Trends in computational modeling on facial expression recognition

Abstract

This paper surveys studies aimed at building computational models of human cognitive processes of facial expression, done mainly in the machine learning domain. By aggregating a set of labels obtained through online crowdsourcing, the ground truth of item/instance is estimated while maintaining a psychological perspective. These models are closely connected with theories established in psychology such as signal detection theory and item response theory. In that respect, the boundary between the machine learning and psychology fields seem to be considerably narrowed. The purpose of this paper is to clarify the relationship between these two research areas to promote the development of both.