A Data-Driven Model to Identify Fatigue Level Based on the Motion Data from a Smartphone

Abstract

The fatigue due to repetitive and physically challenging jobs may result in workers poor performance and Work-related Musculoskeletal Disorder (WMSD). Thus, it is imperative to frequently monitor fatigue and take necessary recovery actions. Our purpose was to develop a methodology to objectively classify subjects fatigue level in the workplace utilizing the motion sensors embedded in the smartphones. An experiment consisting of twenty-four participants (12 M, 12 F) with a smartphone attached to their right shank was conducted using a fatiguing exercise (squatting), targeted mainly the lower extremity musculature. After each set of an exercise (2-min squatting), participants were asked about their ratings of perceived exertion (RPE), then a reference gait data were collected during a straight walk of 20-32 steps. This process was continued until they reported strong fatigue ([≥]17). Using the RPE to label the gait data, we have developed machine learning algorithms (i.e., binary and multi-class SVM models) to classify the individuals gait into two (no-vs. strong-fatigue) and four levels (no-, low-, medium-, and strong-fatigue). The models reached the accuracies of 91% and 61% for two and four-level classification, respectively. The outcomes of this study may facilitate the implementation of a proactive approach in continuous monitoring of operators fatigue level, which may subsequently increase the workers performance and reduce the risk of WMSDs.