Deep Learning for Sensor-Based Sport Performance and Health Monitoring: A Review of Wearable, Vision-Based, and...

Area of Science:

• Sports Science
• Artificial Intelligence
• Biomedical Engineering

Background:

• Wearable, vision-based, trajectory, physiological, and multimodal sensing technologies have advanced significantly.
• Deep learning models enable continuous, objective, and individualized assessment of athlete performance and health.
• Previous reviews often focused narrowly on single sensing modalities or specific applications.

Purpose of the Study:

• To integrate diverse sensing streams (wearable, vision, trajectory, physiological, multimodal) with deep learning for sports intelligence.
• To clarify relationships between sensing modalities, tasks, and deep learning models in performance analysis and health monitoring.
• To identify translational limitations and challenges in applying these technologies.

Main Methods:

• Synthesized recent progress in sensor-based sports intelligence.
• Reviewed applications transforming heterogeneous data streams into decision support for performance and health.
• Discussed suitability of various deep learning architectures (CNNs, LSTMs, Transformers, GNNs, etc.) for different data types.

Main Results:

• Covered applications including athlete/ball perception, tracking, pose estimation, action recognition, tactical analysis, load monitoring, injury prediction, and rehabilitation.
• Highlighted deep learning models' roles in processing visual, temporal, spatial, and physiological data.
• Identified key challenges: data heterogeneity, annotation scarcity, generalization, real-time deployment, interpretability, privacy, and ethics.

Conclusions:

• Future research should focus on standardized datasets, multimodal fusion, self-supervised/transfer learning, and edge/cloud deployment.
• Enhancing explainable AI (XAI) and implementing closed-loop, individualized monitoring systems are critical.
• This review provides a practical reference for optimizing athletic performance, injury prevention, rehabilitation, and long-term athlete health.