Detection of Ball Spin Direction using Hitting Sound in Tennis

Naoki Yamamoto, Kenji Nishida, Katsutoshi Itoyama, Kazuhiro Nakadai, Kazuhiro Nakadai


This paper describes the detection of rotation direction using the hitting sound of tennis balls. Since each ball rotation direction has a slightly different rotation direction and trajectory, there should be a difference in the hitting sound. To distinguish the characteristics of ball rotation direction, a database was constructed that combines the hitting sound recorded experimentally with ball rotation direction. Since it is difficult to distinguish audible differences in hitting sounds by ear, it is necessary to identify them using measuring instruments. For this purpose, after extracting the amplitude spectrum by fast Fourier transform of the shot sound, the entire data was normalized and classified by a support vector machine. As a result of evaluating this method, a high accuracy was obtained in identifying the sound associated with slice among other hit sounds. The proposed method also evaluated the ball hit sound from a YouTube video in an unknown environment and achieved a perfectly correct identification of spin and slice.


Paper Citation