Related Experiment Video
Updated: Jun 27, 2026
06:52
An Inertial Measurement Unit Based Method to Estimate Hip and Knee Joint Kinematics in Team Sport Athletes on the Field
Published on: May 26, 2020
Comparative Analysis of Markerless Motion-Capture Models for Assessing Football Kinematics During 30 m Long-Pass
Donghao Wang1, Junkai Yu2, Shiqin Chen3
1School of Physical Education, Shanghai University of Sport, Shanghai 200438, China.
Sensors (Basel, Switzerland)
|June 26, 2026
Summary
This study compared MediaPipe Pose, DWPose, and YOLO-Pose against Xsens for football long-pass analysis. While not precise enough for high-accuracy field measurements, MediaPipe Pose and DWPose show potential for screening movement patterns.
Area of Science:
- Sports Science
- Biomechanics
- Human Movement Analysis
Background:
- Accurate kinematic analysis is crucial for understanding and improving athletic performance.
- Mobile-based motion capture systems offer a potentially accessible alternative to marker-based systems.
Purpose of the Study:
- To evaluate the agreement and practical utility of MediaPipe Pose, DWPose, and YOLO-Pose against Xsens for analyzing football long-passes.
- To determine the suitability of these models for real-field kinematic data collection.
Main Methods:
- Twelve elite male football players performed 30m long-passes.
- Data collected simultaneously using Xsens inertial sensors and smartphones running MediaPipe Pose, DWPose, and YOLO-Pose.
- Statistical analysis included Intraclass Correlation Coefficient (ICC), Bland-Altman, Root Mean Square Error (RMSE), Coefficient of Multiple Determination (CMD), Statistical Parametric Mapping (SPM), and Statistical non-Parametric Mapping (SnPM).
Main Results:
- Coefficient of Multiple Determination (CMD) ranged from 0.13 ± 0.17 to 0.67 ± 0.25.
- Root Mean Square Error (RMSE) ranged from 9.88 ± 8.20° to 39.92 ± 10.44°.
- Significant kinematic differences were observed primarily in the hip, knee, and ankle joints, particularly with SPM and SnPM analysis.
Conclusions:
- Current mobile-based pose estimation models (MediaPipe Pose, DWPose, YOLO-Pose) are not yet suitable for high-precision field-based kinematic measurements in football.
- MediaPipe Pose and DWPose demonstrate potential for rapid screening of movement patterns and analyzing general movement trends in football techniques.