Computational Intelligence Methods for Facial Emotion Recognition: A Comparative Study

Abstract

Emotion recognition plays a critical role in the human communications. It is one of the major ways to be in touch with others. Four parameters including eye opening size, mouth opening size, ratio of eye opening size to eye width and mouth width are used as a reduced-size feature set in this study. This paper compares the performance of facial emotion recognition classification models based on the following computational intelligence methods: fuzzy logic, chaotic gravitational search algorithm (CGSA), and artificial neural network (ANN) from eyes and mouth features tested on the FACES database. Experimental results show the superior performance of ANN-based method compared to fuzzy- and CGSA-based methods. In addition, this comparative study triggers the idea of a hybrid system based on these computational methods that outperforms the human detection system.