Periocular Biometric Authentication Methods in Head Mounted Display Device

Background/Objectives: Recently, the use of Virtual Reality (VR) devices has increased and their content has also diversified. Therefore, content handling personal information is increasing, and a personal authentication method is needed. Currently, many VR devices are on the market; however, there is no method for personal authentication. Methods/Statistical analysis: We acquire an eye image via an infrared camera attached inside a Head Mounted Display (HMD) for a VR experience, and propose a periocular biometric authentication method utilizing the eye image. The proposed method does not utilize high frequency components of the image, such as iris recognition; thus it has an advantage in that the recognition speed is fast, and the quality of the image is minimally affected. We used L1 distance, Local Binary Pattern (LBP), and Scale Invariant Feature Transform (SIFT) matching methods for eye image comparisons. In the matching process, a method for considering movement in horizontal and vertical directions was used to compensate for the position variation of the image. Findings: Experimental results showed that the Equal Error Rate (EER) was the best at 6.83% for matching through the L1 distance. However, from a security viewpoint, it is confirmed that a False Rejection Rate (FRR) of approximately 10% is obtained when the False Acceptance Rate (FAR) is reduced to 0% through threshold adjustment. This result indicates that the proposed method can be fully utilized as a biometrics method for personal authentication. Improvements/Applications: The proposed method is expected to be used as a biometric for personal authentication in existing HMD environments because it shows excellent performance with an EER of 6.83%, even when processing low frequency eye image components. Future research will investigate methods to improve in case of closed eye.