Pose estimation for pickleball players' kinematic analysis through MediaPipe-based deep learning: A pilot study.

Abstract

Pickleball has gained popularity across diverse age groups. This sport has particular balls that require different hitting styles, like hitting dinks. This study focuses on introducing pickleball players' kinematic analysis through a MediaPipe-based deep learning (DL) tool and analyzes the dominant leg's femur angle, knee angle, and wrist motion of pickleball players during the hitting dink shots, comparing pickleball players with high-level and beginner levels. Fourteen male pickleball players (aged 46.5 ± 10.5) participated in performing a dink shot during warm-up while being recorded by a GoPro camera and analysed by the DL tool. Statistical analysis, including T-tests and One-way ANOVA, showed significant differences between athletes and non-athletes in femur angle during the dink shot (p < 0.001), where high-level athletes demonstrated more femur flexion. Knee angles did not differ significantly, but advanced athletes maintained continuous wrist motion after the ball hit (p < 0.001). The MediaPipe-based DL tool estimated joint angles and motion patterns, offering an approximate alternative to visual analysis by coaches. With the developed DL tool, the coaches and players can rapidly monitor kinematics parameters and identify improvement areas. Future studies should further investigate foot positioning and trunk rotations in different shot types to assess pickleball biomechanical behaviours.