Ear Biometrics: A Comprehensive Study of Taxonomy, Detection, and Recognition Methods

Abstract

Due to the recent challenges in access control, surveillance and security, there is an increased need for efficient human authentication solutions. Ear recognition is an appealing choice to identify individuals in controlled or challenging environments. The outer part of the ear demonstrates high discriminative information across individuals and has shown to be robust for recognition. In addition, the data acquisition procedure is contactless, non-intrusive, and covert. This work focuses on using ear images for human authentication in visible and thermal spectrums. We perform a systematic study of the ear features and propose a taxonomy for them. Also, we investigate the parts of the head side view that provides distinctive identity cues. Following, we study the different modules of the ear recognition system. First, we propose an ear detection system that uses deep learning models. Second, we compare machine learning methods to state traditional systems' baseline ear recognition performance. Third, we explore convolutional neural networks for ear recognition and the optimum learning process setting. Fourth, we systematically evaluate the performance in the presence of pose variation or various image artifacts, which commonly occur in real-life recognition applications, to identify the robustness of the proposed ear recognition models. Additionally, we design an efficient ear image quality assessment tool to guide the ear recognition system. Finally, we extend our work for ear recognition in the long-wave infrared domains.