Combination of SVM and Score Normalization for Person Verification based on Face and Fingerprint Biometrics
Hongzhou Zhang · Microcomputer Information · 2009
Normalization based fusion and classifier based fusion are two kinds of basic methods in score level multi-modal biomet rics fusion. The former transforms scores from different biometrics system into a common domain to make them comparable, whilst the latter focuses on designing a better classifier which can be classified the scores from various biometrics systems efficiently. A novel fusion strategy combines these two methods to take advantages of both, the scores are normalized firstly and then the normalized scores are classified by an SVM classifier. Experimental results carried on a pseudo multi-biometric authentication system using face and fingerprint show that this strategy greatly improved the fusion performance compared with the two basic methods.