Automated fingerprint pattern classification error analysis

Abstract

Fingerprint pattern classifiers are expensive means to reduce the search space of fingerprints identification systems. The more fingers of the search subject are used for pattern classification, the more search space reduction can be achieved. However, the more fingers of the search subject are used for pattern classification, the higher the classification errors will incur. The use of reference class is a practical way of improving the pattern classification accuracy. Its cost is an important factor in deciding whether the solution is cost effective. It is often the case that the number of comparisons made by the matcher is limited by hardware capacity. Hence, the search space size becomes constrained given the response time requirements for an AFIS. Thus, the utilization of reference classes might be limited.