Prediction of Gold Standard Gait Data from Inertial Data: A Machine Learning Approach

Le résumé fourni par la source

Biomechanics laboratories typically use kinematic cameras and force platforms as the gold standard for gait assessment. However, these systems are expensive and have limited availability. Wearable devices, equipped with sensors—primarily inertial sensors—can capture movement data, enabling the inference of human gait behavior. To enhance the quality of measurements obtained from wearables, this study investigates the feasibility of predicting kinematic parameters from inertial data collected by wearable sensors. Machine learning techniques, including Random Forest, XGBoost, and Gradient Boosting, were used to correlate inertial measurements with data from traditional motion capture systems. Feature importance analysis and SHAP highlighted the significance of velocity and acceleration in predicting kinematic parameters. Experimental results indicate that tree-based models, particularly Gradient Boosting and XGBoost, achieved the best performance, with coefficient of determination values close to 0.989, demonstrating the feasibility of the proposed approach.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.