Multimodal eye biometric system based on contour based E-CNN and multi algorithmic feature extraction using SVBF matching

Abstract

Recent advancement in biometric system prefer multimodal biometric system instead of single biometric system to overcome challenges faced by unimodal biometric system such as intra class variation, noise sensitivity, non universality, spoofing attack, etc. Most of the existing iris biometric systems are dependent on ideal condition which needs user cooperation during image acquisition with help of NIR camera to avoid noise. Such system performance significantly degrades when images are taken under visible light without user cooperation called unconstrained environment. Proposed multi modal eye biometric system provides improvement in segmentation accuracy using entropy based convolution neural network (E-CNN) based on contour feature. It also reduces the time required for segmentation up to 0.9second. Multi algorithmic feature extraction for color, texture features of iris and pupil and Y-shaped features of sclera exploit the improvement in feature extraction performance. Proposed feature level support value based fusion (SVBF)approach provide better performance of multimodal eye biometric system and achieves good improvement in recognition accuracy 93.33% and 97% when framework is tested for the images taken from the MMU and UBIRIS.v2 unconstrained eye image database respectively as compared to the related competing approaches.