Abstract
In recent years, a user interface based on human body action is adopted in some entertainment systems such as fitness-oriented video games. In order to obtain the body action from a single camera, we propose a method for estimating 3D pose of a person in a video sequence. The method expresses the human body by a skeleton model and estimates its positions and joint angles using the particle filter algorithm. Experiment results indicate that the proposed method can successively estimate the 3D pose at a processing speed of 15 fps on a laptop PC with a USB camera.
