A Sensing Based Approach to Assessing and Predicting Musculoskeletal Symptoms of Work-From-Home Employees

Abstract

The shift to remote work during the COVID-19 pandemic has raised concerns about physical inactivity and musculoskeletal discomfort. This study proposes a sensing-based approach to assess and predict musculoskeletal symptoms among work-from-home employees. We collected heart rate (HR), metabolic equivalents (MET), and self-reported pain data across four time points from 36 participants. Using Temporal Convolutional Networks (TCN) and comparative models (Informer, LSTM, SVM, LR), we evaluated the relationship between physical activity and region-specific pain. TCN achieved the highest accuracy (97.86%) using combined MET and HR data. Our findings suggest that sensor-based monitoring paired with deep learning enables accurate, non-invasive pain prediction and can inform ergonomic interventions for remote workers.