Deep Learning Based Impostor Detection By Invariant Features From Nir Finger Vien Imaging

Biometrics mostly used as personal identification accordingly securing a client against the unapproved utilization of his or her identity. Acquiring biometric information is getting to be simpler. Smartphones and other advanced technologies exist from which biometric data can be collected easily without the knowledge of others. Finger vein authentication is a method for biometric verification that depends on a vein pattern, which is located beneath the human finger's skin. Veins are covered with skin that cannot be copied by others. In this research, our focus is on the invariant features of finger veins. We have extracted invariant features from various state of the art deep learning techniques and then classified with multiclass SVM. For this purpose, we have used publicly available finger vein image databases. The performance has been evaluated by various evaluation metrics and comparative analysis of various deep learning approaches has been presented to describe the performance of these models on the said data set.