Event-related pupillary response-based authentication system using eye-tracker add-on augmented reality glasses for...

Area of Science:

• Biometrics
• Human-Computer Interaction
• Neuroscience

Background:

• Biometric authentication is crucial for security.
• Existing methods often require physical contact or cumbersome sensors.
• Augmented reality (AR) offers a novel platform for noncontact interaction.

Purpose of the Study:

• To develop and evaluate a noncontact authentication system using event-related pupillary response (ErPR) in an AR environment.
• To compare the efficacy of ErPR with event-related potential (ERP) for biometric authentication.
• To assess the performance of machine learning classifiers for ErPR-based identification.

Main Methods:

• Thirty participants viewed familiar and unfamiliar faces in an AR setting.
• Event-related pupillary response (ErPR) and event-related potential (ERP) were recorded.
• Linear support vector machine and quadratic discriminant analysis classifiers were employed for authentication.

Main Results:

• Both ERP and ErPR amplitudes were significantly higher for familiar faces.
• An ERP-based system achieved perfect accuracy.
• An ErPR-based system using quadratic discriminant analysis demonstrated 97% accuracy with low false acceptance and rejection rates.
• Good agreement was observed between ERP and ErPR measurements.

Conclusions:

• ErPR-based authentication is a viable, noncontact method for person identification.
• This technology offers a low-cost, noninvasive, and easily implementable solution for AR environments.
• The system eliminates the need for sensor attachment, enhancing user convenience.