Classifying Cognitive Workload Based on Brain Waves Signal in the Arithmetic Tasks' Study

Cognitive workload is a quantitative usage measure of the limited amount of working memory. Its measuring is of great importance for understanding human mental effort processing, evaluating information systems or supporting diagnosis and treatment of patients. The paper presents the results of cognitive workload classification of electroencephalographic (EEG) data. The performed study covered arithmetic tasks realised in several intervals with the increasing difficulty level. Brain waves data in the form of EEG signal were gathered and processed in the form of frequency spectra. The paper discusses the process of features selection performed with several methods including ranking methods (K-Fisher), Feature Selection By Eigenvector Centrality (ECFS) and Mitinffs mutual information-based approach. What is more, the paper presents results of participant cognitive workload classification based on such methods as Support Vector Machines (SVM), boosted trees and k-nearest neighbours (KNN) algorithm. The paper discusses the efficiency of features selection methods and accuracy of applied classification methods.