Face recognition based on wavelet transform and SVM

This paper proposed a new scheme for human face recognition using wavelet transform combined with support vector machine as well as clustering method. The features in our research are: 1) using low frequency subband coefficients LL of wavelet decomposition as input for SVM, to attenuate the influence of natural differences, 2) do fine recognition by multi-method of PCA, LFA on pre-accepted image to decrease FAR and for machine learning, 3) conduct homomorphic filter to face image for pre-processing to deal with illuminations influence, 4) machine learning while recognition, update or adjust mode vectors by results of fine recognition, 5) clustering before doing face recognition on multi-target gallery to reduce search time. Experiments on ORL face dataset and self-build face library show efficient results.