Pair of Iris Recognition Using Feedforward Neural Networks

Abstract

Pair of iris recognition is very effective for person
identification due to the iris unique features and the protection of the iris from the environment and aging. In addition it is well suitable to embark upon accidental or ophthalmological disease issue. This paper presents a simple methodology for pre-processing pair of iris images which means both left and right eye of human(instead of
either right or left eye) and the design and training of feedforward artificial neural network for iris recognition system. Three different iris image data partitioning techniques and two data coding are proposed and explored. We also experiment with various number of hidden layers, number of neurons in each hidden layer, input format (binary vs. analog) percent of data used for training vs testing, and with the addition of noise. Our recognition system achieves high accuracy despite using simple data preprocessing and a simple neural
network.