Vision Based Motion Estimation Method Using Ego-Exo Cameras

Abstract

Recently, vision based robust pose estimation method for human computer interaction has been developed by many researchers. It plays an important role in the development of interaction in consumer electronics. Existing pose estimation using monocular camera employed either ego-motion or exo-motion which are not acceptable for fine motion (e.g. a breath, wrist motion, and etc.). In this paper, we propose a hybrid vision method for fine motion estimation. For this method, we integrated ego-camera and exo-camera coordinates. We expect that the proposed method can provide a practical solution for robust natural interaction with digital information display systems and interactive serious game or remote rehabilitation care systems.