Human Body Pose Recognition Using Spatio-Temporal Templates

We present a novel approach to detecting human silhouettes in monocular sequences that achieves very low rates of both false positives and negatives by combining shape and motion information. To this end, we use sequences of moving silhouettes built using motion capture data that we match against short image sequences. We demonstrate the effectiveness of our technique using both indoor and outdoor images of people walking in front of cluttered backgrounds and acquired with a moving camera, which makes techniques such as background subtraction impractical.