A New Strategy for Mental Fatigue Detection Based on Deep Learning and Respiratory Signal

Mental fatigue is often associated with decreased mental alertness and worsening performances. But its detection method is still a difficult issue due to the contradiction between practicability and accuracy. In the current study, we attempt to provide a new method for mental fatigue detection to realize the unity of practicability and accuracy based on deep learning and respiratory signal. To this end, respiratory signals were collected and two deep learning models, convolutional neural network (CNN) and Long Short Term Memory (LSTM), were constructed. Wavelet scale maps and time series of respiratory signal were as input to CNN and LSTM respectively. The data set was divided into training set, verification set and test set according to the ratio of 6:2:2. Bayesian optimization was used for hyperparametric optimization of CNN and LSTM. The results showed that LSTM model has a better performance of test accuracy of 89.16% than CNN that of 77.29%. Our findings indicated that respiratory signal combined with LSTM classifier may be a practical and effective strategy for mental fatigue detection.