EEG-based fatigue detection using binary pattern analysis and KNN algorithm

Fatigue is a prevalent issue that disrupts the overall well-being of individuals, leading to impaired cognitive functions such as learning, thinking, reasoning, remembering, and problem-solving. Chronic fatigue significantly increases the risk of accidents due to reduced focus, vigilance, and delayed reaction times. Traditional self-assessment methods for detecting fatigue are subjective and often unreliable. Recent advancements in neuroimaging have demonstrated that EEG signal analysis can objectively classify an individual's mental state. This research aims to develop a reliable and accurate EEG signal fatigue detection system. The EEG signals are decomposed into four levels using a one-dimensional discrete wavelet transform (1D-DWT). Textural features are extracted using binary pattern (BP) analysis and combined with seven statistical features. Then, these features are fed into a k-nearest neighbors (KNN) classifier to distinguish between the rest and fatigue states. Utilizing a dataset from the Mendeley Data website, the proposed system achieved an accuracy of 93.75%, precision between 93% and 95%, recall ranging from 92% to 95%, and an F1-score of 93% to 94%. This study highlights the potential of EEG-based systems to provide objective and accurate assessments of fatigue levels, thereby reducing the risks associated with chronic fatigue in daily life.