Recognition of Facial Expression Using Centroid Neural Network

Abstract

A novel approach to recognize facial expressions from static images is proposed in this paper. The local binary pattern (LBP) operator is adopted as an effective feature extraction tool for facial image data. An unsupervised competitive neural network, called a centroid neural network with x2 distance measure, CNN-x2, is then utilized as the classification tool for the histogram data obtained by the LBP operator on facial image data. The proposed recognition scheme is applied to the JAFFE database and compared with several conventional approaches to facial expression recognition problems. The results show that the proposed recognition scheme compares favorably with conventional approaches in terms of recognition accuracy.