On ear biometrics

Abstract

Today the most successful biometric based identification technologies such as fingerprint, iris, retina, palm and face recognition are used worldwide in both criminal investigations and high security facilities. These technologies are well-studied, but research shows they have many drawbacks which decrease the success of the methods applied. Ear images are not affected by emotional expression, illumination, aging, poses and alike. In this study principal component analysis (PCA), fisher linear discriminant analysis (FLDA), discriminative common vector analysis (DCVA), and locality preserving projections (LPP) were applied to ear images for personal identification. The error and hit rates of four algorithms were calculated by random subsampling and k-fold cross validation.