Emotional Geometrical Features And Support Vector Machine-based Facial Expression Recognition

During the past three decades, facial expression analysis has attracted more and more attention in the computer vision field for its many applications, such as human-machine interaction, image understanding, synthetic face animation, and web services. Especially, human's expression classification plays a vital role in the human-machine interaction, because considerable research in social psychology has shown that facial expressions help communication between two speakers.We have done some researches on algorithm and applications of facial expression when I study for the master degree.1) In terms of applications of expression recognition on the healthcare industry, we classify happiness, sadness and neutral expression, using the Emotion Geometry Feature (EGF) we proposed and Support Vector Machines. And the recognition rate gets 92.22%.So the doctors or nurses can do the matching services according to the recognized expression.2) In order to classify the patients'expressions, we organize the six basic expressions into three classes, which are positive, negative, and neutral expression according to what the doctor and psychologist said. The positive expression is happiness, and the negative expression includes the anger, sadness, fear, surprise, disgust. We use the extended EGFs to classify the three classes, and the result can get 93.8%.the positive and the neutral expression means that the patient is normal, and the negative expression means that the patient needs to be cared.3) In order to improve the recognition result of pattern recognition, we propose a novel classification model that is AtoC model. The AtoC model is used in the facial expression by us and we can get the accuracy 93% for the six basic expressions. The idea of AtoC model originates from the way human beings learn one thing, which is a process of recognizing one object from abstractly to concretely.