Résumé
Musculoskeletal disorders (MSDs) are a major concern in workplace ergonomics, often stemming from poor posture or the handling of heavy loads. We study binary detection of load carriage state from full-body IMU data collected using the Movella Xsens system from 15 participants, during walking and load-transfer tasks. We compare interpretable, threshold-based rules using wrist and elbow flexion–extension angles with supervised models trained on biomechanical features. For combined walking and transfer tasks, a Random Forest achieves 70.6% accuracy, demonstrating that IMU data can facilitate real-time risk screening for load-related ergonomic assessments.