BIOGRAPHY
My research addresses estimating and understanding motion in image sequences. For humans and other biological creatures, the perception of motion in the world seems effortless and is exploited for tasks as varied as locomotion and courting. Endowing computers with similar perceptual capabilities, however, has proven to be a challenge. Fundamentally, the problem involves inferring structure from a changing pattern of measurements acquired by a sensor (typically a sequence of images from a video camera). Problems exist at various levels of abstraction, from the measurement of two-dimensional image velocity (optical flow) to the interpretation or recognition of complex patterns of motion. Since the late 1980s motion estimation has improved greatly, thanks in part to robust statistical techniques, parameterized models of motion, learning methods, and layered image representations. At the same time, multimedia computers are putting video cameras in homes and offices. This creates an opportunity for research on motion estimation to make an impact on other problems and other disciplines. In particular, I have been working on estimating and understanding human motions including facial expressions and gestures. The hope is that this work will lead to new and better ways for humans and computers to interact and to a better understanding of how to represent and reason about complex temporal processes.
My path to motion research is somewhat complicated but began as an undergraduate majoring in psychology. I took an introductory computer science class partly because I thought it would be good for me, partly because it scared me, and partly because it was the only class that would fit in my schedule. It was Fortran programming on punched cards, I loved it, became a computer scientist, and never looked back. I did my graduate work at Stanford and Yale, I was a manager at Xerox PARC, and along the way, I've tried to combine my interests in humans and computation.
|
