A Novel Estimation Method of Fatigue Using EEG Based on KPCA-SVM and Complexity Parameters

In this study, Kolmogorov complexity (KC) and approximate Entropy (AE) were adopted to characterize the irregularity and complexity of EEG data. Fifty subjects were instructed to perform two different mental tasks to induce two kinds of fatigue states. Then the Kernel Principal Component Analysis (KPCA) and Support Vector Machine (SVM) are combined to differentiate these two states. The KPCA was used to extract nonlinear features from the complexity parameters of EEG and to effectively reduce the dimensionality of the feature vectors. SVM was used to classify two fatigue states. The experimental result shows that complexity parameters are significantly decreased as the fatigue level increases, which suggests that the proposed parameters can be used to characterize mental fatigue level. Furthermore, compared with several typical classification models, the joint method KPCA-SVM can achieve higher classification accuracy (85%) of mental fatigue with less training and classifying time, which indicates that KPCA-SVM is suitable for the estimation of mental fatigue.