Automatic Gait Recognition and Extraction

As much of the computer vision team's research interest now focusses on the extraction and description of moving shapes, our application domain has focused on recognising people by their gait. As with many applications of pattern recognition, these studies concern statistical recognition, and model-based recognition. In order to extract (moving) subjects, we have also developed techniques for locating moving perambulatory subjects.

The DARPA effort on Human ID at a distance is a sub-program of our gait research. The DARPA project is described elsewhere.

We have been fortunate to collaborate with Southampton's Dept. of Psychology on gait recognition. Also, as we are one of the few groups in this new area (as we were once in automatic face recognition), we were pleased to be invited to provide the survey chapter on gait in the world's first book on biometrics together with other surveys. A concurrent advantage of gait is that it provides an excellent vehicle for testing our new motion-based evidence gathering techniques, as demonstrated in much of our research.

If you would like to try out some of our data, we have provided some sequences which can be found from the downloads page