CLASS DISCRIMINATING FEATURE VECTOR-BASED SUPPORT VECTOR MACHINE AND FACE MEMBERSHIP AUTHENTICATION BASED ON IT

摘要

PURPOSE: A class discriminating feature vector-based SVM(Support Vector Machine) and a face membership authentication method based on the same are provided to authenticate a face despite the changes in illumination, facial pose, registrant class configuration, and so on by providing an efficient class discriminating feature vector-based SVM. CONSTITUTION: A registrant face image set of a training set is normalized, and PCA(Principal Component Analysis) mode spatial Gabor vectors are obtained by projecting a face Gabor bunch onto a PCA mode vector partial space. The face Gabor similarity of registrant face images are required by projecting the face Gabor bunch of the face images onto the PAC mode vector partial space. The class discrimination feature vector of the face images are calculated by suing the face Gabor similarities.