Classification of mental states from rational dilation wavelet transform and bagged tree classifier using EEG signals

Abstract

The automation in production, aviation, autopilots, and healthcare industries transformed the work from man to machine. Such automation requires an artificial intelligence system (AIS) for the safe and productive working of machinery. These AIS-based models must study the physiological and behavioral changes in humans accurately to develop brain-computer interface (BCI) systems. The human mental states are one such example that can be used for the development of BCI systems. Electroencephalogram (EEG) signal-based analysis offers a perfect platform for the development of such systems as it provides a noninvasive and cost-effective solution. However, the nonlinear and complex nature of EEG signals makes their analysis and classification difficult. These EEG signals are required to split into multibands for proper analysis and classification. Therefore, an automatic mental states classification using rational dilation wavelet transform and ensemble tree classifier-based chapter is proposed. EEG signals of three states namely focused, unfocused, and drowsiness are decomposed into filter banks of low pass and high pass subbands. Multiple features are extracted and most discriminant features are selected using analysis of variance test. The selected features are fed as input to the three ensemble classifiers namely bagged tree, boosted tree, and random undersampling boosted tree. The average accuracy of 91.77% and the highest classification accuracy of 94.71%, 93.97%, and 86.62% are achieved with bagged tree classifier for focused, unfocused, and drowsiness states, respectively. © 2022 Elsevier Inc. All rights reserved.