HumanFaceRecognition UsingGeneralized Kernel Fisher Discriminant andWavelet Transform
Yu Sun · 2006
Inthispaperthegeneralized kernel fisherConsequently, this method isalso referred toaskernel fisher discriminant (GKFD)methodisusedtodopattern featurediscriminant (KFD). extraction forhumanface image. First, weextend theKFD Infact, theKFD hasbeensuccessfully usedinpattern originally usedinpattern classification problems tothe recognition problems, butitcanonlysolve those problems generalized KFD (GKFD), whichwill beusedinfeaturewithtwoclasses. Formoredetails, please refer to(5)(6). In extraction problems. Compared toseveral commonly used this paper wewill extend theKFDtodofeature extraction for feature extraction methods, theGKFD cannotonlyreduce multiple classes ofproblems. Thus, themethod isalso named thedimension ofinput pattern, butalsoprovide theusefulasgeneralized KFD(GKFD). Thosefeatures obtained bythis information forpattern classification. Further, this GKFD GKFD aredirectly classified bythenearest neighbor method. alsoperforms wellforlinearly nonseparable patternInthis paper, we takethehumanfacerecognition datafor classification problemsforitpossesses a nonlinearexample toconduct therelated experiments. Theexperimental transformation capability. To reducethecomputationresults verify theeffectiveness andefficiency ofourproposed complexity, theoriginal faceimages arepre-processed by GKFD. wavelet transform. Finally, theexperimental results on 1.ANoVELMETHODFORFEATUREEXTRACTIONBASED ON human facerecognition problemsdemonstrate the