A Proposed Biometric Technique for Improving Iris Recognition

Abstract

Recently, the Iris Recognition system has been considered an effective biometric model for recognizing humans. This paper introduces an effective hybrid technique combining edge detection and segmentation, in addition to the convolutional neural network (CNN) and Hamming Distance (HD), for extracting features and classification. The proposed model is applied to different datasets, which are CASIA-Iris-Interval V4, IITD, and MMU. For validating the results of the proposed models, detailed modeling and simulation procedures took place using the mentioned three datasets. A comparison between the obtained results from the current work and published results from open literature was carried out as well. The Proposed Biometric Technique showed desirable recognition accuracies of 94.88% based on applying HD on CASIA, 96.56% based on applying CNN on IITD, and 98.01% based on applying CNN on MMU. The obtained accuracies illustrated the superiority of such a classifier compared to other classifiers used in the published literature.