Author(s): OZAN, S., FOUREL, L., TOUZARD, P., DEGHAIES, K., MARTIN, C., KULPA, R., Institution: UNIVERSITY OF RENNES2, Country: FRANCE, Abstract-ID: 2277

INTRODUCTION:
To perform an effective tennis serve, it is essential to hit the ball from a high point. This provides a margin above the net and makes it easier to reach wide angles [1]. To achieve this, the centre of mass (CoM) must be elevated by maximising the angular velocities of the lower limb joints (knee and ankle) during the leg drive [2]. State of the art experiments usually assess these parameters using a 3D motion capture (MoCap) system based on markers placed on players skin. This tool is considered as the most accurate but is not suitable for in-situ assessment. Markerless MoCap, a new MoCap system based on videos and artificial intelligence, offers the possibility of analysing motion in ecological situations [3]. The aim of this study is to compare those kinematic variables using both MoCap methods.
METHODS:
Eight tennis players equipped with 49 markers placed on anatomical landmarks, performed 42 serves (336 serves in total). Markers positions were recorded at 300 Hz using an infrared MoCap system and processed to compute joint centre positions, segmental orientations, and CoM position using a musculoskeletal software (CusToM) [4]. Markerless MoCap was simultaneously recorded using 10 synchronized video cameras at 300Hz and the same parameters was computed from a commercial software (Theia3d) [3]. Ankle and knee angles between ball toss and impact, CoM height at impact, and peak angular velocities of the rear ankle and knee during the leg drive were computed from both systems. Validity of markerless MoCap for joint angles was assessed with respect to marker-based MoCap using Pearson’s correlation coefficient (CC) and root-mean-square difference (RMSD), while differences in peak velocities and CoM height at impact were assessed using CC and Bland-Altman bias and limit of agreement (LoA) [5].
RESULTS:
RMSD of ankle and knee angles are 5.1° and 8.8° with CCs of 0.99 (p<0.05) and 0.99 (p<0.05) respectively. In markerless MoCap, CoM height at impact is lower by 22.2mm on average with LoA values of [-32.9, -11.4]. In addition, ankle plantar flexion velocity peak is lower by 19.7°/s on average and knee extension velocity peak is greater by 14.8°/s on average with LoA values of [-175.6, 136.3] and [-104.9, 134.5], respectively. CC means of ankle plantar flexion and knee extension velocities are 0.93 (p<0.05) and 0.97 (p<0.05), respectively.
CONCLUSION:
This study highlights the potential of markerless MoCap as a valid alternative to marker-based systems for assessing lower limbs key variables in tennis serve performance. Some discrepancies remain, particularly in joint angle velocity peaks. Further investigations are still required to validate the markerless method on the upper limbs joints kinematics.
References:
[1] Elliott et al., 2009.
[2] Reid et al., 2008.
[3] Kanko et al., 2021.
[4] Muller et al., 2019.
[5] Bland & Altman 1986.