Abstract
Mental fatigue is complex disorganization that affects the human being's efficiency in work and daily activities (e.g., driving, exercising). Encephalography is routinely used to discern this fatigue. Several automatic procedures have deployed conventional approaches to support neurologists in mental fatigue detection episodes (e.g., sleepy vs. normal). In all of the traditional procedures (e.g., support vector machine, discrimination fisher, K-nearest neighbor, and Bayesian classification), only a low accuracy is achieved when a binary classification task (e.g., tired vs. normal) is applied. The convolutional neural network model identifies the correct mathematical manipulation to turn the input into the output. In this study, a convolutional neural network is trained to recognize brain signals recorded by a wearable encephalographic cap. Unfortunately, the convolutional neural network works with large datasets. To overcome this problem, an augmentation scheme for a convolutional neural network model is essential because it can achieve higher accuracy than the traditional classifiers. The results show that our model achieved 97.3% compared to the state-of-the-art traditional methods (e.g., SVM and LDA).