Comparison of human and machine based facial expression classification

Facial expressions are most commonly used for interpretation of human emotion. Over the last few decades, major advances in understanding and analysis of facial expression was achieved by application of computer vision, image processing, and machine learning techniques. In this paper we propose a method to classify facial expression in two classes using the Zernike moments. The proposed system consists of two parts: facial feature extraction and facial expression classification. The facial features were extracted using higher order Zernike moments and the features were classified by an ANN based classifier. The facial expressions were classified into groups that represented either positive or non-positive emotion. The system was tested on Cohn-Kanade (CK) dataset and 69% accuracy was achieved. Also a survey was conducted where human subjects were asked to classify a given facial image (of the same dataset) as having positive or non-positive emotion. The results from the survey shows that the human subjects could classify expressions with 78% accuracy. Comparison of these two studies show that machine based facial expression classification can be performed with similar accuracy as achieved by human operators.