Related Experiment Videos
A multimodal spatio-temporal graph neural network framework for fatigue detection in tennis serving.
1College of Physical Education, Xuzhou University of Technology, Xuzhou, China.
Frontiers in Bioengineering and Biotechnology
|June 25, 2026
Summary
This study introduces a novel multimodal framework using IMUs and sEMG with ST-GCN for accurate tennis serve fatigue detection. The system achieves 95.2% accuracy, offering a feasible solution for on-court injury prevention.
Area of Science:
- Sports Science
- Biomechanics
- Machine Learning
Background:
- Tennis serves involve complex kinetic chain energy transfer, vulnerable to fatigue-induced injuries like rotator cuff tendinopathy.
- Existing wearable tech lacks neuromuscular insight, failing to detect compensatory mechanisms preceding performance decline.
Purpose of the Study:
- To develop a multimodal framework integrating IMUs and sEMG with ST-GCN for advanced fatigue detection in tennis players.
- To analyze neuromuscular and biomechanical alterations during fatigue in the tennis serve.
Main Methods:
- Recruited 15 tennis players for an endurance protocol involving serves until exhaustion.
- Utilized a Spatio-Temporal Graph Convolutional Network (ST-GCN) with skeletal graph topology and sEMG data.
- Employed paired t-tests and ANOVA for statistical significance and performance comparisons.
Main Results:
- The ST-GCN framework achieved 95.2% fatigue classification accuracy, outperforming LSTM, CNN, and ST-TR models (p < 0.001).
- Identified a significant "neuromechanical lag" (p < 0.01) and compensatory patterns, including reduced knee flexion (-12.4°).
Conclusions:
- The ST-GCN system offers a computationally efficient (8.4 ms latency) fatigue detection solution for tennis.
- This framework represents a feasible step towards on-court smart sports engineering and precision rehabilitation.
