Automatic Facial Expression Recognition Using Extended AR-LBP

The Local Binary Pattern (LBP) based operators are sensitive to localization errors. To mitigate these errors input images are manually aligned, face is localized using eyes co-ordinates in the image before feature extraction and multi-scale or multi operators are used, which restricts the use of LBP based operators for automatic facial expression recognition. This paper proposes an Extended Asymmetric Region Local Binary Pattern (EAR-LBP) operator and automatic face localization heuristics to mitigate the localization errors for automatic facial expression recognition. The proposed operator along with face localization heuristics was evaluated for person-independent facial expression recognition on JAFFE and CK+ databases using a multi-class SVM with Linear and Radial Basis Function (RBF) as kernels. It is observed that face localization and the EAR-LBP method are able to mitigate the localization errors to produce reasonably better performance. Maximum 10-fold cross validation average performance of 58.74% and 60.35% were obtained on JAFFE and in case of CK+ database, maximum performance of 83.09% and 82.21% were obtained using Linear and RBF kernels for SVM multi-class classifier respectively.