Automatic Deception Detection in RGB videos using Facial Action Units

Abstract

The outcome of situations such as police interrogatory or court trials is strongly influenced by the behaviour of the interviewed subject. In particular, a deceptive behaviour may completely overturn such sensible situations. Moreover, if some specific devices such as polygraph or magnetic resonance are used, the subject is aware of being monitored and thus he may change his behaviour accordingly. To overcome this problem, in this paper a method for detecting deception in RGB videos is presented. The method automatically extracts facial Action Units (AU) from video frames containing the interviewed subject, and classifies them through an SVM as truthful or deception. Experiments on real trial court data and comparisons with the current state of the art show the effectiveness of the proposed method.