Dynamical Simulation Priors for Human Motion Tracking

Abstract

Abstract-We propose a simulation-based dynamical motion prior for tracking human motion from video in presence of physical ground-person interactions. Most tracking approaches to date have focused on efficient inference algorithms and/or learning of prior kinematic motion models; however, few can explicitly account for physical plausibility of recovered motion. Here, we aim to recover physically plausible motion of a single articulated human subject. Towards this end, we propose a full-body 3D physical simulationbased prior that explicitly incorporates a model of human dynamics into the Bayesian filtering framework. We consider the motion of the subject to be generated by a feedback "control loop" in which Newtonian physics approximates the rigid-body motion dynamics of the human and the environment through the application and integration of interaction forces, motor forces and gravity. Interaction forces prevent physically impossible hypotheses, enable more appropriate reactions to the environment (e.g., ground contacts) and are produced from detected human-environment collisions. Motor forces actuate the body, ensure that proposed pose transitions are physically feasible and are generated using a motion controller. For efficient inference in the resulting high-dimensional state space, we utilize an exemplar-based control strategy that reduces the effective search space of motor forces. As a result, we are able to recover physically-plausible motion of human subjects from monocular and multi-view video. We show, both quantitatively and qualitatively, that our approach performs favorably with respect to Bayesian filtering methods with standard motion priors.

Resolving 3D Human Pose Ambiguities With 3D Scene Constraints

“…Most prior methods that have used world constraints focus on interaction with a ground plane [69] or simply constrain the body to move along the ground plane [74]. Most interesting among these is the work of Vondrak et al [69] where they exploit a game physics engine to infer human pose using gravity, motor forces, and interactions with the ground. This is a very complicated optimization and it has not been extended beyond ground contact.…”

3D Human motion tracking by exemplar-based conditional particle filter

“…In [30], a dynamical simulation prior was proposed based on the truth of physical ground-person interaction. This prior is able to constraint human motion in the range of …”

Posture reconstruction using Kinect with a probabilistic model

“…[Taylor et al 2010] introduced a probabilistic latent variable model called the Implicit Mixture of Conditional Restricted Boltzmann Machines for human pose tracking. [Vondrak et al 2013] proposed to use a simulation-based dynamical motion prior for human motion tracking. However, these dynamic based approaches are computational costly and hence not applicable for Kinect based interactive applications.…”