EEG based Mental Workload Assessment by Power Spectral Density Feature

In the field of cognitive neuroscience, mental workload assessment plays an important role. In this work, the power spectral density (PSD) feature of Electroencephalogram (EEG) signals is extracted based on spectrum analysis, and the problems of medium-level and high-level mental workload identification are studied. The classification accuracy of spectral features of each frequency band is evaluated by using AdaBoost, Decision Tree (DT), KNN and support vector machine (SVM). In addition, the features are selected according to the change of relative PSD of each frequency band. The results show that the classification accuracy of the data after feature selection can reach 76.62%, which has been improved with different levels in almost classifier than original data.