Smart Occupational Health: A Machine Learning Approach to Ergonomic Hazard Identification using Body-Mounted Sensors

Abstract

This research is motivated by the need for vigorous identification and prevention of musculoskeletal injuries in occupations that involve physically demanding body movement including construction, repair and maintenance, freight handling and shipping, installation, and sports training. The common denominator amongst such occupations is that physical requirements of performed activities may at times exceed the natural bodily limits of human participants. We introduce a machine learning (ML) technique capable of determining ergonomic risk levels of human tasks with high accuracy based on activity duration and frequency information. The ML classifier is support vector machine (SVM) with a cubic kernel function, and is trained and tested with data captured by the built-in sensors of body-mounted mobile devices (i.e., smartphones). Given the ubiquity of smartphones, our approach offers an affordable, low-power, and easy to maintain, synchronize, and operate alternative to current practices of ergonomic hazard identification.