METHODS OF DIGITAL EEG PROCESSING AND MACHINE LEARNING FOR ASSESSING HUMAN COGNITIVE STATES
TL;DR
Experimental results demonstrate that the joint use of spectral characteristics and functional connectivity measures improves classification accuracy compared to the use of spectral features alone and highlights their potential for further development of real-time cognitive monitoring systems.
Abstract
This paper presents methods of digital processing of electroencephalographic (EEG) signals and machine learning algorithms for assessing human cognitive states based on an open-access EEG dataset. Electroencephalography is a widely used non-invasive technique for studying brain functional activity due to its high temporal resolution. However, EEG signal analysis is challenged by low signal-to-noise ratio, inter-subject variability, and the presence of artifacts, which necessitates the use of advanced digital signal processing methods. The study employs an open EEG dataset recorded during the performance of cognitive tasks associated with creative thinking. The proposed processing pipeline includes band-pass and notch filtering, artifact suppression, spectral analysis in standard frequency bands, and functional connectivity analysis based on phase synchronization metrics. Spectral and connectivity features are combined to form an informative feature space for machine learning-based classification of cognitive states. Experimental results demonstrate that the joint use of spectral characteristics and functional connectivity measures improves classification accuracy compared to the use of spectral features alone. The obtained results confirm the effectiveness of digital EEG processing and machine learning methods for automated assessment of cognitive states and highlight their potential for further development of real-time cognitive monitoring systems.