Innovative Methods and Applications of Soft Computing in Ear Bio Metrics

Abstract

vii newlineAbstract newlineHuman ear is considered to be a powerful non-ambiguous articulation for biometric data. That is extensively used for personal identification in recent years. The evidence proclaiming the uniqueness of the ear was produced empirically way back in 1890 by a French criminologist who called human ear to be spoken portrait . Despite such a long standing evidence in support of uniqueness of shape of ears, the momentum in ear biometrics related research has picked up only recently. Thus far, entirely automated system for ear biometric based person recognition system has not been tried, though there are several significances of ear as a biometric articulation. Applications of soft computing techniques, machine learning approaches and machine vision paradigms are particularly scanty. This is exactly the motivation behind the research work presented in this thesis. newlineThe work presented in this thesis mainly hinges on five pronged approach; newlinei. Elicitation of novel features treating ear to be a projected planar surface which is first of its kind. newlineii. Development of personal identification system making use of the novel features. newlineiii. Application of hard limiting and soft computing paradigms in grouping the ears leveraging uniqueness of ear features from person to person. Then, designing the group based person identification system. newlineiv. Feature reduction using hard limiting and evolutionary computing paradigms. newlinev. Augmenting convex hulls to capture the shape of the ear, and to obtain rotation or orientation invariant features followed by design of rotation invariant person identification system. newlineFor the purpose of this research, images were garnered by two exclusively live photo sessions involving more than 600 subjects mostly students and teachers in the age group 21-50 years. In all, about 1500 ear images were gathered in to database gallery. newlineAs a first component of this research, novel feature set based on the physical property of the planar surface, namely the moment of inertia (MI) were extracted from ear i