EEG Signal Processing for Semantic Feature Extraction: A Study on Perception, Imagination, and Vividness

The extraction of semantic features from EEG signals presents a novel approach to understanding the neural mechanisms underlying perception, imagination, and vividness. This study examines the correlation between EEG-based brain activity and subjective imagery ratings derived from the Vividness of Visual Imagery Questionnaire (VVIQ) and the Bucknell Auditory Imagery Scale (BAIS-V). We used a dataset containing recordings of 15 subjects subjected to audio, pictorial and orthographic stimuli. Before proceeding with feature extraction, the signal was appropriately preprocessed. As regards feature extraction, Power Spectral Density (PSD) and frequency bands were used. In this way we ensure that the neural representations that we consider are significant. To classify perception and imagination states, we implemented machine learning models—Support Vector Machines (SVM), Random Forest (RF), Logistic Regression (LR), and Multilayer Perceptron (MLP)—and evaluated their performance using accuracy, area under the curve (AUC), and cross-validation techniques. In the end, to reveal the influence of self-reported vividness scores on classification accuracy, statistical analysis were performed. The findings indicate that EEG-based features have a great potential for modeling semantic cognition, offering valuable insights into the neural basis of mental imagery. This research contributes to advancing EEG-based brain-computer interface (BCI) applications and cognitive neuroscience studies on imagination and perception.