An Efficient Algorithm for Fingercode-Based Biometric Identification

With the emerging trend of incorporating biometrics information in e-financial and e-government systems arisen from international efforts in anti-money laundering and counter-terrorism, biometric identification is gaining increasing importance as a component in information security applications. Recently, fingercode has been demonstrated to be an effective fingerprint biometric scheme, which can capture both local and global details in a fingerprint. In this paper, we formulate fingercode identification as a vector quantization (VQ) problem, and propose an efficient algorithm for fingercode-based biometric identification. Given a fingercode of the user, the algorithm aims to efficiently find, among all fingercodes in the database of registered users, the one with minimum Euclidean distance from the user’s fingercode. Our algorithm is based on a new VQ technique which is designed to address the special needs of fingercode identification. Experimental results on DB1 of FVC 2004 demonstrate that our algorithm can outperform the full search algorithm, the partial distance search algorithm and the 2-pixel-merging sum pyramid based search algorithm for fingercode-based identification in terms of computation efficiency without sacrificing accuracy and storage.