Feature Selection in Audiovisual Emotion Recognition Based on Rough Set Theory

Abstract

Affective computing is becoming an important research area in intelligent computing technology. Furthermore, emotion recognition is one of the hot topics in affective computing. It is usually studied based on facial and audio information with technologies such as ANN, fuzzy set, SVM, HMM, etc. Many different facial and acoustic features are considered in emotion recognition by researchers. The question which features are important for emotion recognition is discussed in this paper. Rough set based reduction algorithms are taken as a method for feature selection in a proposed emotion recognition system. Our simulation experiment results show that rough set theory is effective in emotion recognition. Some useful features for audiovisual emotion recognition are discovered.