A Novel Gabor Filter Selection Based on Spectral Difference and Minimum Error Rate for Facial Expression Recognition

A new feature selection approach is proposed for facial expression recognition system. The features are extracted using Gabor filters from Grey-scale images for characterizing facial texture. Then, an adaptive filter selection (AFS) algorithm is applied to choose the best subset of Gabor filters with different scales and orientations. In AFS algorithm, the filters are selected based on spectral difference between the original image and the noisy image in Gabor wavelet domain. After that, the optimum subset of filters is selected based on minimum error rate. This subset of Gabor filters is used for feature extraction. The extracted features are classified by adopting a multiple linear discriminant analysis (LDA) classifier. Experiments on different databases are carried out that the method is efficient for facial expression recognition.