Feature selection for person-independent 3D facial expression recognition using NSGA-II

In this paper, the problem of person-independent facial expression recognition from 3D facial features is investigated. We propose a methodology for the selection of features that uses a multi-objective genetic algorithm where the number of features is optimized to improve classification accuracy. The facial feature selection aims to derive a set of features from the original expression images, which minimizes the within-class separability and maximizes the between-class separability. We used non-dominated sorted genetic algorithm II (NSGA II) which is one of the latest genetic algorithms developed for resolving problems of multi-objective aspects with more accuracy and higher convergence speed. The proposed methodology is evaluated using 3D facial expression database BU-3DFE. Facial expressions such as anger, sadness, surprise, joy, disgust, fear and neutral are successfully recognized with an average recognition rate of 88.18%.