Optimization of Neural Networks for the Accurate Identification of Persons by Images of the Human Ear as Biometric Measure

Abstract

This chapter describes the application of modular neural network architecture to improve the recognition of persons using Ear images as a biometric measure [80]. The database used was obtained from the University of Science and Technology Beijing (USTB). We show the results obtained with the modular neural network, its optimization using genetic algorithms and their integration using different methods: Winner Takes All (WTA), type-1 fuzzy integration and fuzzy integration optimized by genetic algorithms. The behavior of the simulations shows a good identification, using the appropriate pre-processing, fuzzy integrators and the best structure found by the genetic algorithm.