ABSTRACT
An electroencephalogram (EEG) measures neuronal activities. It records the brain’s electrical activity by placing metal electrodes over the scalp. EEG signals can be analyzed in mainly two ways: for the identification of neurological disorders like epilepsy, major depressive disorders and alcohol use disorders and for indirect medical applications like emotion recognition, sleep stage classification, eye state detection, and motor imagery. In the biomedical field, signal processing allows for obtaining information from EEG signals for diagnostic purposes. Different feature extraction and machine learning techniques have been compared with the pros and cons of EEG signal processing. A new technique of graph signal processing (GSP) for the special handling of multidimensional data has been found very effective in the analysis of EEG signals. Hence, its usage in EEG analysis has been discussed specially. GSP is now an obvious choice for many applications involving brain image processing and disorders. GSP provides tools for analyzing irregular domain data and graph networks. According to neuroscience, brain activity is connected to intricate functional and structural networks. GSP is used for handling such data with irregular structures, dynamic and non-Euclidean domains. Such applications include social networks, sensor feeds, online traffic, supply chains, and biological systems. Geometric deep learning is a field of research that extends deep learning to non-Euclidean data. To take into account the merits of GSP, graph convolutional neural network classifiers have been developed in some papers.
