A video-based assessment tool using machine learning for ergonomic risk prediction in manual lifting tasks

Manual lifting tasks are commonplace in industries such as logistics, healthcare, and manufacturing, which can potentially lead to significant ergonomic risks, including musculoskeletal disorders. This study proposes a video-based ergonomic risk assessment tool that integrates MediaPipe Pose Landmarker for joint coordinate tracking, classification and regression tree models for lifting stage classification, and a back-propagation neural network for parameter compensation. Utilising the revised NIOSH lifting equation, our proposed tool calculates the recommended weight limit and lifting index to quantify ergonomic risks from smartphone videos. Validated on laboratory and field datasets, the tool demonstrates adaptability, scalability, and the potential to provide cost-effective ergonomic assessments.