Fingerprint recognition by deep neural networks and fingercodes
Alper Baştürk, Nurcan Sarikaya Basturk, Orxan Qurbanov · 2018
In this study, fingerprints are recognized using a deep neural network. Two autoencoders and a softmax layer are used to build the deep neural network structure. Required time for training has been reduced using graphics processor. The data set used is created using four different images obtained from each of 165 different fingers. This data set is extended by using additional rotated versions of fingerprint images. By filtering these images with directional Gabor filters, feature vectors of the fingerprint images (fingercodes) are obtained. Finally, a deep neural network is trained by using this feature data set. Obtained simulation results show that deep neural networks can be used for recognition of fingerprints.