RCIT: An RSVP-Based Concealed Information Test Framework Using EEG Signals

摘要

Concealed information detection has a strong connection with human cognition. Existing deception detection approaches, such as polygraph testing usually exploit psychophysiological changes. However, they often suffer from low accuracy and attacking by training. Compared with conventional physiological responses, electroencephalogram (EEG) directly reflects activity responses in the brain, thus can be potentially more accurate in deception detection. In this article, we propose an EEG-based concealed information detection framework. In this framework, different test trials are designed to evoke effective brain responses corresponding to concealed information while keeping the subject focused. Meanwhile, we employ rapid serial visual presentation (RSVP) for deception detection. Our framework presents the stimuli on the fringe of awareness, making it not easy to take the countermeasures. Furthermore, we develop a method based on neural network to detect brain responses. Experimental results with 10 subjects show that, our approach can effectively detect concealed information with an accuracy of 87.13%.