Improved Face Authentication Based on Statistical Models: A Frequency Domain Approach

• Activity Number: 20
• Type: Topic Contributed
• Date/Time: Sunday, August 7, 2005 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statisticians in Defense and National Security
• Author(s): Sinjini Mitra*+
• Companies: Carnegie Mellon University
• Keywords: authentication ; biometrics ; face ; phase ; spectrum ; statistical model

Abstract

The modern world has seen a rapid evolution of the technology of biometric authentication, prompted by an increasing urgency to ensure a system's security. The need for efficient authentication systems has skyrocketed since 9/11, and the proposed inclusion of digitized photos in passports shows the importance of biometrics in homeland security today. Based on a person's essentially unique biological traits, these methods are potentially more reliable than traditional methods like PINs and ID cards. This paper focuses on establishing a firmer statistical foundation for face authentication systems and evaluating the accuracy of existing empirical methods. We first present an existing nonmodel-based authentication system based on a linear filter called the Minimum Average Correlation Energy (MACE) filter and describe how simple statistical models can be used to evaluate its performance on large, real-world databases containing diverse images. We then propose a novel mixture model-based approach in the frequency domain by exploiting the well-known significance of phase-in face identification. Some classification results, inference, and associated challenges are discussed.