A combinational deep learning approach for automated visual classification using EEG signals

According to the literature, electroencephalographic (EEG) data are among the most important physiological signals, and EEG classification has remained a major challenge. The classification of EEG signals that obtain from visual stimuli is the focus of this research. The goal is to accurately predict, from EEG signals, the categories of visual stimuli. In order to achieve this, we provide a novel technique that enables the classification of EEG signals from 40 different classes of visual stimuli. For EEG signal classification, we employed a novel combination of Residual network (ResNet) and long short-term memory network. In order to feed inputs into a 2D convolutional network, we also used wavelet transform to convert signals into images. Our suggested network has an average classification accuracy of 99.85% for training data, 98.75% for validation data, and 98.7% for test data. Finally, we investigated different areas of the brain as well as separate EEG frequency bands for cognitive investigations. When compared to previous studies, our framework performs exceptionally well in the image-based EEG classification analysis.