期刊名称:International Journal of Multimedia and Ubiquitous Engineering
印刷版ISSN:1975-0080
出版年度:2015
卷号:10
期号:3
页码:35-44
DOI:10.14257/ijmue.2015.10.3.04
出版社:SERSC
摘要:In this paper, we present a method for fully automatic facial expression recognition in facial image sequences using feature extracted from tracking of facial landmarks. The facial landmarks at the first frame of the image sequence under examination are initialized using elastic bunch graph matching (EBGM) algorithm and tracked in the consecutive video frame over time. At first, the most discriminative geometric features in terms of triangle are selected using multi-class AdaBoost with extreme learning machine (ELM) classifier. The features for facial expression recognition (FER) are extracted from AdaBoost selected most discriminative set of triangles composed of facial landmarks. Finally, the facial expressions are recognized using support vector machines (SVM) classification. The results on the extended Cohn-Kanade (CK+) and Multimedia Understanding Group (MUG) facial expression database shows a recognition accuracy of 97.80% and 95.50% respectively using proposed facial expression recognition system.