AI-Enabled Wearable Sensor Technologies for Athlete Performance Monitoring: A Critical Review of Signal Processing, Machine Learning, and Real-Time Embedded Systems

Wearable sensing and artificial intelligence (AI) are increasingly being used to quantify athlete movement, physiological response, training load, and recovery in field and laboratory settings. However, the practical value of these systems depends not only on the predictive capability of an algorithm but also on sensor validity, signal quality, hardware constraints, real-time processing, and the transferability of models across athletes and sports. This review examines the technical pipeline linking wearable electronic sensors to athlete-monitoring decisions, with particular attention to inertial measurement units, global navigation satellite systems, photoplethysmography, electrocardiography, electromyography, pressure sensing, and multimodal systems. It synthesizes evidence on signal conditioning, segmentation, feature engineering, classical machine learning, deep learning, sensor fusion, edge inference, and embedded implementation. Existing reviews indicate that wearable devices can support useful monitoring, but accuracy varies by metric, device, movement pattern, and validation setting; performance under controlled conditions should not be assumed to generalize to competition. Key challenges include motion artifacts, inconsistent protocols, small and imbalanced datasets, leakage between training and test sets, limited external validation, battery and communication constraints, privacy, and unclear decision thresholds. The review proposes a layered reference architecture and a validation framework that links engineering performance to sport-specific usefulness. Future research should prioritize synchronized multimodal datasets, athlete-independent testing, interpretable and uncertainty-aware models, energy-efficient edge AI, secure data governance, and transparent reporting. The article’s contribution is an engineering-oriented synthesis that connects sensor physics and embedded constraints with the needs of physical education and sports science.