A NON-FEEDBACK EEG-BASED BRAIN COMPUTER INTERFACE (BCI) FOR COGNITIVE STATE CLASSIFICATION IN TIME DOMAIN

This paper introduces an innovative EEG-based Brain-Computer Interface (BCI), aiming to discern two cognitive states experienced by students during learning sessions. Focusing on “Relaxation” and “Engagement in learning tasks”, the study identifies attentive students and students exhibiting disengagement. Utilizing single EEG channel and signals from the fronto-polar region, it aims to develop a real-time engagement detection system compatible with portable devices. Employing a basic machine learning pipeline, the research focuses on time-domain feature extraction and capturing heterogeneous high-level features. Through feature analysis, selection, and support vector machine(SVM) classification, the BCI system differentiates between relaxed and learning states, achieving 60.36% accuracy with 10-fold cross-validation. The subject-wise analysis yields impressive results, reaching up to 93% accuracy. Despite challenges in EEG signal non-stationarity, the model’s accuracy underscores the efficacy of the time-domain parameters.