Mental Fatigue Estimation Based on Multichannel Linear Descriptors and Support Vector Machine with Optimizing Parameters

In this paper, three parameters of multichannel linear descriptors i.e. Omega, Phi and Sigma are used to measure the level of mental fatigue for the first time. The multichannel linear descriptors of electroencephalogram (EEG) from electrode arrays (Fpl, Fz, and Fp2) are extracted as the features of brain activity in different mental fatigue state. Then support vector machine (SVM) with optimizing parameters is used to identify two mental fatigue states. For comparison, linear discriminant analysis based on mahalanobis distance (MDBC) is also used to identify such two mental fatigue states. The experimental results show that SVM classifier with optimizing parameters is much suitable for our study, and it achieves the average maximum classification accuracy of 91.67%. The investigation also suggests that the values of Omega, Phi and Sigma of EEG strongly correlate with mental fatigue, which can effectively distinguish two mental fatigue states, so that they are expected to serve as the indexes to evaluate mental fatigue level objectively.