Time-Wheel based Multi-Series Sequence Partitioning for Athlete Pain Detection Modeling

Athlete Pain detection is a crucial research area in sports science, with significant implications for improving performance, preventing injuries, and optimizing training plans. With the continuous development of Electronic Health Records (EHRs) technology, athlete health management and pain detection increasingly rely on multivariate time series data. In this paper, we propose a novel pain detection method for athletes, called Time-Wheel based Multi-Series Time-aware Toeplitz Inverse Covariance-based Clustering (TWM-TICC). The method models the athlete’s health status and pain progression by partitioning physiological data into subsequences and applying dynamic clustering to identify different stages of pain. TWM-TICC incorporates the multi-series nature and irregular time intervals of EHRs, allowing it to effectively capture pain patterns that may emerge during training. We first validate the method through experiments using real-world athlete physiological dataset from our university’s affiliated hospitals. The results show that TWM-TICC perform well in the accuracy and interpretability of pain detection.