A study on the load coupling model of basketball athletes based on multi-source wearable sensor data fusion and unscented Kalman filter

Real-time and precise measuring of sports load is crucial to optimize training

Design: , avoid injuries, and enhance basketball player performance. Existing single-parameter monitoring

Methods: are not able to fully reconstruct the complex physiological and biomechanical loads in athletes. As an attempt to break the limitation, this paper proposes a novel model of load coupling for basketball players through multi-source wearable sensor integration. The model collects large amounts of data, including heart rate, acceleration, and angular velocity, across several body regions to construct an integrative representation of the workload imposed on the athlete. We employ the Unscented Kalman Filter (UKF) algorithm for combining data since it can handle the non-linear relationships between various load parameters appropriately. The model simplifies the overall load into three separate components, i.e., physiological load, upper-body biomechanical load, and lower-body biomechanical load. The adaptive noise covariance process is also added to improve the filter’s robustness during dynamic states. The new model was tested experimentally on undergraduate basketball players under different training conditions. The

Results: indicate that our model provides a more accurate and better estimation of sport load than traditional

methods and can differentiate the patterns of the load for every task effectively. The ablation study confirmed that the adaptive UKF greatly improved the stability of the estimates of the loads.