Model-Based Human Ear Localization and Feature Extraction

Abstract

Nowadays the viability of ear-based biometric identification and the uniqueness of ears is beyond question, but reliable technical solutions as the basis for a successful commercial application have yet to appear. As opposed to face recognition, in which amodel-based approach is widely used, surprisingly little effort has been put into using ear models in automatic identification, even though ear shape is more robust than facial characteristics, being unaffected by emotional expressions or other changes like facial hair or glasses. In this paper we will introduce our latest work in a model-based approach for ear identification, featuring schemes for both ear localization and feature extraction. The first tests of our proposed solution have proved that the method is accurate and fast enough to allow its application in intelligent video surveillance, where people can be remotely identified and their identity can be continuously tracked using the ears visible on video surveillance camera images.

Keywords:

• Ear biometrics
• image processing
• model-based identification
• active contours
• intelligent video surveillance