Concealed knowledge identification using facial thermal imaging

In this paper, we present a non-intrusive lie detection system based on thermal imaging technologies. The system consists of the following modules: thermal camera, face detection and tracking, face landmark detection, feature extraction, and pattern recognition for concealed knowledge inference. We have discovered the most sensitive areas on the human face to monitor facial temperature changes. Detection algorithms are then developed to identify concealed knowledge from thermal imaging automatically. Face landmark tracking is used directly on the thermal video images to detect regions of interest (ROI) and extract features for concealed knowledge inference. We achieved an equal error rate (EER) of 16.5% in concealed knowledge recognition for 16 subjects on test data. Our non-contact method of concealed knowledge detection using thermal data achieves similar or better recognition accuracy as traditional intrusive methods, such as polygraph or EEG.