Patrick Blauberger

As pose estimation has garnered considerable attention for kinematic analysis, numerous pose estimation models have been developed in recent times. A pose estimation model is a trained neural network that predicts human body landmarks from an image. Each model contains different strong and weak points, which make it difficult for users to decide which model to use for kinematic analysis. The accuracy of the model can be one big factor for model selection, but there are not many studies investigating this critical point. Therefore, this study aims to investigate the accuracy of different models and variants by comparing the measurements from the models and variants against reference measurements. Five male participants were invited to this study. Each participant was asked to perform five exercises: squat, squat jump, counter movement jump, walk, and jog while being recorded by twelve normal RGB cameras (Contemplas) and ten marker-based tracking cameras (VICON). The video recordings from the Contemplas were processed by six different pose estimation models and variants: Mediapipe, MeTRAbs Small, MeTRAbs X Large, YOLO, MoveNet Lightning, and MoveNet Thunder to detect joint positions. From the detected joint positions, four joint angles, left hip, right hip, left knee, and right knee, were calculated. Three-way repeated measures ANOVA and Tukey HSD post-hoc analysis were applied to compare the pose estimation models with VICON measurements. The ANOVA result showed that exercise and model factors had a significant impact on the measurement errors although angle factor did not. In the post-hoc analysis, knee joint angle errors from YOLO, MoveNet Lightning, and MoveNet Thunder in jog and walk were significantly higher than those from Mediapipe, MeTRAbs Small, and MeTRAbs X Large. In conclusion, differentiated recommendations can be given for optimum model and variant choice in different conditions in kinematic analyses.

This study compared the kinematic demands of FIFA’s Electronic Performance and Tracking Systems (EPTS) validation circuit with an alternative sport-specific circuit (SSC). The aim was to determine which circuit presents high-intensity demands known to challenge EPTS accuracy. Four amateur soccer players (age: 29.4 ± 5.9 years) performed both circuits using global positioning system (GPS). Speed, acceleration and change of direction (COD) was divided into zones and compared between the circuits. Results showed that the SSC resulted in greater distance, time and entrances in speed zones 3 and 4, as well as in moderate and high acceleration/deceleration zones. SSC also exhibited more time and entries in high-speed COD zones. FIFA’s circuit demonstrated higher peak speed and more time in speed zone 5. Differences in COD between circuits reflect the intensity and type of exercises performed, with the SSC including more sprints and rapid direction changes. In conclusion, the SSC presents more kinematic demands in high-intensity situations critical for EPTS validation, whereas the FIFA circuit demonstrates less challenging movement patterns. FIFA could improve the validation protocol by updating the circuit to include more demanding sport-specific movements thus increasing the evaluated devices validity in high-intensity situations which are of very high interest for sports.

Thanks to the advancement of computer vision technology and knowledge, the accuracy of human pose estimation has improved to the level that can be used for motion capture. Especially, human pose estimation has been gaining attention in research due to its efficiency and accuracy. The traditional motion capture system is not accessible to everyone. Human pose estimation could be a solution to replace the traditional system. However, the validity of human pose estimation has not been investigated enough yet in athletic and sports contexts. For this reason, this research aims to validate the kinematic measurements of human pose estimation by comparing them against the measurement of marker-based motion capture system. Five participants were recruited and asked to perform eight athletic and nine sports movements, respectively while being captured by normal and infrared cameras. Human pose estimation was run on frames from the RGB cameras to estimate human landmarks. From estimated landmarks in human pose estimation and marker-based motion capture system, elbow, shoulder, hip, and knee joint angles on the left and right sides were calculated and compared. Mean absolute error was used to evaluate the accuracy of human pose estimation measurements. The mean errors for athletic and sports movements were 9.7 ± 4.7 degrees and 9.0 ± 3.3 degrees, respectively. Errors were generally largest for elbow joint angles. The errors might be due to occlusion and systematic differences between human pose estimation and marker-based motion capture system. In conclusion, human pose estimation contains room for improvement, but has the potential to be used in some applications in which strictly precise measurements are not required.

This study aimed to analyze the effect of laterality and instructional video on the soccer goalkeepers’ dive kinematics in penalty. Eight goalkeepers from youth categories (U15, U17, U20) were randomly divided into control (CG) and video instruction groups (VG). The latter performed 20 penalty defense trials on the field with balls launched by a machine, ten before and after watching a video instruction to improve the diving kinematics. The CG only performed the dives. Three cameras recorded the collections. A markerless motion capture technique (OpenPose) was used for identification and tracking of joints and anatomical references on video. The pose data were used for 3D reconstruction. In the post-instruction situation, the VG presented differences in comparison to the CG in the: knee flexion/extension angle, time to reach peak resultant velocity, frontal step distance, and frontal departure angle, which generated greater acceleration during the dive. Non-dominant leg side dives had higher resultant velocity during 88.4 – 100% of the diving cycle, different knee flexion/extension angle, and higher values ​​in the frontal step distance. The instructional video generated an acute change in the diving movement pattern of young goalkeepers when comparing the control and the video instruction group in the post condition.

This study describes a method for extracting the stride parameter ground contact time (GCT) from inertial sensor signals in sprinting. Five elite athletes were equipped with inertial measurement units (IMU) on their ankles and performed 34 maximum 50 and 100-m sprints. The GCT of each step was estimated based on features of the recorded IMU signals. Additionally, a photo-electric measurement system covered a 50-m corridor of the track to generate ground truth data. This corridor was placed interchangeably at the first and the last 50-ms of the track. In total, 863 of 889 steps (97.08%) were detected correctly. On average, ground truth data were underestimated by 3.55 ms. The root mean square error of GCT was 7.97 ms. Error analyses showed that GCT at the beginning and the end of the sprint was classified with smaller errors. For single runs the visualization of step-by-step GCT was demonstrated as a new diagnostic instrument for sprint running. The results show the high potential of IMUs to provide the temporal parameter GCT for elite-level athletes.

American football is an appealing field of research for the use of information technology. While much effort is made to analyze the offensive team in recent years, reasoning about defensive behavior is an emergent topic. As defensive performance and positioning largely contribute to the overall success of the whole team, this study introduces a method to simulate defensive trajectories. The simulation is evaluated by comparing the movements in individual plays to a simulated league average behavior. A data-driven ghosting approach is proposed. Deep neural networks are trained with a multi-agent imitation learning approach, using the tracking data of players of a whole National Football League (NFL) regular season. To evaluate the quality of the predicted movements, a formation-based pass completion probability model is introduced. With the implementation of a learnable order invariant model, based on insights of molecular dynamical machine learning, the accuracy of the model is increased to 81%. The trained pass completion probability model is used to evaluate the ghosted trajectories and serves as a metric to compare the true trajectory to the ghosted ones. Additionally, the study evaluates the ghosting approach with respect to different optimization methods and dataset augmentation. It is shown that a multi-agent imitation learning approach trained with a dataset aggregation method outperforms baseline approaches on the dataset. This network and evaluation scheme presents a new method for teams, sports analysts, and sports scientists to evaluate defensive plays in American football and lays the foundation for more sophisticated data-driven simulation methods.

The aim of this study was the validation of player and ball position measurements of Kinexon’s local positioning system (LPS) in handball and football. Eight athletes conducted a sport-specific course (SSC) and small sided football games (SSG), simultaneously tracked by the LPS and an infrared camera-based motion capture system as reference system. Furthermore, football shots and handball throws were performed to evaluate ball tracking. The position root mean square error (RMSE) for player tracking was 9 cm for SSCs, the instantaneous peak speed showed a percentage deviation from the reference system of 0.7–1.7% for different exercises. The RMSE for SSGs was 8 cm. Covered distance was overestimated by 0.6% in SSCs and 1.0% in SSGs. The 2D RMSE of ball tracking was 15 cm in SSGs, 3D position errors of shot and throw impact locations were 17 cm and 21 cm. The methodology for the validation of a system’s accuracy in sports tracking requires extensive attention, especially in settings covering both, player and ball measurements. Most tracking errors for player tracking were smaller or in line with errors found for comparable systems in the literature. Ball tracking showed a larger error than player tracking. Here, the influence of the positioning of the sensor must be further reviewed. In total, the accuracy of Kinexon’s LPS has proven to represent the current state of the art for player and ball position detection in team sports.