Human body tracking using EM based on a 2-D articulated body model

Visual tracking of human body movement is a key technology in a number of areas, such as visual surveillance and monitoring. We present a 2-D model-based method of human body tracking from a monocular video sequence. Morris and Rehg put forward a 2-D scaled prismatic model(SPM) for figure registration which has far fewer singularity problems than 3-D models. Here we extend it in a 2-D cardboard human body model with additional one DOF of width change. Based on this modified 2-D model rather than 3-D model in Bregler and Malik's work, we also set up a mixture motion model for body movements and then solve motion parameters of the articulated body using EM in a statistical framework, where the model-based kinematic constraints are incorporated in a linear form. Tracking results from real video sequences are encouraging.