On Distinctiveness and Symmetry in Ear Biometrics

Previous works show that human ears can be used for identification, gender classification, and kinship verification have investigated whether there is a symmetry between person's ears; however, the performances been less than satisfactory. Our paper extends analysis of classification on ear images analyses bilateral ears, in both cases aiming to determine parts from which recognition derived. We use model-based deep learning methods capitalize structure performance, respectively. consider rotation under an affine transformation, by modelling as flat plane attached head. address question it possible given image one ear, person then recognized their other ear. Such symmetry-based strategy could reduce constraints applications biometrics. recognise with 90.9% success rate rim (the upper helix lobe) dominates performance. To investigate symmetry, we compare mirrored version achieve 93.1% CCR, current state-of-the-art, important regions different those determined gender. extend construct two groups images, contains same subject subjects. The 100% CCR confirms existence subject's ears. By these approaches actually high chance exists would prudent systems concentrate inner rather outer