Facial emotion recognition: A comparative analysis using 22 LBP variants

Facial expression is a significant form of non-verbal communication for human being. It includes much important information about the feeling, the mental and the emotional state of a person which can be useful in several real-world applications and fields like image processing and computer vision. Face can be seen as a composition of micro-patterns of textures. Over the last decades, LBP operator, which shown its robustness in extracting useful features characteristics from an image, has been successfully applied in diverse range of problems including facial expression recognition. Nowadays, many LBP variants have been proposed in the literature. This paper reviews 22 LBP-like descriptors and provides a comparative analysis on facial expression recognition problem using two benchmark databases, the Japanese female facial expression (JAFFE) and Cohn-Kanade (CK) databases. The experiments show that several of the evaluated methods achieve performances that are better than those recorded by the state-of-the-art systems. Recognition rates of 97.14% and 100% have been reached on JAFFE and Cohn-Kanade databases respectively.