UC Berkeley

Work-related musculoskeletal disorders (WMSDs) represent a significant challenge in occupational health, with risks increasing when the physical demands of a job surpass an employee's capabilities. Accurate Physical Demands Analysis (PDA) is essential for developing job descriptions, pre-employment screening, conducting ergonomic assessments, and supporting return-to-work programs. PDAs require identifying Occupational Physical Activities (OPAs) and quantifying physical exposures. Traditionally, OPAs have been identified through self-reports, observation and direct measurements. While effective, these methods can be less reliable, labor-intensive, and limited in dynamic work settings. Emerging technologies, such as wearable devices, provide promising alternatives by offering automated and scalable solutions. However, research applying these methods to recognize OPAs in complex work tasks remains limited. This study utilized inertial measurement units (IMUs) and deep learning techniques to identify thirteen distinct OPAs, including seven whole body and six upper body activities, both in isolation and within three simulated work tasks designed to replicate real-world conditions. Building on OPA identification, the study focused on quantifying physical exposures critical for ergonomic evaluation, particularly in manual lifting. IMUs were used to measure lift origins and destinations, reducing intrusiveness and enabling the capture of more lifting events. These findings establish reliable methods for both quantifying OPAs and assessing physical exposures, with significant potential for improving workplace safety and reducing the risk of WMSDs.