A novel Monogenic Directional Pattern (MDP) and pseudo-Voigt kernel for facilitating the identification of facial emotions

Abstract

Facial expressions are the best way of communicating human emotions. This paper proposes a novel Monogenic Directional Pattern (MDP) for extracting features from the face. To reduce the time spent on choosing the best kernel, a novel pseudo-Voigt kernel is chosen as the common kernel for dimension reduction proposed as pseudo-Voigt kernel-based Generalized Discriminant Analysis (PVK-GDA). The pseudo-Voigt kernel-based Extreme Learning Machine (PVK-ELM) is used for better recognition of facial emotions. The efficiency of the approach is proved by experimenting with the Japanese Female Facial Expression (JAFFE), Cohn Kanade (CK+), Multimedia Understanding Group (MUG), Static Facial Expressions in the Wild (SFEW) and Oulu-Chinese Academy of Science, Institute of Automation (Oulu-CASIA) datasets. This approach achieves better classification accuracy of 96.7% for JAFFE, 99.4% for CK+, 98.6% for MUG, 35.6% for SFEW and 88% for Oulu-CASIA, which is certainly higher when compared to other techniques in the literature.