Entropy and Complexity Assisted EEG-based Mental Workload Assessment System

As the era of Brain-Computer Interfacing (BCI) arrives, computationally measuring human mental workload via Electroencephalography (EEG) signal has become a crucial research field. Conventionally, mental workload assessment studies are mainly based on time-statistics, frequency, and wavelet domain features. In this paper, we present a mental workload assessment system in discriminating high and low mental workload by extracting EEG features from two new domains: time-complexity and entropy domains features. According to statistical analysis, the result demonstrates that the Frontal and Frontal-Central are two dominating regions. In addition, by fusing the traditional and new features, we boosted the classification performance from 69% to 88%. It indicates time-complexity and entropy domain features are able to extract some non-linear characteristics of EEG, which could not be achieved by traditional approaches. We conclude that the new features are feasible to assess human mental workload, and could provide complementary information to traditional features.