The Effect of Classification Methods on Facial Emotion Recognition ‎Accuracy

The interests toward developing accurate automatic face emotion recognition ‎methodologies are growing vastly, and it is still one of an ever growing research field in the ‎region of computer vision, artificial intelligent and automation. However, there is a ‎challenge to build an automated system which equals human ability to recognize facial ‎emotion because of the lack of an effective facial feature descriptor and the difficulty of ‎choosing proper classification method. In this paper, a geometric based feature vector ‎has been proposed. For the classification purpose, three different types of classification ‎methods are tested: statistical, artificial neural network (NN) and Support Vector ‎Machine (SVM). A modified K-Means clustering algorithm has been developed for ‎clustering purpose. Mainly, the purpose of using modified K-means clustering technique ‎is to group the similar features into (K) templates in order to simulate the differences in ‎the ways that human express each emotion. To evaluate the proposed system, a subset ‎from Cohen-Kanade (CK) dataset have been used, it consists of 870 facial images ‎samples for the seven basic emotions (angry, disgust, fear, happy, normal, sad, and ‎surprise). The conducted test results indicated that SVM classifier can lead to higher ‎performance in comparison with the results of other proposed methods due to its ‎desirable characteristics (such as large-margin separation, good generalization performance, etc.). ‎