Ear recognition system using adaptive approach Runge-Kutta (AARK) threshold segmentation with cart classifier

Abstract

In the recent years, researchers have drawn attention towards the human external ear due to being a new class of relative stable biometric. The significant of this paper is to employ an approach in biometric ear recognition. Previously, there have been a number of researchers carried out on ear recognition systems and various methods have achieved good performances. This paper has implemented a modern approach for segmentation of ear images that includes an adaptive approach Runge-Kutta (AARK) segmentation and CART (Classification and Regression Tree) classifier. AARK segmentation is typically used to locate objects and boundaries of ear images. It also increases the speed of segmentation, good shape matching and it results in good shape connectivity. CART classifier is used to produce better accuracy when performing ear classification. Also, this paper consists with preprocessing, ring projection, information normalization and feature extraction of DWT (1D-Discrete Wavelet Transform). The obtained results show that the significance of the proposed method carry out advanced ear recognition techniques, and increases the PSNR values, and consequently improve the accuracy rate.