A Computation of Fingerprint Similarity Measures Based on Bayesian Probability Modeling

摘要

One of the primary functions of minutia-based fingerprint recognition algorithms is to compute a similarity measure between two fingerprints. The similarity measure is generally based on the type, angle-difference, and position-difference of corresponding minutiae. This paper proposes a Bayesian probability modeling method for computing the fingerprint similarity measure. The proposed method models the distributions of the angle-differences and the position-differences according to the type-difference between the corresponding minutia pairs. Also, the similarity measure is represented by a posteriori probability assuming that their distributions are statistically independent. This method has been applied to two different cases of fingerprint verification and demonstrated its effectiveness by reducing the equal error rates with the average of 40%.