Detecting Fatigue Status of Pilots Based on Deep Learning Network Using EEG Signals

Abstract

This article presents a solution for fatigue recognition through a new deep learning model that has a characteristic input of the power spectrum of an electroencephalogram (EEG) signal. First, four rhythms are obtained through the designed FIR filters, and the curve areas of their power spectrum density are coupled into four fatigue indicators. Second, a deep sparse contractive autoencoder network is proposed to learn more local fatigue characteristics, and the recognition results of pilots mental fatigue status are given. Compared with the state-of-the-art models, the results show that our model has good learning performance in extracting local features and fatigue status detection.