Head Impact Detection Using Machine Learning Algorithms

Numerous research studies have emphasized the significance of accurately detecting head impacts and implementing safety measures. This study addresses this crucial need by utilizing machine learning algorithms applied to data from piezoelectric sensors on a simulated head model. Employing a systematic approach, this work utilizes Random Forest (RF) and Extreme Gradient Boosting (XGBoost) models to process the normalized sensor data, aiming to pinpoint impact locations with high precision. Through rigorous k-fold cross-validation and comprehensive performance analysis, the study reveals that the XGBoost model slightly outperforms the RF model, achieving an RMSE of 0.4764 and an R2 of 0.9485. Feature importance evaluations suggest an optimal sensor placement strategy, potentially reducing the model complexity while retaining predictive accuracy. The superior performance of the XGBoost model, combined with strategic sensor placement, highlights the study's contribution to enhancing head impact safety measures in sports and industrial settings. The findings pave the way for future research into the deployment of intelligent safety systems, leveraging the synergy between wearable technology and machine learning.