Machine vision based physical fitness measurement with human posture recognition and skeletal data smoothing

摘要

A machine vision based measurement system for physical fitness is designed and implemented. Compared with other existing systems, our system only utilizes one Kinect sensor without bulky wearable sensors, thus enabling testees limber and free. To improve the test accuracy, a series of skeletal data smoothing methods and posture recognition algorithms are developed or used. The tests among university students and experimental results show that the performance of our system is increased and it is comparable with human beings, and therefore more practical and labor-saving.