An embedded 6-axis sensor based recognition for tennis stroke

In recent years, sensors are widely used in outdoor activities for returning motion information to participants. This paper proposes a product that could send back information about the tennis stroke while playing tennis in real time. This product, consisting of sensor module, controller module and transmission module, is embedded in tennis handle. There are three steps for stroke recognition. Shot detection as the first step in this process uses fluctuation of acceleration based on moving windows. Secondly, data on windows that detected as shots are used to divide those shots into 3 stroke types (forehand, backhand, serve) according to acceleration as well as angular velocity. At last, more information about how the ball rotates would be acquired based on the angular velocity. This step would divide forehand or backhand into topspin or backspin. Different from widespread wearable sensors, this product is embedded in the racket handle without attachment to body and the information of motion would be sent to our mobile phone in real time. The experimental result shows high accuracy for motion recognition with 98% accuracy for shot detection and 96% accuracy for stroke types recognition.