Exploring the Ability to Classify Visual Perception and Visual Imagery EEG Data: Toward an Intuitive BCI System

Providing an intuitive interface for the actual use of brain–computer (BCI) can increase BCI users’ convenience greatly. We explored possibility that visual imagery be used as a paradigm may constitute more intuitive, active BCI. To do so, electroencephalography (EEG) data were collected during perception and experiments. Three image categories (object, digit, shape) three different images per category stimuli. EEG from seven subjects in this work. types perception/imagery preprocessed classification: raw time series data; time–frequency maps; common spatial pattern (CSP). Five classifiers (EEGNet, 1D convolutional neural network (CNN), MultiRocket, MobileNet, support vector machine (SVM)) applied to each applicable type among types. Thus, we investigated feasibility classifying three-category or nine-class over various found MultiRocket showed best classification performance: yielding approximately 57.02% (max 63.62%) 46.43% 71.38%) accuracy imagery. However, no meaningfully improved performance was achieved either imagery, although yielded slightly higher than From our extensive investigation, classified; however, it is somewhat doubtful whether system. It believed introducing better-designed advanced deep learning networks together with informative feature extractions improve classifications. In addition, sophisticated experimental design enhance potential achieve