Cardboard People: A Parameterized Model of Articulated Image Motion

In this paper we extend the work of Black and Yacoob [5] on tracking and recognition of human facial expressions to the problem of tracking and recognizing the articulated motion of human limbs. We make the assumption that a person can be represented by a set of connected planar patches: the cardboard person model illustrated in Figure 1. In the case of faces, Black and Yacoob [5] showed that a planar model could well approximate the motion of a human head and that it provides a concise description of the optical flow within a region. This motion can be estimated robustly and it can be used for recognition.