Machine learning- and sensor-based posture recognition

Many workers in manual material handling jobs experience high physical demands associated with work-related musculoskeletal disorders. Musculoskeletal disorders have high incidence rates and are costly to employers due to absenteeism, lost productivity, increased health care, and worker’s compensation costs. Quantifying the physical demands of a job is essential for identifying high-risk jobs and returning to work progress following injury. Rather than utilizing traditional observation-based methods, which lack accuracy and reliability, we used wearable technologies (inertial measurement units and pressure insoles) and deep learning methods to recognize occupational, and physical activities. We aim to provide a more reliable and valid approach to quantifying physical exposures for performing risk assessments for manual material handling tasks.

 

If you are interested in the most recent research updates for the project, please feel free to contact Yishu Yan (yishuyan at berkeley dot edu)