EEG decoding for detection in rapid serial visual presentation with small target objects

Abstract

In this study, a brain-computer interface-based EEG signal analysis algorithm is proposed for detecting small targets during rapid serial visual representation (RSVP). This research also suggests a unique system where the EEG decoding algorithm cooperates with both computer vision artificial intelligence and user’s cognitive response to detect targets, thereby overcoming the limitations of existing systems and improving the overall performance. To deal with the problems caused by target objects smaller than the objects in the conventional RSVP tasks, the proposed algorithm learns to detect target signals that can be evoked at varying times during visual stimuli, and its performance was confirmed by comparison with existing methods. In addition, the new collaborative system enables both humans and EEG decoding algorithms compensate for the shortcomings of each other by repeating the process of both human recognition for visual stimuli and subsequent EEG signal decoding for the stimuli that have not been recognized by the human user. Enhancements in the system performance were confirmed through experiments to show the potential of brain-machine interface research in a new direction.