A novel approach to enhance automatic 3D facial expression recognition

Researches on 3D facial expression recognition have been extensively promoted in recent years, yet automatic 3D facial expression recognition is still a challenging problem. In this paper, we propose an easy-handled approach to address this problem and consequently improve its performance. In the approach, a 2D-image-like structure is utilized to represent the 3D models so that we can automatically extract the facial features from either its depth values or the texture information. Then the feature-based irregular divisions are specially designed to depict the facial features more accurately. Finally, a novel block weighted strategy which emphasizes the contribution of different facial regions is additionally applied to enhance the facial descriptors for classification. Using the general protocol for 3D facial expression recognition on the BU-3DFE database, each of the steps is validated and the proposed approach displays comparative performance which draws a promising direction for automatic 3D facial expression recognition.