Prediction of Occupational Physical Activities using Inertial Measurement Units and Deep Learning Models

Abstract

Many workers engaged in manual material handling (MMH) jobs experience high physical demands that are associated with work-related musculoskeletal disorders (WMSD). Quantifying the physical demands of a job is an important legal requirement in the US that is used by human resources in the job hiring process. Most physical demands analysis (PDA) are performed using observational and semi-quantitative methods. The lack of accuracy and reliability of these methods can create problems when assigning acceptable tasks to an injured worker. In this study, various deep learning models were applied to data from eight inertial measurement units (IMUs) to predict 15 occupational physical activities (OPA). Overall, a 95% accuracy was reached by convolutional neural network (CNN) for predicting occupational physical activities when performed in isolation. More work is needed to estimate the accuracy of the model when OPA elements are combined into a more complex task.