A Study On Facial Expression Recognition

With the increase of people's interest in the interface of human and machine, affective computing is becoming a research focus. If computers and robots could understand and deliver emotion just like human, the relation between human and machine would be changed radically, and then the computer could provide better service for people.Facial expression which carries rich information of body behavior is the leading carrier of human affective and the symbol of intelligence. As an essential branch of affective computing, facial expression recognition is groundwork of emotion understand, and is the precondition for computers to understand human's emotion. Human affective needs much more research. Today, there is still no acknowledged definition of human affective itself. That is an inconvenience to analyze facial expression what conduced by emotion. So, the research of facial expression analysis mostly focuses on facial expression synthesization and recognition. Facial expression synthesization emphasizes on the simulation of facial expression. Facial expression recognition deals with the classification of facial motion and facial feature deformation into abstract classes that are purely based on visual information. But the facial expression in daily life usually is caused by multiple emotions. It is difficult to classify the facial expression into one special class. Achieve a rational result of daily life facial expression is a great challenge for human facial expression recognition.The main purpose of this paper is to recognize human facial expression. The research in this paper includes the extraction and distribution of the facial features, the modeling of facial expression space, and facial expression analysis of different face-shape. At last a facial expression recognition system is built based on special human. The system could test the model and methods presented in this paper. Also the system could be used as experimental flat for other facial expression research.There are some distinct differences between the work of this paper and present methods of facial expression recognition. Considering the physiological knowledge that facial expression is conduced by human emotion this paper studies the facial expression and its classify in virtue of the theory of emotion psychology. Then a new facial expression space model is achieved. The emotion state of the facial expression could be better represented by the new model. On the same while, considering the fact that every people have a unique face, the facial expression recognition framework based on face-shape is proposed by the analysis of the difference of facial expression among different face-shapes.The main innovations are as follows:(1) Considered the known contour character, a new eye-contour extraction algorithm is proposed based on the combination of the shape parameters got from the transform projection and the image information in the eye field.(2) The facial expression features are usually represented in the imperceptible change of the face. Gabor kernels are similar to the receptive field profiles in cortical simple cells, which are characterized as localized, orientation selective, and frequency selective. Then a multiple Gabor features based facial expression recognition method is presented in this paper. This method can improve the performance of a facial expression recognition system.(3) Modeling of facial expression plays a very important role in the research of synthesis and recognition of facial expression. This paper analyzes the nature of the facial expressions and gives a qualitative description of the corresponding facial expression space, and then proposes a new facial expression space model with the characters of both discrete affective space model and continuous affective space model. The experimental results show that the facial expression model proposed in this paper can rationally represent the daily facial expressions.(4) A facial expression recognition idea based on face-shape classifying is proposed. First, the face-shape classifying is analyzed, and the difference of the facial expression among different face shapes is described. Then the facial expression recognition framework is presented. The facial expression recognition experiment based on 3d data and 2d data is designed at last. The experiment results showed that the facial expression recognition based on face-shape classifying is an efficient method and can highly improve the recognition rate.(5) Combined facial expression space model and the facial expression recognition framework based on face-shape, a facial expression recognition system is presented. The system could be used as a flat root for the relative research of facial expression analysis.