Workbench Review for Biometrics Face Verification Systems Analysis Studies

Abstract

Computer vision is a promoted field of artificial intelligence (AI) that uses machine learning algorithms to teach computers and systems to derive meaningful information from digital images. Pattern recognition has long been a goal of computer vision, but recently reliable automated face recognition has become a realistic target of biometrics development research. In this paper pattern recognition versus verification systems has been diagnosed, the contribution of classifier analysis to Biometrics Face Verification performance is examined. The discussion covers classification paradigms, the use of multiple observations, and their judicious fusion at the data; it should improve the correct decision performance at the different decision fusions levels. The fusion tasks reported in this development work were carried through fusion of two well-known face recognisers, ICA I and ICA II. Novel strategy based on Likelihood Ratio Fusion within scores is employed, in which face recognition accuracy is optimised and limitations of single recogniser have been reduced. The performance of the analysis studies were tested based on three different face databases and the Dynamic Still Face Databases eNTERFACE2005 (Stylianou et al., 2005), with simulation results demonstrating significant performance achievements.