Research On Facial Expression Recognition Based On Local Facial Texture Feature Extraction

For human beings, facial expression is a basic way to express their emotions and is an effective means to conduct nonverbal communication. People can express their thoughts and feelings accurately and subtly by the expression, and also they can identify others’ attitudes and minds through it. Therefore, researchers recognized that it could be a better way to serve humanity if the machines have the ability to recognize and understand human facial expression.In recent years, with the depth improved of facial expression of face detection, face tracking and face recognition technology, facial expression recognition research has gradually become a hot research field of pattern recognition and artificial intelligence. At the same time, many of the central issues in the progress of facial expression recognition are stand out, such as how to detect the exact position of the face in the image, how to extract the active expression feature and how to choice the right classifier to get the best result with extracted facial texture featuresThis dissertation studies the research status and some application problems of facial expression recognition at home and abroad systematically, and raise some proposes to improve some algorithms for some key issues, the main research work includes:1. Face image preprocessing. Detect the exact position of the face in the image by the gray integral projection method. Using the secondary gray integral projection method to locate the key feature points (eyes, nose, mouth) of face in the face image. Through the Image preprocessing, we got precise facial expression, which is important for the latter expression feature extraction and expression classification.2. Extracts static texture expression feature of image faces. In this process, the main purpose is to got the effective low-dimensional expression feature, and to study the advantages and disadvantages of Traditional Local Binary Pattern (LBP) method, propose the Local Directional Binary Pattern (LDBP) method on the basis of the merits of the traditional LBP method, and then combine the high-dimensional expression features extracted by LDBP with locally linear embedding (LLE) method to achieve low-dimensional effective expression to extract texture features.3. Research the ideas about LBP method and related method, propose the Local Vector Method (LVM). In this method, express the local texture feature by vector instead of scalar quantity, this way can describe the local texture variations more detail. To enhance the robustness of the face recognition system, Image Euclidean Distance (IMED) method is introduced, to reduce the amount of calculation, the vector representation is converted to a plural, and then embedded the IMED method in a complex representation of features. Through the analysis of experimental data, we can ultimately prove that the algorithm is more effective, and has better robustness for the slight deformation caused by the disturbance.