I'm a graduate student in the EECS department working with Prof. Pentland and his students in Vismod, a group in the MIT Media Laboratory. I am working on a project to aid human motion tracking through the marriage of dynamic and behavioral models with machine vision. Knowledge of the constraints on human motion should increase the computer's ability to track and interpret people.

Hopefully the result will be better vision-based Human/Computer interfaces. Eventually better interfaces could lead to many less-invasive, natural, human uses for computers.

The first step toward that lofty goal was the creation of a program that can track a person, and their various parts, in real-time. For more details see the tech report describing the system.

I have a keen interest in the human perceptual system, and have studied biological perception and behavior. Over the past several years I seem to have become well versed in the fine art of computer vision and decision. I'm especially interested in using real-world constraints to solve computer vision problems: especially the analysis of human motion.

Recent work has focused on the problem of modeling the contraints on human motion. The physical limitations of the body, the patterns encoded into our physiological structures, and even the habits of motion that we acquire over time, all combine to provide strong contraints on how we move. An article in the Proceedings of FG'98 , discusses progress in this direction. The idea in a nutshell: perception is mediated by expectation.

In Vismod we don't just build vision technology. We also explore novel human-computer interface designs that use vision and other sensing technologies. An article in the AAI Journal summarizes five years of experience in building these interfaces.

Go back to the Top