Expressive features for movement exaggeration

Given a single motion-capture sequence of a person performing a dynamic activity at a particular intensity (or effort), our goal is to automatically warp that movement into a natural-looking exaggerated version of that action. Consider warping a movement of a person lifting a lightweight box to make the movement appear as if the box were actually very heavy. We describe an efficient data-driven approach applicable to animation re-use that learns the underlying regularity in an action to select the most "expressive" features to exaggerate. Other "style-based" approaches are presented in [Gleicher 1998; Brand and Hertzmann 2000; Vasilescu 2001].

[1]  Michael Gleicher,et al.  Retargetting motion to new characters , 1998, SIGGRAPH.

[2]  Aaron Hertzmann,et al.  Style machines , 2000, SIGGRAPH 2000.

[3]  M. Alex O. Vasilescu Human motion signatures for character animation , 2001, SIGGRAPH 2001.