MULTIMODAL BIOMETRIC DECISION FUSION FOR LIVENESS AUTHENTICATION / ANTI-SPOOFING ENGINE

摘要

Biometrics is the science and technology of measuring and analyzing an individual's physical and behavioral characteristics to authenticate a person's identity. In this paper, a fusion algorithm for the purpose of liveliness authentication/anti-spoofing engine is introduced and implemented which consists of three major modules. Liveness Detection, Multimodal Recognition and Decision Fusion. The live face detection is carried out based on eye image and its movement analysis. Hit-Miss transform, Morphological Shared Weight Neural Network (MSNN) for face feature extraction and Daubechies wavelet (db2) & Back Propagation Neural Network (BPNN) for iris feature extraction are used. The individual scores of two traits, face and iris, are combined at classifier and trait levels. The simulation results show that the proposed multimodal biometric system outperforms the unimodal system with an accuracy of 98.68%. Algorithms, fusion techniques and simulated results are presented.