Physics-Based Person Tracking Using the Anthropomorphic Walker

Abstract

We introduce a physics-based model for 3D person tracking. Based on a biomechanical characterization of lower-body dynamics, the model captures important physical properties of bipedal locomotion such as balance and ground contact. The model generalizes naturally to variations in style due to changes in speed, step-length, and mass, and avoids common problems (such as footskate) that arise with existing trackers. The dynamics comprise a two degreeof-freedom representation of human locomotion with inelastic ground contact. A stochastic controller generates impulsive forces during the toe-off stage of walking, and springlike forces between the legs. A higher-dimensional kinematic body model is conditioned on the underlying dynamics. The combined model is used to track walking people in video, including examples with turning, occlusion, and varying gait. We also report quantitative monocular and binocular tracking results with the HumanEva dataset.

Detecting end-effectors on 2.5D data using geometric deformable models: Application to human pose estimation

“…In their work, trajectories of segmented body parts are mapped on linear subspaces to model the global body movement. Brubaker et al [16] use a simple lower-body model based on physical walking movement called Antropomorphic Walker, proposed by Kuo [17]. Hasler et al [18] propose a pose estimation algorithm which performs on mono and multiple uncalibrated cameras.…”

Stick It! Articulated Tracking Using Spatial Rigid Object Priors

“…Most work on improving the predictive system, is focused on learning motion specific priors, such as for walking [6][7][8][9][10][11][12]. Currently, the most popular approach is to restrict the tracker to some subspace of the joint angle space [7][8][9][10]13].…”

“…In general, as above, the environment is usually not incorporated in the tracking models. One notable environmental exception seems to be the ground plane [6,17]. Yamamoto and Yagishita [17] use a linear approximation of the motion path by linearising the forward kinematics function.…”

“…Promising results are, however, shown on constrained situations, such as when the position and orientation of a persons feet is known. Brubaker et al [6] explicitly model the ground plane in a biomechanical model of walking. Their approach is, however, limited to interaction with the ground while walking.…”

Metric Regression Forests for Human Pose Estimation

“…Generative approaches model the likelihood of the observations given a pose estimate. The pose is typically inferred using local optimization [4,5,13,22,27] or stochastic search [8,10,21]. Regardless of the optimization scheme used, such approaches are susceptible to local minima and thus require good initial pose estimates.…”