The belief theory for emotion recognition

Abstract

This paper presents a facial expression classification system based on a data fusion process using the theory of belief. Such expressions correspond to the six universal emotions (happiness, surprise, disgust, sadness, anger, and fear) as well as the neutral expression. The suggested algorithm rests on the decision fusion of both approaches: the global analysis and the local analysis of facial components. The classification result, throughout these two approaches, will be enhanced by fusion. The performance and the limitations of the recognition system and its ability to deal with different databases are identified through the analysis of a large number of results on the FEEDTUM database.