Real-time physics-based motion capture with sparse sensors

Abstract

We propose a framework for real-time tracking of humans using sparse multi-modal sensor sets, including data obtained from optical markers and inertial measurement units. A small number of sensors leaves the performer unencumbered by not requiring dense coverage of the body. An inverse dynamics solver and physics-based body model are used, ensuring physical plausibility by computing joint torques and contact forces. A prior model is also used to give an improved estimate of motion of internal joints. The behaviour of our tracker is evaluated using several black box motion priors. We show that our system can track and simulate a wide range of dynamic movements including bipedal gait, ballistic movements such as jumping, and interaction with the environment. The reconstructed motion has low error and appears natural. As both the internal forces and contacts are obtained with high credibility, it is also useful for human movement analysis.