Recognition of the fatigue status of pilots using BF–PSO optimized multi-class GP classification with sEMG signals

Air crashes caused by human factors pose a problem. Many researchers have focused on aviation human factors and found that pilots’ fatigue status is the key factor. In this study, a hybrid multi-class Gaussian process model is proposed to identify the fatigue status of pilots by analyzing the surface electromyogram signals on the back of their neck and upper arm muscles. Instead of using the traditional conjugate gradient technique to determine the optimal parameters, a hybrid bacterial foraging and particle swarm method is proposed to optimize the unknown parameters to improve the classification accuracy of the multi-class Gaussian process. In the proposed method, the entropy-based features are extracted by wavelet translation from the collected signals to estimate the fatigue status of pilots. Experiments are performed through flight simulation in a full-flight simulator to provide three situations for the fatigue level of the subjects. Comparison of experimental results validates the feasibility of the proposed method to identify the fatigue status of pilots and the further enhancements by the proposed classification system in terms of classification accuracy. Results also show that the developed method helps prevent air crashes caused by pilots’ fatigue.