Vision-based techniques to estimate 3D human pose from single camera are important in many applications like entertainment and games industries, sports science, gait analysis and video surveillance. While de-facto optical-based motion capture system is highly accurate in estimating 3D human pose, vision-based techniques offer better alternative because the latter are far cheaper and non-intrusive. In this paper, we review recent significant advances made in single-camera 3D human pose estimation techniques. We discuss about challenges faced, typical image observations used, ways to address the challenges and highlighting limitations in state-of-the-art. We conclude this paper with speculation of future research directions as well as open research problems.
展开▼
