Automated recognition of construction workers’ physical fatigue based on foot plantar patterns captured from a wearable insole pressure system

Construction workers are exposed to numerous non-fatal occupational injuries (e.g., fall accidents, work-related musculoskeletal disorders) due to physically demanding activities such as repetitive lifting tasks. One of the key preventive measures to mitigate these occupational injuries among construction workers is by recognizing workers’ physical fatigue. However, previous approaches for recognizing workers’ fatigue are subjective, time-consuming, and based on localized muscle fatigue. Therefore, the objective of this study is to develop a non-invasive approach to recognize workers’ physical fatigue by capturing foot plantar patterns measured by a wearable insole pressure system after a fatiguing repetitive lifting task. The experimental protocol was designed to recruit construction workers to participate in this study by collecting their foot plantar patterns during normal gait and after a fatiguing repetitive lifting task. The performance accuracy was evaluated by adopting five types of supervised machine learning classifiers and different window sizes. The results showed that the Random Forest classifier obtained the best classification performance with an accuracy of 95.8% and sensitivity of 97.8% using a sliding window of 2.56s. The findings indicate that the proposed approach would provide useful ergonomic intervention guidelines for early detection of workers’ physical fatigue, and thus enable safety managers to mitigate non-fatal occupational injuries among construction workers.