Micro expression classification using facial color and deep learning methods

Micro emotions are a unique type of facial expression as they are involuntary and very brief. They usually occur when a person attempts to suppress or hide their emotions. Lasting less than 500 ms, they can be very hard to detect even for the human eye, however, since they reveal a person's true feelings, they are of extreme interest in many fields of study. Most approaches to automatic detection and classification of micro emotions rely on detecting the small residual facial movements. In this paper, we propose to exploit an aspect of the human face which is much harder to subdue, namely, facial color change due to blood flow during expression of emotion. We propose a system that evaluates color change during micro emotion expression and successfully classifies the emotion type. This approach is unique in that it disregards the motion related aspects of the expression, and relies entirely on the facial color. We show that our system improves over movement based approaches.