ScholarWorks@숙명여자대학교

초록

As brain-computer interface (BCI) systems have gained prominence, the demand for automated interpretation of surveillance images has escalated, driven by increased surveillance capabilities and reduced personnel, especially in military contexts. Electroencephalogram (EEG) plays a pivotal role in BCI advancements, particularly in rapid serial visual presentation (RSVP) tasks requiring high temporal resolution for swift target detection amid rapid visual stimuli. We developed a real-time BCI system using the RSVP paradigm for automatic classification of target responses from EEG signals. Three machine learning-based classifiers were evaluated to classify EEG data into target and non-target responses, which were Support Vector Machine (SVM), Multi-layer Perceptron (MLP), and K-Nearest Neighbor (KNN). The top-performing classifier was selected for integration into the real-time BCI system. In offline evaluations, the BCI system using KNN achieved the highest F2-score of 0.9685 with an area under curve (AUC) value of 0.9562. The online performance of the real-time BCI system using KNN resulted in the best F2-score of 0.5556. This proposed BCI system demonstrates potential in detecting target recognition from EEG signals automatically, even though rapid responses to the target. © 2024 IEEE.

키워드