Video-based deception detection

This chapter outlines an approach for automatically extracting behavioral indicators from video and explores the possibility of using those indicators to predict human-interpretable judgments of involvement, dominance, tenseness, and arousal. The team utilized two-dimensional spatial inputs extracted from video to construct a set of discrete and inter-relational features. Then three predictive models were created using the extracted features as predictors and human-coded perceptions of involvement, tenseness, and arousal as the criterion. Through this research, the team explores the feasibility and validity of the approach and identifies how such an approach could contribute to the broader community.