Combining global and minutia deep features for partial high-resolution fingerprint matching

On mobile devices, the limited area of fingerprint sensors brings demand of partial fingerprint matching. Existing fingerprint authentication algorithms are mainly based on handcrafted features, such as minutiae topological structure and ridge patterns. Their accuracy degrades significantly for partial-to-partial matching due to the lack of features. Optical fingerprint sensor can capture very high-resolution fingerprints (2000dpi) with rich details as pores, scars, shape of ridges, etc. These details can cover the shortage of minutiae insufficiency. However, it is challenging to make good use of them, since they are irregular and unstable. In this paper, we propose a novel matching algorithm for such fingerprints by taking advantage of deep learned features. Our model employs a couple of deep convolutional neural networks to learn both high-level global feature and low-level minutia feature. Then we use score level fusion of global similarity and spectral correspondence of minutiae matching. Experiments indicate that our model outperforms several state-of-the-art approaches. (C) 2017 Elsevier B.V. All rights reserved.