We are interested in finding low-dimensional representations of complex human motions. For example, each frame of the 3-D motion sequence shown to the right can be described by a 36-dimensional vector (12 joints, each with X,Y,Z values).

We have been developing computational techniques for reducing the dimensionality of this data into a lower and perceptually meaningful space. In this example, we are able to find a one-dimensional representation that corresponds to the inherent one-dimensional nature of the motion.

(Collaborative work with Fred Henle and Bruce Donald)