Facial Expression Recognition Based On Gabor Wavelet Transform And Svm

Facial expression recognition can be widely applied in many areas such as e-education, robotics, intelligent human-computer interaction, medical treatment etc, and plays an important role in the development of pattern recognition, computer vision, computer graphics, artificial intelligence, physiology, psychology and so on.Based on synthesizing the key technique of expression recognition, the advantage of Gabor wavelet transform over image feature extraction is utilized. And since SVM has a good performance in the Small Sample Size (SSS) problem, the method is presented, that is, combining Gabor wavelet, Principle Component Analysis (PCA) and Fisher Linear Discriminant Analysis (FLD) to extract the expression features, and then using SVM as the classifier.The location of eyes is determined by integral projection. According to the coordinate, geometry preprocessing is accomplished. Histogram equalization is used to reduce the difference of illumination. On this basis, Gabor wavelets with 5 scales and 8 orientations are applied on images and the amplitude of convolution is used as the feature. Subsample is used to first reduce the feature dimension, PCA and FLD are then implemented in order to father reducing the dimension and training feature space. Thus, the features are more likely to be labeled; meanwhile, time for classification is shortened. Finally, SVM is applied to sort the seven expressions: angry, disgust, fear, happy, sad, surprise and neutral expression.Experiments of human dependent and undependent are performed on the JAFFE database, and the results indicate that the method presented here is an effective one for expression recognition.