The Risk Classification of Ergonomic Musculoskeletal Disorders in Work-related Repetitive Manual Handling Operations with Deep Learning Approaches

Abstract

The injury resulted from the repetitive and load-bearing works is the most frequent work-related musculoskeletal disorders (WMSD) or cumulative trauma disorders (CTD). It comes from the overload of repetitive load-bearing actions, which resulting in fatigue, inflammation, even injuries of musculoskeletal system. According to the annular report of Labor Insurance Bureau in Taiwan, WMSD is up to 85-88% payment. Thus, the aim of this study is to evaluate the risk of WMSD during work by using the simple, quick, and correct methods by using the deep learning algorithms. In the proposed research method, after collection the videos of hand repeated movements, the ergonomic injuries are evaluated by using the 2D human pose estimation method, which is based on the Key Indicator Method - Manual Handling Operations (KIM-MHO). Then, a model of predefined classifications through deep learning approaches for manual handling operating tasks is built. The analysis results show that the classification accuracy is more than 80%, compared with the doctor's judgment. The goal of this study is to get the accuracy up to 90%, so as to achieve fast and accurate assistance for deciding the risk of ergonomics, and immediately give proper feedback.