Fast and Accurate Fingerprint Recognition in Principal Component Subspace

Abstract

In the case of fingerprint-based person recognition, the most widely used discriminating features are minutiae (end points and bifurcations of ridges). Majority of fingerprint matching algorithms are dealing with comparing the parameters directly derived from or relative to minutiae points extracted from the templates. Hence eventually fingerprint matching based on minutiae can be reduced to a 2D point set matching problem. Various security pitfalls like impersonation using one’s minutiae coordinates and performance issues related to enhancement as well as spurious minutiae removal are obvious in such a system. Certain non-minutiae based schemes are able to give acceptable performance at the cost of increased complexity which results in increased execution time. In order to overcome these issues, we propose a simple yet efficient and faster fingerprint alignment and matching scheme based on statistical features which will not reveal the unique local features of the template. Proposed matching technique is based on the weighted similarity score obtained by comparing the principal component subspaces of fingerprint templates. Proposed method also utilizes an alignment scheme based on principal components calculated for the 2D coordinates of fingerprint region with minimal overhead without any helper data.