Physics-Based Person Tracking Using Simplified Lower-Body Dynamics

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, 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 model dynamics comprises a two degree-offreedom representation of human locomotion with inelastic ground contact. A stochastic controller generates impulsive forces during the toe-off stage of walking and spring-like forces between the legs. A higher-dimensional kinematic observation model is then conditioned on the underlying dynamics. We use the model for tracking walking people from video, including examples with turning, occlusion, and varying gait.

Physical simulation for probabilistic motion tracking

“…In particular, many of the produced results lack plausible physical realism and often violate the constraints imposed on the body by the world (resulting in out-of-plane rotations and foot skate). Such artifacts can be attributed to the general lack of physically plausible priors [2] (that can account for static and/or dynamic balance and ground-person-object interactions) which provide an untapped and very rich source of information.…”

“…In contrast, we incorporate full body human dynamical simulation into a Particle Filter, suited for multi-modal posteriors that commonly arise from ambiguities in monocular imagery. More recently, Brubaker at el [2] introduced a low-dimensional biomechanically-inspired model that accounts for human lower-body walking dynamics. The low-dimensional nature of the model [2] facilated the tractable inference; however, the model, while powerful, is inherently limited to walking motions in 2D.…”

Estimating Human Dynamics On-the-fly Using Monocular Video For Pose Estimation

“…The Anthropomorphic Walker proposed and pioneered by McGeer [19], and later modified by Kuo [17] by introducing impulsive push-off combined with torsional spring between the legs permitted the generation of a range of stable walking [6] for details regarding the dynamics model. motion on level and slanted ground.…”

“…Brubaker et al [6] introduced the notion of a physics-based biomechanical prior-model characterizing lower-body dynamics. A key benefit is the natural accommodation of variations in style due to changes in speed, step-length, and mass leading to 3D person tracking with physically plausible poses.…”

Staying Well Grounded in Markerless Motion Capture

“…In [3] further prior knowledge about light sources and shadows is applied to exploit more information about the scene observed by the cameras and to introduce virtual shadow cameras. A physically motivated model for tracking the lower body has been proposed in [6]. Moreover, it is also possible to impose fixed joint angle limits, as suggested in [7].…”