Match score integration of iris and fingerprint in multibiometrics system

Unimodal biometric systems encounters variety of limitations such as noisy input data, universality of traits, inability to enroll, unacceptable error rates and spoofing. Multimodal biometric system aims to fuse output of multiple biometric classifiers to alleviate some of these limitations. This paper proposes a novel multibiometric system using two most used biometric traits fingerprint and iris. Feature vector of each of traits are extracted from texture pattern of biometric images, using discrete wavelet transform and principal component analysis technique. Classification of these feature vectors is carried out using Euclidean distance. The match score obtained are fused together using simple sum rule assigning equal weight to both the modalities. Experimental results on combining these two biometric shows that the proposed multimodal system provides significant improvement in performance over individual unimodal systems. Equal error rate of 0.354500 has been achieved for the proposed system.