Development of exercise evaluation prototype system based on musculoskeletal analysis for motivation

Abstract

A daily training helps fall prevention for elderly people. We develop a prototype of training support system that shows users training effectiveness using muscle motion analysis to keep motivation. Our system is composed of low cost stereo camera and muscle-skeletal modeling software. The joint coordinates are obtained from stereo camera images by image recognition. The muscle force and joint moment are estimated using inverse kinematics and inverse dynamics. In our system the problem is instability of coordinate in general stereo camera. We propose the coordinate correction based on the human body structure and the human movement. In our method the energy function based on two constraint is optimized to correct coordinates. The simulated data is corrected by our proposed method. The corrected coordinate is similar to true coordinate and the standard deviation of error is 15.9mm.