Iris Feature Reduction Using Bottleneck Neural Network

Abstract

The human iris is one of the best biometric features in the human body for pattern recognition due to its stability, invariant and distinctive features for personal identification. The proposed Iris Recognition System (IRS) is consisting of four major fundamental stages: image preprocessing, feature extraction, feature reduction and image pattern recognition. In the third stage the Bottleneck Neural Network (BNN) is used as a reduction algorithm which gives the reduced iris feature set that is recognized with Support Vector Machine (SVM) algorithm. The accuracy of using BNN with SVM is increased from (67%- 100%) for variable number of persons (10-100).