System Realization Of Verification Mode Multi-modal Biometrics

Biometrics is an identity authentication technology full of wide development prospect rising recently in security field. Compared with those traditional identity authentication methods, biometrics is more reliable and secure. Now the main research direction of biometrics is multi-modal biometrics based on information fusion which fuses several biometnc traits of different modals on the various levels during the operation of the authentication system and then gives the integrated result. Multi-modal biometrics has greatly improved the recognition performance and user acceptance where single-modal biometrics is usually deficient and enhanced the robustness and practicability of biometric system. And for these reasons, it's more suitable for the security inspection applications under various environments and circumstances.The main research content of this article is the system realization of verification mode multi-modal biometrics and construction of a real-time multi-modal biometric system based on face and voice. In the preface, it gives some brief introduction of the basic concepts and recent development status of biometrics and also indicates the advantages of multi-modal biometrics. The second chapter lays stress on the research work of multi-modal biometric system. It puts forward a universal system framework model, introduces the mathematical description of system performance and the authentication mechanism under verification mode, and then expounds the sensor level fusion, feature level fusion, match level fusion and decision level fusion in detail. In the third and forth chapters, this article respectively introduces recently prevalent face recognition and speaker recognition technologies, in which a face detection algorithm based on hausdorff distance and an improved LBG algorithm using codewords shifting are brought forward. The fifth chapter has constructed a real-time multi-modal biometric system based on face and voice. According to the experimental results, it can be seen that the recognition performance of this multi-modal system is evidently better than those single-modal systems based on the one of the two biometric traits above which shows the superiority of multi-modal biometrics using decision level fusion. And the last chapter of the article introduces two practical cases of multi-modal biometric systems including the case of the ATM system based onpassword, face and voice which has great feasibility.