HUMAN EAR RECOGNITION USING CONVOLUTIONAL NEURAL NETWORK

Abstract

Today's digital world, security plays an important role in everyday process in computer. The existing security levels are hacked by someone at anytime. A biometric based security system is expected to fulfil user's demand such as low error rates, high security levels, possibility of fake detection etc. This paper proposes an efficient ear based recognition technique using Convolutional Neural Network (CNN). It is non –intrusive methodology and attributes are probably the most common biometric feature used by humans to recognize one other. There are many advantages of using the ear as a source of data for human identification. An ear biometrics system consists of ear detection and ear recognition (authentication and identification) modules. In this paper, we used AMI (translated from Spanish as Mathematical. Analysis of Images) dataset. These images were captured using Nikon D100 camera. Once the pre-processing of image gets completed, the images are converted into one dimensional space. LBP method can be used for classification, recognition and detection of original image pixel further, histogram is used to obtain more discrimination features.  In an existing work, researchers used K-Nearest Neighbours (KNN) method. It doesn’t work with large number of dataset and also it is difficult to calculate the distance in each dimension. The aim of the proposed system is to improve the security level with the help of Convolutional Neural Network (CNN). The proposed system gives the overall accuracy of 98.99 %.