Recognition of Facial Emotion through Face Analysis based on Quadratic Bezier Curves

Emotion recognition is a challenging task of human-computer interface in wireless communication. Emotion recognition from speech has a problem with the quality of the input voice, which is difficult to ensure in the mobile environment. On the contrary, facial emotion recognition is one of the interesting subjects due to the relevance of the expressions on human emotions. This paper proposes an automatic extraction and interpretation method of facial expression using extraction of feature points and variation of the Bezier curve from still image. The proposed algorithm has three steps to recognize the facial emotion: (1) Detecting facial regions with feature map, (2) Drawing the Bezier curve on eye and mouth, and (3) Classifying the emotion of characteristic with Hausdorff distance. To evaluate the proposed recognition scheme, we estimate a success-ratio with emotionally expressive facial image repository. Experimental results show average 76.1% of success to interpret and classify the facial expression and emotion.