Principal component analysis for ear-based biometric verification

Abstract

Biometrics is an active research field that is increasingly being integrated into current technology. As a result, more and more people are becoming familiar with biometric technics such as fingerprint or facial recognition. Nevertheless, there are innovative techniques such as ear-based biometrics which are not very well known yet because they are at initial stages of research. In this work, an ear geometry-based biometric verification system oriented to recognition through mobile phones is presented and evaluated. Feature extraction is carried out by means of Principal Component Analysis and feature matching is performed by a distance-based classifier including Euclidean and Eigen distances. The evaluation has been made according to the specifications included in the ISO/IDE 19795 norm following a zero-effort falsification scenario. To this end, a database including ear images of different users has been captured using the camera of a smartphone. The obtained Equal Error Rate matches 1.11% when Euclidean distance is applied for feature comparison and 5.56% in the case of Eigendistance.