Research On Description And Recognition Method Of Facial Emotion Feature

Facial emotion is a research topic in the field of artificial psychology. It has much potential applications in theoretical research and practical application. Around the key technologies of facial emotion recognition, the paper summarizes the facial emotion research’s background and significance and introduces the current facial expression recognition development status at home and abroad. Then Analysis mainstream method used in facial expression feature extraction and expression classification, as well as illustrate several widely used facial expression database. This paper mainly discusses the feature extraction and classification algorithm in facial expression recognition system. The concrete work is as follows.In order to extract human facial expression features more quickly and efficiently, the paper firstly pre-process face image. The first step of pretreatment is edge detection of expression images using canny operator. Then locate the center of eyes using pixels’ gray information in edge image. According to the distance between right and left eye, normalize image, including size and rotation normalization.Secondly, extract expression feature using hybrid improved Active Shape Model (ASM) and Active Appearance Model (AAM). Both ASM and AAM are based on the point distribution model (PDM). PDM describes the object shape through a set of discrete control points and establish a motion model corresponds to each control point. Set constraints according to the reference position and moving mode of control points. The method could reduce the impact of noise and deformation owing to introducing target feature information. ASM is accurate in locating facial contour feature point. At the same time, AAM shows an excellent performance in internal key point location. Therefore, in this paper, the combination algorithm of ASM and AAM will be used in facial expression feature extraction. At the same time, in order to improve the matching speed, the graphic information is added to the measure function in the place of the traditional Mahalanobis to judge whether the algorithm convergence.Thirdly, it is in line with the SVM’s advantage that facial expression recognition is a small sample of nonlinear classification problems. As a classification method developed on the basis of statistical learning theory, support vector machine is a two-class classifier. In order to realize the identification of six basic expressions, the paper constructs the expression classifier using a plurality of SVM according to combination thought.Finally, in the foundation of the algorithm, program a simple face recognition system using C++. And use the face database images as the experimental object to test the robustness and accuracy of the facial expression recognition system.