A machine learning approach for detecting fatigue during repetitive physical tasks

Abstract Prolonged and repetitive stress on muscles, tendons, ligaments, nerves can have long-term adverse effects the human body. This be exasperated while working if environment nature of tasks puts significant strain body, which may lead to work-related musculoskeletal disorders (WMSDs). Workers with WMSDs experience generalized pain, loss muscle strength, ability continue working. Most injuries are caused by ergonomic risks, such as physical movements, awkward postures, inadequate recovery time, muscular stress. Fatigue seen a detector risk, accumulation fatigue significantly increase possibility injury. Thirty participants completed series over six-hour period wearing sensors capture data related heart rate movement, external embedded captured ground reaction hand exertion force. They also provided subjective ratings at start end experiment. Classifiers for (high vs low) were constructed using three methods: linear discriminant analysis (LDA), k-nearest neighbor (kNN), polynomial kernel-based SVM (P-SVM) validated tenfold cross-validation technique that was repeated hundred times. Results our supervised machine learning approach demonstrated maximum accuracy 94.15% P-SVM binary classification fatigue.