Humans Verification by Adopting Deep Recurrent Fingerphotos Network

Abstract

Fingerphoto can be considered as one of recent and interesting biometrics. It basically means a fingerprint image that is acquired by a smartphone in contactless manner. This paper proposes a new Deep Recurrent Learning (DRL) approach for verifying humans based on their fingerphoto image. It is called the Deep Recurrent Fingerphotos Network (DRFN). It compromises of input layer, sequence of hidden layers, output layer and essential feedback. The proposed DRFN sequentially accepts fingerphoto images of all personal fingers. It has the capability to change between the weights of each individual fingerphoto and provide verification. A huge number of fingerphoto images have been acquired, arranged, segmented and utilized as a useful dataset in this paper. It is named the Fingerphoto Images of Ten Fingers (FITF) dataset. Average accuracy result of 99.84 % is obtained for personal verification by exploiting fingerphotos.