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

Area of Science:

• Sports Science
• Data Science
• Wearable Technology

Background:

• Analyzing collaborative dynamics in team sports is crucial for performance enhancement.
• Existing methods often lack objectivity and rely on subjective assessments.
• Wearable sensor networks offer potential for objective data collection in sports.

Purpose of the Study:

• To propose and validate a novel multi-level data fusion method for analyzing collaborative dynamics in team sports.
• To integrate data from various wearable sensors (IMU, GPS, physiological, positioning) for enhanced analysis.
• To develop an objective system for quantifying team coordination and predicting match outcomes.

Main Methods:

• Developed a multi-level data fusion architecture with adaptive weight allocation and asynchronous alignment.
• Validated the method using controlled experiments with semi-professional athletes in basketball and soccer.
• Conducted cross-sport testing on basketball, soccer, volleyball, and handball, measuring accuracy and response times.

Main Results:

• Achieved 8.6 dB improvement in signal quality and 42.3% enhancement in positional accuracy compared to single-source methods.
• Demonstrated consistent performance across sports (84.2-91.4% accuracy) with real-time response (192-312ms).
• Found strong correlation between temporal coordination and team performance (r=0.73); four metrics predicted outcomes with 73.6% accuracy.

Conclusions:

• The multi-level data fusion method effectively quantifies team collaborative dynamics using wearable sensor networks.
• This approach provides objective tools for coaches and analysts to assess team coordination and performance.
• The findings highlight the importance of temporal coordination in team sports success and enable data-driven insights.