Combination of Annealing Particle Filter and Belief Propagation for 3D Upper Body Tracking

Abstract

Abstract. 3D upper body pose estimation is a topic greatly studied by the computer vision society because it is useful in a great number of applications, mainly for human robots interactions including communications with companion robots. However there is a challenging problem: the complexity of classical algorithms that increases exponentially with the dimension of the vectors' state becomes too difficult to handle. To tackle this problem, we propose a new approach that combines several annealing particle filters defined independently for each limb and belief propagation method to add geometrical constraints between individual filters. Experimental results on a real human gestures sequence will show that this combined approach leads to reliable results.

Distributed Estimation Techniques for Cyber-Physical Systems: A Systematic Review

“…On the contrary, those approaches based on Kalman filters usually require a lot of information, as in [43], or consensus iterations between estimation steps, as in [39,50], with the consequent communication effort. A similar drawback appears in [38], where a lot of information must be sent between the particles of the filter between two consecutive sampling instants.…”

“…Mainly they can be divided into the following groups:Adaptive observer, used in [46] for multi-agent systems, in [48] for leader–follower systems and in [55] for heterogeneous multi-agent system.Bayesian filter, also used in many areas, such as collaborative human–robot systems [36], joint attack detection and secure state estimation [51] or 3D upper body tracking, with a combination of annealing particle filter and belief propagation inference [38]. H∞ filter, which has been applied for the detection of biasing attacks on distributed estimation networks [45] and for the joint attack detection and secure state estimation [52].…”

“…Bayesian filter, also used in many areas, such as collaborative human–robot systems [36], joint attack detection and secure state estimation [51] or 3D upper body tracking, with a combination of annealing particle filter and belief propagation inference [38].…”

“…Other applications: In this group, three particular studies not related to the previous groups were included. In [38], the authors propose a new approach for 3D upper body pose estimation, using a combination of an annealing particle filter and belief propagation inference. The work [43] is concerned with the distributed estimation problem, adopting a Kalman-like filter, for industrial automation over relay assisted wireless sensor networks.…”