APPLYING MULTIPLE INSTANCES AND MATCHING ALGORITHM IN FACE VERIFICATION SYSTEM

摘要

This paper investigates the benefits of a multi-modal biometric system for a single biometric. A multi-modal biometric system generally means the multiple biometrics system (e.g. combining face and iris). But, in this paper we apply multiple instances (capturing multiple impressions of the same face) and multiple matching algorithms (combining different approaches to matching of faces) for a single face biometric. For an experiment, we design the face verification system based on PCA and evaluate the system with our realistic evaluation model and database. As a result, the performance of the proposed approach outperforms that of a single instance and matching approach by about four times. This result proves that the proposed approach is one of the essential considerations for building a real-world face verification system.