Estimation of workload using EEG data and classification using linear classifiers

Cognitive workload is a subjective term operantly defined as a worker’s perception of a work performance and work difficulty. To estimate workload through Electroencephalogram (EEG) requires good algorithm with best features. Objective of this research study was to estimate workload using linear classifiers on non linear data. Workload was presented by varying levels of Multi Attribute Task Battery II (MATB-II task). Two non linear features of Hurst exponent and Higuchi Fractal Dimension have been extracted from the data which was acquired from 28 subjects who were all male in the age group 25-40. The classification has been performed using three prominent classifiers i.e. K-Means, K Nearest Neighbor (KNN) and Support Vector Machine (SVM) to test their efficiency in the case of workload data. We have hypothesized SVM classifier to give best results out of the three classifiers. Comparing the performance accuracy of the selected classifiers, we propose a classifier that will give best results for workload classification