Human identification verification from biometric dorsal hand vein images based on deep learning generative adversarial networks

For the purpose of this research, biometric hand vein technique was used to recognize individual identity. In this study, authentication was done using the hand "dorsal venous network" vein. For this, a new method DL-GAN has been developed by combining deep learning and productive adversarial network (GAN). With the DL-GAN authentication method, the authentication rate has been increased. Many biometric methods, including vein information on the hand, are used to identify people. Information on how the superficial subcutaneous shallow vein, which can be detected by various technological methods such as infrared camera, is placed in the skin tissue is a new method of identifying people. Vein locations on the hand are idiosyncratic and are a good new option for identifying individuals. The vascular system on the upper hand and near the wrist, the subcutaneous artery on the upper hand, the veins and the vascular network (metecarpal, venous network, basilic) are used to authenticate. In order to recognize the biometric hand vein image, appropriate software coding has been made for the MATLAB 2020a programming language. The developed DL-GAN method has been tested on two separate databases, Jilin University – hand held database and 11K hand held database. The results of the experiments performed on the hand-held vein data set, on the other hand, show that the DL-GAN method has reached 98.36% accuracy and has an error rate of 2.47% and a standard accuracy of 0.19%. The experimental results in the second data set, on the other hand, have an accuracy of 96.43%, an equal error rate of 3.55% and a standard accuracy of 0.21%. The improved DL-GAN method obtained better results from biophysical methods such as LBP, LPQ, GABOR, FGM, BGM and SIFT compared to the same databases. Keywords: Identification, biometrics, detection system, dorsal hand veins, deep learning, generative adversarial network.