The Research On Facial Expression Analysis Based On Bi-Dimensional Principal Component Analysis And Fisher Linear Discriminate

Facial expression recognition is a very challenging and very meaningful issue, itrelates to computer vision, psychology, physiology and some technology in patternrecognition and image processing. With the advancement of technology, people needmore and more from the computer, they want computer to be more convenient,friendly and interact more intelligently, besides, facial expression recognitiontechnology has broad prospects in many fields.According to the advantages and disadvantages of current methods, this paperfirst introduces the basic processes of facial expression recognition and thenconcentrates on the key issue of facial recognition, feature extraction, and expressionclassification, the main work is as follows:In face detection and image preprocessing, this paper introduces face detectionbased on the Haar-Like characteristics and theAdboost learning method, and thentakes some steps to preprocess the face image, like normalize the size of the image,remove the noise and histogram equalization.In feature extraction, this paper introduces feature extraction based on principalcomponent analysis, then introduces dimensional principal component analysis toimprove the method, and later improved into a bi-dimensional principal componentanalysis, since Fisher linear discriminate can maximum the distance between differentclasses, we combine the bi-dimensional principal component analysis fisher lineardiscriminate to extract features.In classification, this paper introduces the principle and advantages of SVMclassifier, the experiments use radial basis function and one-to-one SVM classifier toidentify different expressions, and contract the results of different extraction methods.The result shows with the bi-dimensional principal component analysis and Fisherlinear discriminate feature distraction and the SVM classifier, the expressionrecognition achieves good results.