ABSTRACT
Cognitive load level identification is an interesting challenge in the field of brain-computer-interface. The sole objective of this work is to classify different cognitive load levels from multichannel electroencephalogram (EEG) which is computationally though-provoking task. This proposed work utilized discrete wavelet transform (DWT) to decompose the EEG signal for extracting the non-stationary features of task-wise EEG signals. Furthermore, a support vector machine (SVM) implemented to classify the task from the DWT-based extracted features. . The proposed methodology has been implemented on a renowned EEG dataset that captured three levels of cognitive load from the n-back test. In this work, two different approaches: i) Low vs High cognitive load (0-back vs [2-back+3-back]) and ii) Low vs Medium vs High (0-back vs 2-back vs 3-back) are investigated for the performance measurement. The linear SVM achieved the highest average classification accuracy that is 77.20 ± 6.63 and 87.89 ± 7.3 for 3-class and 2-class approaches, respectively.
Acknowledgments
The authors would like to acknowledge the contribution of Dr. Md. Asadur Rahman, Assistant Professor, Dept. of Biomedical Engineering, MIST, Dhaka, Bangladesh for his guidelines to improve the quality of this work.
Disclosure statement
No potential conflict of interest was reported by the authors.
Ethical Approval
The authors of this study used a renowned dataset (available in: https://www.nature.com/articles/sdata20183) that was conducted obeying the Helsinki declaration and also approved by the Ethics Committee of the Institute of Psychology and Ergonomics, Berlin Institute of Technology (approval number: SH_01_20150330).