Motion Data-Driven Unsafe Pose Identification through Biomechanical Analysis

About 33% of non-fatal occupational injuries and illness are due to work-related musculoskeletal disorders (WMSDs), which had the highest percentage of injuries or illnesses in construction in 2009. Though techniques previously used to prevent WMSDs (e.g., ergonomic rules, checklists based on surveys, and laboratory experiments) provide valuable insights into the prevention of WMSDs, these techniques may not be suitable for measuring the physical demands required for manual ongoing works under real conditions. In an effort to address this issue, we propose a motion capture approach to obtain a worker's posture information for measuring physical loads on body parts (e.g., shoulder, back). The human postures extracted by motion capture are expressed by rotation angles between body joints, and these angles are then converted to joint angles which are the inputs for biomechanical analysis. The resulting information contains the internal forces of body parts (e.g., hands and feet) and is capable of identifying allowable strength ranges, thereby helping determine overexertion and awkward postures during a task. As a test case, motion data for ladder-climbing is obtained using a motion capture system (e.g., Microsoft Kinect) in a laboratory, and the postures are biomechanically analyzed frame by frame. The results show that the proposed method performs well at computing the physical loads, which promises its great potential to understand the causes of WMSDs during ongoing manual construction work. Further, it can automatically predict potentially hazardous postures on certain body parts, which has otherwise only been achieved by tedious and time-consuming manual investigation.