Analyzing and recognizing walking figures in XYT

Abstract
We describe a novel algorithm for gait analysis. A person walking frontoparallel to the image plane generates a characteristic "braided" pattern in a spatiotemporal (XYT) volume. Our algorithm detects this pattern, and fits it with a set of spatiotemporal snakes. The snakes can be used to find the bounding contours of the walker. The contours vary over time in a manner characteristic of each walker. Individual gaits can be recognized by applying standard pattern recognition techniques to the contour signals.

This publication has 6 references indexed in Scilit: