UCLA

This thesis presents the design of an on-bed rehabilitative exercise guidance and monitoring system that utilizes a high density sensor bedsheet to sense certain type and range of motion exercises. I propose and develop a new method to recognize rehabilitation actions and quantize body moving angles through analyzing the

progression of pressure images. I also present a way to construct 3D models of patients from 2D pressure images. The 3D character model can not only reflect the rehabilitation exercises, but also the visualization synchronizes with the actions of patients. Our experimental results on five on-bed exercises show that this system

can accurately recognize patient's exercising type and track compliance of patients to prescribed treatment programs. The system is easy to setup and is suitable for long term monitoring.