Modelling human perception of static facial expressions

Abstract

Data collected through a recent web-based survey show the perception (i.e. labeling) of a human facial expres- by a human observer is a subjective process, which in a lack of a unique ground-truth, as intended in standard classification framework. In this paper we the use of Discrete Choice Models(DCM) for hu- perception of static facial expressions. Random utility are defined in order to capture the attractiveness, by the human observer for an expression class, asked to assign a label to an actual expression image. utilities represent a natural way for the modeler to for- her prior knowledge on the process. Starting with model based on Facial Action Coding Systems (FACS), subsequently defines two other models by adding two sets of explanatory variables. The model parameters learned through maximum likelihood estimation and a cross-validation procedure is used for validation purposes. © 2008 IEEE.