Exploiting the Deep Learning with Fingerphotos to Recognize People

Abstract

Biometrics based user authentication for mobile devices are now being commonly popular. In this paper, an authentication approach is proposed based on recognizing images of fingerphotos. Dual Deep Fingerphoto Learning (DDFL) is suggested. It consists of two Deep Learning (DL) networks. The first DL network is utilized to recognize a fingerphoto image of an index finger. The second DL network is used to recognize a fingerphoto image of a middle finger. The outputs of both networks are fused together, then, the result is benchmarked. The proposed approach is evaluated by employing fingerphoto images from the IIITD smart phone fingerphoto database (Version 1). Experimental results yield superior performance of 97.66% for our suggested method.