Human motion analysis

Abstract

This project aims to analyze human motion in an image sequence of swimming. The proposed system consists of three main modules. First, a background subtraction module segments the human region in spite of large scene variations in the monitored pool areas. Second, the analysis module obtains the skeletonization of the segmented region of human. Finally, a tracking module is used to acquire the trajectories of the body parts. The background subtraction module employs a region-based background modeling approach that can differentiate the moving human subject from the nonstationary water surface. The skeletonization is obtained by locating the interested points such as hands, elbows and shoulders of the human. The tracking module utilizes the Kalman filter to find the position of the human body and the trajectories of the body parts. We test the human motion analysis system using video clips of breaststroke swimming. The experimental results are very good.