Multimodal Biometric Verification using the Iris and Major Finger Knuckles

Abstract

The drawbacks of unimodal biometric systems such as non-universality, noisy sensor data and spoofing can be mitigated using multiple biometric traits. In this study, a novel multibiometric system to authenticate users based on their major knuckle finger patterns using four fingers (i.e., little, ring, middle, and index) and iris is proposed. A local texture descriptor namely binarized statistical image features (BSIF) has been employed to extract the features for each of the biometric traits considered in order to improve biometric-based personal verification. The comparison results on PolyU contactless hand dorsal images database and IIT Delhi-1 iris database indicate that the proposed multibiometric authentication with grouping function based score fusion outperforms the existing transformation-based fusion approaches in literature (e.g., t-norms, symmetric-sum), attaining a correct recognition rate of 95.54%.