The Extraction and the Recognition of Facial Feature State to Emotion Recognition Based on Certainty Factor

Psychologically, emotion is related to someone’s feeling in particular condition. Some fields like: health, psychology, and police investigation need the information of emotion recognition. Human’s emotion can be classified into six types include happy, sad, angry, fear, disgusted, and normal. Psychologically, there are some methods that can be used to emotion recognition, like self-report analysis, automatic measure, startle response magnitude, FMRI analysis (Functional Magnetic Response Imaging), and behavior response. Each of those methods has their own advantage and disadvantage. The aim of this research is to determine someone’s emotion in a video scene. The video was decomposed into image frame and in each image frame was extracted into feature (component) face, which include mouth, eyes, nose, and forehead. The feature extraction was done by combining two methods based on the color and the facial geometric figure. The state of each face feature related to AU’s (Action Unit) face that used to emotion recognition. State recognition of mouth and eyes can be seen based on the feature elongation, state on the forehead and nose are known based on the wrinkle density. In the last part of emotion, recognition is done with certainty factor method to determine the quality of each emotion, the classification of actor’s emotion is determined based on the quality level of maximum emotion. The results showed recognition of emotion in a single image, the average accuracy of 77%, while in the video, the average accuracy of 76.6%.