A Novel Approach for Biometric Authentication System Using Ear, 2D Face and 3D Face Modalities

Abstract

Biometric system using face recognition is the frontier of the security across various applications in the fields of multimedia, medicine, civilian surveillance, robotics, etc. Differences in illumination levels, pose variations, eye-wear, facial hair, aging and disguise are some of the current challenges in face recognition. The ear, which is turning out to be a promising biometric identifier having some desirable properties such as universality, uniqueness, permanence, can also be used along with face for better performance of the system. A multi-modal biometric system combining 2D face, 3D face (depth image) and ear modalities using Microsoft Kinect and Webcam is proposed to address these challenges to some extent. Also avoiding redundancy in the extracted features for better processing speed is another challenge in designing the system. After careful survey of the existing algorithms applied to 2D face, 3D face and ear data, we focus on the well-known PCA (Principal Component Analysis) based Eigen Faces algorithm for ear and face recognition to obtain a better performance with minimal computational requirements. The resulting proposed system turns out insensitive to lighting conditions, pose variations, aging and can completely replace the current recognition systems economically and provide a better security. A total of 109 subjects participated in diversified data acquisition sessions involving multiple poses, illuminations, eyewear and persons from different age groups. The dataset is also a first attempt on the stated combination of biometrics and is a contribution to the field of Biometrics by itself for future experiments. The results are obtained separately against each biometric and final decision is obtained using all the individual results for higher accuracy. The proposed system performed at 98.165 % verification rate which is greater than either of the dual combinations or each of the stated modality in a statistical and significant manner.