Reda ADJOUDJ, Aouad BOUKELIF
Artificial Neural Network & Multilevel 2-D Wavelet Decomposition Code-Based Iris Recognition

Abstract.
An iris recognition involves analyzing features found in the colored ring of tissue that surrounds the pupil. This biometric has the potential for higher than average template-matching performance. Easy of use and system integration has not traditionally been strong points with iris scanning devices but as new products emerge, improvements should be expected. This document demonstrates how an iris recognition system can be designed by artificial neural network as a matching/recognizing algorithm, which use Multilevel 2-D wavelet decomposition code of iris image. Note that the training process did not consist of a single call to a training function. Instead, the network was trained several times on various ideal inputs and noisy images coded by Multilevel 2-D Wavelet decomposition, these images contain the iris. In this case, the system is scored a perfect recognition rate of 98.55% and 87.11% for the two databases, compared to the case when the scanned image of iris is applied directly, where the rate was 89.90% and 73.60%. Finally, the training of the network on different sets of noisy images forced the network to learn how to deal with noises, a common problem in the real world.