Abstract
With increased levels of security threats and the long-term consequences of falsely accusing the innocent and freeing the guilty, there is a growing need for reliable and efficient deception detection systems. Polygraph tests are invasive and require elongated time and human expertise, which is subject to bias and error. In this paper, we analyze thermal and visual clues of deception using a dataset collected from 30 subjects and multiple scenarios. We analyze expressions and other visual features and provide the first comparison between thermal facial regions to identify areas with higher capability of indicating deceit. Our experimental results show that our non-contact feature fusion model outperforms traditional physiological measurements, paving the road for non-invasive deception detection methodologies.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have