Detection of fatigue on gait using accelerometer data and supervised machine learning

In this paper, we aim to detect the fatigue based on accelerometer data from human gait using traditional classifiers from machine learning. First, we compare widely used machine learning classifiers to know which classifier can detect fatigue with the fewest errors. We observe that the best results were obtained with a Support Vector Machine (SVM) classifier. Later, we propose a new approach to solve the feature selection problem to know which features are more relevant to detect fatigue in healthy people based on their gait patterns. Finally, we used relevant gait features discovered in a previous step as input in classifiers used previously to know its impact on the classification process. Our results indicate that using only some gait features selected by our proposed feature selection method it is possible to improve fatigue detection based on data from human gait. We conclude that it is possible to distinguish between a normal gait person and a fatigued gait person with high accuracy.