Abstract
Fingerprints are considered as an efficient and viable means of biometrics. An automatic fingerprint verification system (AFVS) can automatically verify an individual based on minutiae matching. The fingerprint verification contains two main stages, fingerprint classification and fingerprint matching. In the stage of fingerprint classification, fingerprints are classified into four classes to reduce the following fingerprint matching. The four classes are arch, left loop, right loop and whorl. In the stage of fingerprint matching, fingerprint minutiae must be detected for matching. A minutiae pattern consists of ridge endings and bifurcations. Based on the differences of distance and direction of minutiae, we compute matching scores of all pair of fingerprints. If one of the matching scores with the other 3 fingerprints acquired from the same individual is among the top 3 places, the fingerprint is accepted for the verification. The error rate of our AFVS tested on the database DBLI101 is 17.08% (69/404) and the other database of 28 individuals gains the error rate of 0.98% (1/112)