Using Nonlinear Dimensionality Reduction in 3D Figure Animation
A. Elizabeth Seward and Bobby Bodenheimer
Abstract
This paper explores a method for re-sequencing an existing set of
animation, specifically motion capture data, to generate new motion.
Re-using animation is helpful in designing virtual environments and
creating video games for reasons of cost and efficiency. This paper
demonstrates that through nonlinear dimensionality reduction and frame
re-sequencing, visually compelling motion can be produced from a set of
motion capture data. The technique presented uses Isomap and
ST-Isomap to reduce the dimensionality of the data set. Two distance
metrics for nonlinear dimensionality reduction are compared as well as
the effect of global degrees of freedom on the visual appeal of the
newly generated motion.
Bobby Bodenheimer 
Last modified: Thu Feb 10 10:06:28 2000