Enhanced Gabor (E-Gabor), Hypersphere-based normalization and Pearson General Kernel-based discriminant analysis for dimension reduction and classification of facial emotions

Abstract

Abstract This paper puts forward an Enhanced Gabor feature descriptor termed as E-Gabor for obtaining high classification accuracy of emotions with low dimension. Two methods have been used for further classification. In the first method, E-Gabor is used as a stand-alone feature for classification. Hypersphere-based normalization has been used for normalizing the E-Gabor features, thereby improving the efficiency in the classification of emotions. In the second method, the E-Gabor feature descriptor is fused with Pyramid Histogram of Gradient (PHOG) feature descriptor and projected to a common subspace of six dimensions using the proposed Pearson General Kernel-based Discriminant Analysis (PGK-DA) before classification. In both the methods, Pearson General Kernel-based Extreme Learning Machine (PGK-ELM) is used for classification. Experiments conducted on Japanese Female Facial Expression (JAFFE), Cohn Kanade (CK+), Multimedia Understanding Group (MUG), Static Facial Expressions in the Wild (SFEW), Oulu-Chinese Academy of Science, Institute of Automation (Oulu-CASIA) and Man–Machine Interaction (MMI) datasets report a classification accuracy of 97.6%, 97.9%, 95.7%, 35.4%, 87.7% and 82.7% with method I and 95.7%, 97.2%, 94.9%, 35.2%, 87.1% and 82.1% with method II, respectively, for seven class emotion detection, which is high when compared to other state-of-the-art methods.