Gender classification using ear biometrics

Abstract

In this work, ear biometrics has been used for gender classification. Identifying a person as male or female is an interesting problem and is required in many practical applications. The earhole has been considered as the primary reference point. Relative distances (Euclidean distance) have been measured between the ear identification points (ear features) and the ear hole. The ear features considered are outer lobe edge, outer and inner curves of the helix, outer and inner curves of the antihelix and two edges of the concha. We have used an extensive internal database of about 342 samples of male and female ears. The Bayes classifier, K-Nearest Neighbour (KNN) classifier and the neural network classifier have been used for the classification. Overall classification rate of 90.42 % is achieved using KNN classifier. © 2013 Springer.