Motion Filtering with Dynamics
This paper describes a technique for filtering and modifying human data using dynamic constraints and simulation. A simple control system calculates torques to track human motion data for a dynamic simulation of an articulated figure. The simulation generates smooth, physically plausible motion that maintains characteristics of the original. This technique is used to filter data for a variety of upper-body motions, animating two models with differing kinematic parameters and degrees of freedom. In addition, the system creates transitions from one motion sequence to another in the spirit of previous techniques. Further, the system imposes task and environmental constraints to generate believable behaviors and dynamic contacts.