Multi-level data fusion enables collaborative dynamics analysis in team sports using wearable sensor networks

Abstract

Abstract This research proposes a novel multi-level data fusion method for analyzing collaborative dynamics in team sports using wearable sensor networks. We developed and validated this approach through controlled experiments with 40 semi-professional athletes across basketball and soccer scenarios. The multi-level fusion architecture integrates IMU, GPS, physiological, and positioning data through adaptive weight allocation and asynchronous alignment algorithms. Experimental validation demonstrated 8.6 dB improvement in signal quality and 42.3% enhancement in positional accuracy compared to single-source approaches. Cross-sport testing across basketball, soccer, volleyball, and handball showed consistent performance (84.2–91.4% accuracy) with real-time response times of 192-312ms. The developed collaborative dynamics indicator system revealed that temporal coordination parameters strongly correlate with team performance (r = 0.73), while four key metrics predict match outcomes with 73.6% accuracy. This methodology provides coaches and analysts with objective tools for quantifying previously subjective aspects of team coordination.