Proposed Method for Evaluating Voice Authentication Systems

Abstract

Security is an important part of our daily lives. This thesis focuses on the biometric component of security. Identity authentication for security is improved when voice authentication is augmented by other identity authentication components. This thesis presents a method for evaluating existing algorithms. This thesis details a method to compare pass/fail rates of various voice authentication algorithms. The method provides a formal mechanism to evaluate algorithms for different security levels. Different security levels have different security requirements. The application sets the security level. The level for entering a level for entering a high security research facility is more stringent for imposters at the risk of inconveniencing employees. The method also can be used to optimize setting the threshold of the algorithm. The developed method works even when the algorithm is less than ideal.