The contrast analysis of facial expression recognition.pdf

Figure 1. Examples of still expression images From left to right: Neutral, Happiness, Anger, Disgust, Sadness, Fear, Surprise. The Contrast Analysis of Facial Expression Recognition by Human and Computer Tang Jinghai 1 Ying Zilu 1,2 Zhang Youwei 1,2 1 School of Electronics and Information Engineering, Beihang University, Beijing China, 100083; 2 School of Information, Wuyi University, Jiangmen, Guangdong China 529020 Email: tangjh999@ Abstract From the trial investigation of human to identify facial expressions, the contrast of facial expression recognition between human and computer is analyzed. The phenomena that one expression may arise from several emotions (“one to more”) and various expressions may reflect one emotion (“more to one”) result in the problem that facial expression recognition is a difficult task because of these mappping uncertainties. In this paper, some points on understanding facial expression recognition are raised and several important aspects of the feature research are also discussed 1. Introduction In speech recognition and face identification, researchers always take the accuracy of classification as the final criterion to evaluate an algorithm with higher accuracy better performance. In the very beginning of our research in expression recognition, we treat the problem just the same as that of face identification while ignore the differences between them. In fact, these two questions are different in nature. The objects need to be recognized in face identification are faces which are much different from each other, and all expressions are undesired factors, but in expression recognition, the objects are several expressions and the difference of facial shape is interference factor that should be eliminated. We tried almost all existing methods of face identification to classify expression person-independently and the accuracies obtained are much lower than those in face identification. This fact giv