Micro-expressions (ME) are brief, fast facial movements that occur in high-stake situations when people try to conceal their feelings, as a form of either suppression or repression. They are reliable sources of deceit detection and human behavior understanding. Automatic analysis of micro-expression is challenging because of their short duration (they occur as fast as 1/15–1/25 of a second) and their low movement amplitude. In this study, we report a fast and robust micro-expression detection framework, which analyzes the subtle movement variations that occur around the most prominent facial regions using two absolute frame differences and simple classifier to predict the micro-expression frames. The robustness of the system is increased by further processing the preliminary predictions of the classifier: the appropriate predicted micro-expression intervals are merged together and the intervals that are too short are filtered out.
Part of the book: Intelligent Video Surveillance