Multimodal Biometric Authentication Systems using Deep Learning
R. Aarthy · International Journal for Research in Applied Science and Engineering Technology · 2021
The use of biometrics for identification operations needs that a specific biometric issue is distinctive for each person that it is calculated, which it's invariant over time.Biometrics similar to signatures, photographs, fingerprints, voiceprints, and retinal vas patterns all have noteworthy drawbacks.Though signatures and images are low cost and simple to get and store, they're impossible to spot automatically with assurance and are simply forged.The human iris, on the other hand, is an interior organ of attention and yet protected against the external environment, however, it is easily visible from among one meter of distance makes it an ideal biometric for an identification system with the benefit of speed, responsibility, and automation.In this work, it's planned to implement a face and iris recognition system, wherever unvaried closest purpose formula (ICP) and deep neural network is employed to section the face, eye, and iris region.In this proposal, the system propose a novel and strong approach for periocular recognition and feature extraction.In the approach, the face is detected in real-time face images which are then aligned and normalized.The proposed system utilizes an entire strip containing both the eyes as a periocular surface.For feature extraction, the model computed the magnitude responses of the image filtered with a filter bank of harder Gabor filters.Feature dimensions are reduced by applying the Grassmann algorithm.Hence, the system produces High-level security and the identification accuracy is well-maintained.Also, the real-time camera-based implementation and readability of multiple features at the time are focussed.The results show that the proposed method is effective for multi-modal biometric recognition.