Author : Miss.Vishakha Shinde 1
Date of Publication :7th December 2016
Abstract: Since last many years it has been observed in contrast to other physiological characteristics the iris pattern have a rich and wonderful structure with the full of complex structure. In between the pupil and sclera iris is present as coloured circular part of the eye. There are Many approaches have been developed for the recognition of iris. It is the process of recognizing a person’s as unique identity by analyzing the appropriate pattern of his/her eye. Artificial neural networks have dimensional and multimodal approach which is usually polluted by noises and missing data. One of the IRIS recognition system developed which used local histogram and image statistics method but it failed to locate boundaries of IRIS and also optimization problem which is related to the process of weight training. We are using feed forward neural network. Artificial neural networks (ANNs) are computational modelling tools that are defined as structures comprised of densely interconnected adaptive simple processing elements. They are able to perform massive parallel computations for data processing and knowledge representation. One more advantage is that a learning algorithm is significantly simplified when the RNN model is of the feed-forward type compared to the recurrent type. A RNN model with multiple classes of signals is introduced that can be used in applications associated with the concurrent processing of different streams of information, such as colour image processing.To tackle the complexity of ANN training problem, meta-heuristic optimization algorithms such as genetic algorithm, particle swarm optimization and ant colony optimization have been highly proposed to search for the optimal weights of the network. Here we propose new optimization technique to overcome the existing problems. This technique is BMO (Bird Mating Optimization) which is population based metaheuristic search method which tries to imitate mating ways of bird species for designing optimum searching techniques. Metaheuristic algorithms do not use any gradient information, and have more chance to avoid local optima by sampling simultaneously multiple regions of search space.This technique is applied for weight training of neural networks for clarification of human iris.
Reference :