Title :
Improved neural network-based recognition of irises with sector and block partitioning
Author_Institution :
Fac. of Inf. Technol., UAE Univ., Al Ain, United Arab Emirates
Abstract :
High performance biometrics helps in reliably identifying persons for access authorization and other purposes. Iris recognition is very effective in identifying persons due to the iris´ unique features and the protection of the iris from the environment and aging. We focus on the design and training of a feed-forward artificial neural network for high-performance iris recognition and investigate the impact of various image data partitioning techniques on the recognition accuracy of the biometric system. Several iris image data partitioning techniques are proposed and explored. Simulation results reveal that 100% recognition accuracies with sector and block data partitioning techniques can be reached, improving on our prior work results.
Keywords :
authorisation; biometrics (access control); feedforward neural nets; iris recognition; learning (artificial intelligence); access authorization; block partitioning; feedforward artificial neural network; high performance biometrics; image data partitioning techniques; neural network based iris recognition; sector partitioning; training; Accuracy; Artificial neural networks; Biological neural networks; Iris recognition; Neurons; Strips; Training;
Conference_Titel :
Innovations in Information Technology (IIT), 2011 International Conference on
Conference_Location :
Abu Dhabi
Print_ISBN :
978-1-4577-0311-9
DOI :
10.1109/INNOVATIONS.2011.5893819
