Modular neural networks with type-2 fuzzy integration for pattern recognition of iris biometric measure

Abstract

This paper presents a new modular neural network architecture that is used to build a system for pattern recognition based on the iris biometric measurement of persons. In this system, the properties of the person iris database are enhanced with image processing methods, and the coordinates of the center and radius of the iris are obtained to make a cut of the area of interest by removing the noise around the iris. The inputs to the modular neural network are the processed iris images and the output is the number of the identified person. The integration of the modules was done with a type-2 fuzzy integrator at the level of the sub modules, and with a gating network at the level of the modules. © 2011 Springer-Verlag.