Emotion from Motion

Abstract

This paper introduces a model to generate ''emotional'' animation from ''neutral'' human motion. Using techniques from signal processing, our method calculates certain emotional transforms which are then applied to existing motions of articulated figures in order to produce the same motions, but with an emotional quality such as angry or sad. These transforms capture the difference between a neutral and emotional movement with respect to two components: speed (timing), and spatial amplitude (range) of a movement. Since the transforms are applied as global operations, they provide a convenient and efficient way to adapt motion-captured, simulated or keyframed animation of articulated figures to different situations and characters.