) has been widely applied to model-based human motion capture. However, the number of particles required for the algorithm to work increases exponentially with the dimensionality of the model. In order to alleviate this computational explosion, we propose a two-level hierarchical framework. At the coarse level, the conguration space is discretized into large partitions and a suboptimal estimation is calculated. At the ne level, new particles in the vicinity of the suboptimal estimation are created using a more likely and narrow conguration space, allowing the original coarse estimate to be rened more eciently.Our preliminary results demonstrates that this hierarchical framework achieves accurate estimation of the human posture with signicantly reduction in the number of particles.Keywords: coarse-to-ne, bottom-up aggregation of state estimations. we consider a model-based 3D human motion capture using particle lters.Particle ltering has proved to be an eective and robust technique for contour tracking [13,14,15].However, the application of particle ltering to mark- In this work, we address the problem of dimensionality and the associated computational complexity byproposing an ecient coarse-to-ne framework using monocular image sequences. The rest of this paper