Pilots' Fatigue Status Recognition Using Deep Contractive Autoencoder Network

The evaluation of pilots' fatigue status is of substantial significance in aviation safety, which faces two major issues. They are how to get the fatigue status feature representation and how to identify the fatigue behavior status of pilots via electroencephalogram (EEG) signals. To solve the first issue, we propose a novel fatigue evaluation index via different window functions to compute the power spectrum of relative rhythms from EEG signals. Wavelet packet transform is used to decompose EEG signals from pilots to form four major rhythms, i.e., delta wave, theta wave, alpha wave, and beta wave, and the combined representation of their power spectrum curve area is the features of pilots' mental status. To solve the second issue, we propose a new deep contractive autoencoder (AE) network with a softmax (SM) classifier to detect the multistatuses of mental fatigue workload. The recognition results of our model are also compared with that of other models such as the deep AE network with a SM classifier model. The experimental results show that our deep learning model has superior classification performance, and the recognition accuracy of fatigue mental status is up to 91.67%, which shows that the proposed method performs excellently compared with the state-of-the-art methods.