个人身份认证的正面人脸识别算法.pdfVIP

8 (A ) 8 Vol. 8(A) , No . 8 2003 8 Journal of Image and Graphics Aug . 2003 ) ,2) ) 陶　亮 　庄镇泉 ) ( , 230026) 2) ( , 230039) ( SVM) , ( ) . M : ( ) : , M SVM( M ) ; (2) : ( ) SVM, SVM , . ORL , . ( 520 6040) ORL : T P39 .4 : A : 006-896 ( 2003) 08-0860-06 An Effective Approach for Frontal Face Verification TA O Liang ) ,2) , ZHU ANG Zhen-quan2) ) ( Dept. of Electronic Science and Technology , University of Science and Technology of China, H ef ei 230026) 2)( Dep t. of Electronic Engineering and Inf ormation Science, A nhui University , H ef ei 230039) Abstract T his paper presents an effective algorithm for front al face verification based on the w avelet decomposition t echnique and Support Vector Machines ( SVMs) . The process of t he proposed method for face verificat ion of M clients consists of tw o st ages. ( ) Training stage: by the wavelet decomposition, extracting the appropriate features from the facial images in the prepared training dat abase of f aces, training M SVMs for the M client s by the extract ed facial feat ure vectors. (2) Verification stage: select ing a trained SVM from the M SVMs based on the ident ity claim ( such as a name or a passw ord) of an unknow n person and using the t rained SVM t o classify t he facial feature vector ext racted from the f acial image of t he unknown person by the w avelet decomposition, and the classificat ion result w ill show whether or not the identit y claim of the unknow n person is valid. T he ORL dat abase of faces is select ed t o test and evaluate the propo