Facial Expression Recognition Based on Hybrid Approach

Facial expression recognition using various approaches (appearance, geometric or a combination of them) is an interesting and challenging research topic. It has many potential applications in humancomputer interaction, social robots, deceit detection and behavior monitoring. This paper proposes an automatic system for facial expression recognition which consists of a hybrid approach (appearance and geometric) in the feature extraction phase. Appearance features are extracted as Local Directional Number (LDN) descriptors while facial landmark points and their displacements are considered as geometric features. Expression recognition is performed using multiple SVMs and decision level fusion. The proposed method was tested on the Extended Cohn-Kanade (CK+) database and obtained an overall 96.36% recognition rate which outperformed the other state-of-the-art methods for facial expression recognition.