Research And Implementation Of Facial Expression Recognition Based On Local Binary Pattern

Expression is the most direct expression of inner feelings, psychological studies have shown that about 55 percent of emotions through facial expressions manifested; the face is very complex flexible body, is also one of the most important biometric feature;recognition is the main method of computer to understand the world. So the study of facial expression recognition involves cross-content psychology, physiology, mathematics, computer science and many other disciplines, as well as facial expression recognition is the typical application of image processing, pattern recognition, computer vision, graphics, artificial intelligence, human-computer interaction. The successful application of Facial expression recognition will have a profound impact on the field of social, health, education, security, human-machine interface.The main work of this paper is divided into the following parts:(A) Makes a note on the background and significance of facial expression recognition, gives a brief introduction of current research of home and abroad. In expression facial features depicted as a starting point, by enumerating several expression database leads to the common research methods of Expression Recognition and its compared and described.(B)Makes a detailed expatiated about image preprocessing which is before expression feature extraction, including those based on haar-like features and Adaboost cascade classifier algorithm for face detection and eye location, the image gray-scale normalization about gray-scale transformation and histogram equalization, the geometric normalization about rotation scaling correction. These are the necessary measures to outstand characteristics of expression and remove irrelevant factors of expression.(C)Makes a detailed study of local binary pattern(LBP) operator, for the detailed texture features of expression a quick and effective description that LBP operator was taken, in addition takes a improved DLBP operator to boost recognition rate. Mainly elaborated the definition expression rotational invariance equivalent mode of LBP, and made a series of experiments of expression recognition with LBP feature and DLBP feature with support vector machine(SVM) on JAFFE library and received a relevant conclusion.(D)Designed and implemented a facial expression recognition software based on LBP feature. Using OpenCV vision library in VS2010 environment to achieve a software interface based MFC, which consists mainly of face detection module, image pre-processing module, feature extraction module, classification module. And gives a detailed description of the layout and usage of the interface. At last takes a function test with JAFFE and camera.