A data-driven 3D vision feedback system for bilateral rehabilitation application

The patients in bilateral rehabilitation move their paretic limbs to follow the trajectory defined by imparetic limbs. The robot provides power only when the patients need assistance. Capturing the motion posture and feeding back to patients through audio or visual modes can raise training enthusiasm and activeness and obtain better rehabilitation effect. A mini even free vision feedback system is on demand when rehabilitation trainings need to be implemented at home. In this study a vision feedback system is developed based on open source software. The original 3D character is generated by a piece of special software Makehuman and exported as several optional formats such as obj, mtl, bvh and so on. These exported files are re-loaded in a graph interface built with wxWidgets and rendered with OpenGL. The motion sensors are fixed on the patient's limbs and the motion data are sampled to update 3D model rendering. The patients' motion posture is shown on a screen in real time to prompt their rehabilitation training. The new vision feedback system presented in this paper does not depend on any commercial software or expensive hardware. As a fully open source system, it is very appropriate to realize more economical rehabilitation trainings.