Enhance User Authentication via Deep Learning on Face and Periocular Biometrics

For identification purposes, a specific biometric characteristic has to be invariant over time and unique for each individual whom it can be measured for. Some significant disadvantages associated with biometrics consist of voiceprints, fingerprints, images, signatures, and patterns of blood vessels in the retina. Despite their being inexpensive, easily acquired, and readily forged, signatures and photographs cannot be mechanically recognized with precision. However, as the human iris is an internal body organ of the eye and insulated from the outside environment and easily visible up to one meter, it makes an ideal biometric for an automated, fast, and reliable identification system. The most dependable and accurate biometric identification technology available today is Iris recognition. This paper utilizes mathematical pattern recognition techniques applied on photographs of every person's unique, complicated, and random iris patterns for the biometric identification process, which is commonly known as "iris recognition". This work focuses on developing a face and iris recognition system with deep neural networks, Gabor filtering, and the Grassmann algorithm to be able to differentiate the iris, eye, and face regions. Results indicate the success of the proposed strategy for iris-based biometric identification.