1. School of Military Biomedical Engineering, Air Force Medical University, Xi’an Shaanxi 710032, China;
2. Department of Medical Engineering, The 987th Hospital of Joint Logistics Support Force, Baoji Shaanxi 721000, China
Abstract：Objective To explore whether the method of multiscale entropy (MSE) combined with support vector machine (SVM)
can effectively detect mental fatigue, and then to compare the detecting performance of electroencephalogram (EEG) electrodes at
different cerebral cortex locations. Methods We established the mental fatigue model by continuous cognitive load task, and used a
portable device to collect the EEG signals of 12 subjects in the state of waking and mental fatigue, then classified the two states by the
MSE and SVM algorithms. Results After the continuous cognitive load task, the fatigue level of the subjects increased significantly,
and the NASA-TLX and KSS scales showed significant statistical differences (P<0.01). The average classification accuracy of the
two states in frontal lobe Fpz, parietal lobe Pz and occipital lobe Oz EEG electrodes were 92.16%, 81.63% and 90.54%, respectively.
There was no statistical difference between Fpz and Oz electrodes (P>0.05), while they showed significant statistical differences
when referenced to Pz electrode (P<0.05). Conclusion MSE combined with SVM can effectively detect the waking and mental
fatigue state of human body, and Fpz and Oz electrodes can better detect mental fatigue than Pz electrode.
