A Kinect based Golf Swing Score and Grade System using GMM and SVM

Abstract

This study displays a method to recognize and segment the time-sequential postures of golf swing. Golf is a kind of sports that has been adopted by most universities for physical education in Taiwan. The correct posture of golf swing is the most important skill while training a golfer. The training of golf swing is tedious, ineffective, and time-consuming in the traditional university course, because the golf instructor has to correct the swing postures of every student one by one in a class with 50 students. It's crucial to develop a system that can effectively recognize the steps of golf swing and facilitate self-learning of correct golf swing. First, a game controller, Kinect, is used to capture the 3D skeleton coordination of a golfer while performing swing. Second, the time-sequential postures of golf swing represented by Kinect's skeleton coordination are then transformed into a symbol sequence through vector quantization. Third, This system use serial correlation GMM model and GMM-KL divergence kernel to score and recognize grade of Golf Swing. Evaluate result show our method achieve a good accuracy of score and recognize grade and significantly improves than the Standard GMM and other SVM kernel.