The Feature Extraction And Recognition Of Facial Expression

Expression recognition is becoming a hot research field of artificial emotion, artificial psychology and pattern recognition. Expression recognition technology to computer anthropomorphic expression recognition, men communicate through facial expressions, through the expression to read a person’s inner world, which is of great significance to build the natural harmony of human-computer interaction environment. Expression recognition technology is one of affective computing, is the intersection of the field of psychology, physiology, image processing, machine vision, pattern recognition, a challenging subject.Firstly read carefully on the basis of a large number of domestic and foreign literature and books on Facial Expression Recognition subject background research value, application are as and domestic and foreign research status described. The main stream and facial expression recognition system involves various steps today to do a more comprehensive overview. This has laid a solid theoretical foundation for expression recognition technology research and explore.Secondly, the original expression image preprocessing, mainly including geometric the pretreatment and grayscale pretreatment, crop, the illumination normalization one and scale normalized, noise reduction, feature extraction and classification. Then pretreated image feature extraction, the paper analyzes the domestic and international on expression recognition features extract the theory and the results on the basis of proposed based on Gabor wavelet transform and PCA to reduce the dimension combined feature extraction algorithm, which can effectively extract the low-dimensional expression characteristics most conducive to classification.After that, based on the advantages and disadvantage so fartificial neural network, the paper constructs based on multi-decision neural network classifier, and apply it to the aspect of the facial expression recognition, good or bad according to the classification of the classifier set and adjust the neural network parameters, give full play to the advantages of fast learning and generalization ability.Finally, in order to verify the effectiveness of the propose dual gorithmin Matlab platform facial expression recognition system.