Physiological Monitoring and Technical Diagnostic System for Track and Field Training Integrating Intelligent Sensing and Artificial Intelligence

Abstract

Real-time physiological monitoring and technical diagnosis in track and field training are often constrained by fragmented multimodal data acquisition, insufficient sensor synchronization, and limited intelligent analysis capabilities. To address these issues, this study proposes an integrated physiological monitoring and technical diagnostic system based on intelligent sensing and artificial intelligence. A distributed sensing architecture incorporating inertial measurement units (IMUs), electromyography (EMG), heart rate (HR), and plantar pressure sensors is established, where timestamp synchronization and edge computing enable millisecond-level alignment of heterogeneous signals. Sliding-window feature extraction, Long Short-Term Memory (LSTM) networks, and Spatiotemporal Graph Convolutional Networks (ST-GCNs) are jointly employed for physiological load estimation and technical action deviation diagnosis, while a hierarchical feedback mechanism provides real-time training guidance. Experimental results demonstrate an overall action recognition accuracy of 90.2%, an F1-score of 0.859, an average end-to-end latency of 40.6 ms, and a physiological load prediction highly correlated with blood lactate measurements (r > 0.96). The proposed framework establishes an efficient multimodal sensing and intelligent decision-making paradigm for wearable monitoring systems and provides valuable engineering references for electromagnetic sensing networks, wireless signal acquisition, antenna-enabled wearable devices, and intelligent information processing in next-generation human-centered monitoring applications.