Development of a Multi-Intance Fingerprint Based Authentication System

This paper presents a multi-instance fingerprint authentication system, which is immune to various problems encountered in unimodal systems. During enrolment all available fingerprints are captured and their unique features extracted, processed and stored in a database alongside the bio-data of individual subjects. Unique features of fingerprint were extracted using the crossing number (CN) technique which acquires the ridges ending and bifurcation from a single pixel image by examining the local neighborhood of each ridge pixel using a 3 by 3 window. Authentication requires two fingerprints that are randomly requested from the user from which features are then extracted and compared with those in the database template respectively to determine a match score from each. Each match score is then fused together using a weighted sum rule. The fused score is then compared to perform the final match. The experimental result shows an average runtime of 3.82 seconds for verification of subjects and accuracy of 98.67% on an AMD E1-2100 PC with 4GB ram. The developed software possesses a fault tolerance capability and flexibility of switching from multiinstance to a multi-trait with both modules working independently of one another. The model presented in this research work will mitigate the high incidences of impersonation and sharp practices perpetuated by people in various public and private establishments.