Performance Analysis of the ARIA Adaptive Media Processing Workflows using Colored Petri Nets
Electronic Notes in Theoretical Computer Science (ENTCS)
Addressing the vulnerabilities of likelihood-ratio-based face verification
AVBPA'05 Proceedings of the 5th international conference on Audio- and Video-Based Biometric Person Authentication
| Hi-index | 0.00 |
Techniques that treat the face holistically as a vector of pixel values, which we refer to as a monolithic representation, are still widely considered state of the art for the task of face verification in literature. Recently good performance has be attained in the task of face verification, using Gaussian mixture models (GMMs), via estimating a parts (i.e. image patch) face model; where the shape (i.e. spatial) information is largely ignored. In this paper we postulate that the characteristics current algorithms employ for verifying a face using monolithic and parts representations differ and are in many ways symbiotic; lending themselves to synergetic combination and improved verification performance. Results are presented on the BANCA database that demonstrate excellent verification performance in the presence of many common real world variabilities (e.g. camera degradation, minor pose variabilities, some changes in background and lighting).