Periocular Biometrics in Head-Mounted Displays: A Sample Selection Approach for Better Recognition

TLDR: A new normalization scheme to align the ocular images and then, a new reference sample selection protocol to achieve higher verification accuracy is proposed and is exemplified using two handcrafted feature extraction methods and two deep-learning strategies.

Abstract: Virtual and augmented reality technologies are increasingly used in a wide range of applications. Such technologies employ a Head Mounted Display (HMD) that typically includes an eye-facing camera and is used for eye tracking. As some of these applications require accessing or transmitting highly sensitive private information, a trusted verification of the operator’s identity is needed. We investigate the use of HMD-setup to perform verification of operator using periocular region captured from inbuilt camera. However, the uncontrolled nature of the periocular capture within the HMD results in images with a high variation in relative eye location and eye-opening due to varied interactions. Therefore, we propose a new normalization scheme to align the ocular images and then, a new reference sample selection protocol to achieve higher verification accuracy. The applicability of our proposed scheme is exemplified using two handcrafted feature extraction methods and two deep-learning strategies. We conclude by stating the feasibility of such a verification approach despite the uncontrolled nature of the captured ocular images, especially when proper alignment and sample selection strategy is employed.