The method of facial expression recognition based on DWT-PCA/LDA

An efficient facial expression recognition method by combining the Discrete Wavelet Transform (denoted by DWT) with PCA /Fisher Linear Discriminant Analysis(denoted by PCA/FLD) method is proposed. First, each image is decomposed by using DWT, so that the decomposed image contains the large amount of information of the original image. Then, FLD approach is used to extract features from the decomposed low-frequency and high-frequency components. Finally, using the nearest neighbor classifier(denoted by NNC) to recognize the facial expression. High-frequency information characterizes the details of the contour, edge and texture features of images, which are the key to facial expression recognition. So the algorithm selects the low-frequency component meanwhile retains high-frequency information, and through experiments to verify the impact of high-frequency component on facial expression recognition. The algorithm is simple, fast, easy to implement, experiment results show that recognition performance is good, besides it holds some robustness and generalization.