Simple Machines for Scaling Human Motion

Abstract

This paper describes a fast technique for modifying humanmotion sequences in a way that preserves physical properties of the motion. Reference motion may be obtained from any source: motion capture data, keyframed motion, or procedurally generated motion. We show that by deriving a simplified control system from motion data only, we are able to modify the motion in a physically realistic way at nearly real-time speeds, because we can scale and modify the simplified system directly. To demonstrate the effectiveness of this approach, we animate running motion for a variety of characters over a range of velocities. Results can be computed at several frames per second.