Construction of an Intelligent Tennis Coach Based on Kinect and a Sensor-Based Tennis Racket

Abstract

The purpose of this paper is to develop an artificial intelligence coach system which can be used to provide beginners a more economical way to learn tennis and to achieve as similar as the coach to guide the training. To reach the goal, artificial intelligence coach system is composed of three phases that are (1) Tennis swing motion data collection, (2) CAST-based swing motion group construction, and (3) Online tennis swing motion analysis. In the first phase, the tennis motions are collected using the Kinect and the built sensor-based tennis racket. Then, the cluster affinity search technique (CAST) and the dynamic time warping (DTW) are utilized to divide the collected swing motion series data into groups to form swing motion groups. In the third phase, using the swing motion groups, the system can provide possible tennis motion improvement suggestions. At last, experiments were also conducted on the real dataset to show the effectiveness of the proposed system.