Research On Facial Expression Recognition Based On Weighted Rectangle Integral Image

Until recently, Facial expression recognition system has been widely applied in various fields with the development of computer technology. Based on the in-depth researches about expression feature extraction and classification, People proposed a variety of effective methods. Summarized the research background and significance of the topic and learned from the domestic and international discourse and research papers concerning facial feature extraction and classification in recent years, especially studied the facial feature extraction, we develop an extension integral image on the basis of existent integral image. The feature extraction method and support vector machine (SVM) achieve the expression recognition system.The main work is listed here:(1)The paper presents a new binary approach of combining edge detection with Otsu. The binary image is traditionally obtained by Otsu method to locate eye position, which not only can not separate foreground from background but also miss out a lot of useful information. Edge detection can intensify edge. Their combination not only separate foreground from background but also get the exact eye location, so this method makes for feature extraction and classification.(2)Rotation image and extension integral image are presented. According to the Gabor wavelet merits and demerits and its characteristics, new weighted rectangle templates are defined. A method of rotation image is also introduced based on rotation template. To compare with Gabor, integral image is extended for feature extraction, that is, extension integral image. The two methods successfully simulate Gabor multi-scale and multi-direction, obtain the pixels sum of their coverage and receive the value of feature points. The results show that they own the scale and directivity of Gabor especially meet the real-time requirements and improve efficiency of the system.(3)The combination of SVM and Gabor is studied deeply and validated their high recognition rate. Because Gabor evolves into extension integral image, we use SVM to class feature data. The experiment shows that their recognition rate is also high and system has good real time.(4)Facial expression recognition system is designed and realized by object-oriented technology. The system is constituted three modules of image pretreatment, feature extraction and classifier and validated its validity.