Classification of mental workload from beta of EEG data using Stochastic Configuration Network (SCN)

With the advancement of technology, crews of humanmachine systems tend to be at higher mental workload (MW) level, which is more likely to cause accidents. Therefore, monitoring the MW of operators in special industries is very important. In this paper, three levels of MW are set up in Multi Attribute Task Battery II (MATB). To reduce the impact of the order on different MW levels, the experiment was designed with Latin square. The data of Task performance, NASA-TLX questionnaires and electroencephalograph (EEG) of 16 subjects were collected. The scores of NASA-TLX questionnaires for three MW levels had significant differences, which demonstrated that the three tasks of different MW levels were designed successfully. We extracted EEG spectral power in beta frequency band of 16 subjects as the input features for the SCN classifier. The highest classifier accuracy was 98.86%, the lowest was 77.57%, and the average accuracy was 93.30%, which showed a good classification performance. Therefore, the research proposes that the beta rhythm may be the most relevant to the MW of pilots during flight missions. This conclusion makes it possible to simplify the operation of MW classification, and provides the possibility for real-time monitoring and predicting MW in the future.