Fuzzy Criterion Based Wavelet Feature Selection for Face Recognition

This paper proposes a fuzzy-criterion-based wavelet feature selection method for face recognition. Firstly, the optimal WPD (wavelet packet decomposition) with the fuzzy criterion is obtained. Secondly, the classification abilities of the features(WP coefficients) in the optimal WPD are also evaluated and ranked by the fuzzy criterion. Thirdly, the discriminant features are selected and input into EFM(enhanced fisher linear discriminant model) for the purpose of dimension reduction, and the classification with linear discriminant functions based on least square error is made. Finally, the experiment on face recognition-indicates that the recognition rates achieved by fuzzy based wavelet feature selection are higher than those obtained by PCA(principal component analysis) method.