3-D model-based tracking of humans in action: a multi-view approach