Methodology for partial fingerprint enrollment and authentication on mobile devices

Abstract

The reduced platen area of fingerprint sensors in mobile devices results in acquisition of partial fingerprints. Existing fingerprint enrollment schemes for small area sensors are tedious and have an uncertainty of complete finger coverage during scanning. We propose a novel enrollment protocol for small area rectangular sensors that maximizes finger coverage within few scans. Also, due to presence of insufficient minutiae, accuracy of minutiae-based fingerprint matching algorithms degrades significantly when applied for partial-to-partial fingerprint matching. Instead, we propose a matching algorithm that utilizes multi-scale texture descriptors, namely, Accelerated KAZE (A-KAZE). Experiments on FVC 2000, 2002 and in-house databases indicate that A-KAZE gives promising accuracy. On a Samsung Galaxy Note II N7100 (Quad-core 1.6 GHz, 2GB RAM), average time taken for template generation and 1×1 matching of fingerprint of size 237×117 pixels is 86 ms and 19 ms respectively.