Neural Network Classification for Iris Recognition Using Both Particle Swarm Optimization and Gravitational Search Algorithm

Abstract

This paper introduces an iris classification system using FFNNGSA and FFNNPSO. This iris identification system consists of localization of the iris region, normalization, feature extraction and then classification as a final stage. A Canny Edge Detection scheme and a Circular Hough Transform are used to detect the iris boundaries. After that the extracted IRIS region is normalized using Daugman rubber sheet model. Next, Haar wavelet transform is used for extracting features from the normalized iris region then the feature matrix is reduced using the principle component analysis (PCA). Finally, both particle swarm optimization (PSO) and gravitational search algorithm (GSA) are used for training a forward neural network to get the optimum weights and biases. The results showed that training the feed-forward neural network by GSA is better than training it by PSO in an iris recognition system.