چكيده به لاتين
Considering the increasing need of today’s societies to more information security, using biometric technology has become more critical. In the past decade, uni-biometric traits has shown new weaknesses such as lack of uniqueness, lack of availability, fraud and so on, cause muti-biometrics to become more attractive. In fusion levels of biometric information sources, score level fusion is the most popular level, because of appropriate percision and low time complexity.
In this research, our purpose is combination of two or three biometric traits in score level. For each of these biometric traits, appropriate matcher has been chosen. Every matcher has its specific scores in specific distribution and domain.
In order to acquiring more optimal methods in comparison with current ones, we will suggest two main approach; using combination of discriminant features and using specific classifiers for imbalance data. We will test our combination methods using different scenarios in two multi-biometric dataset and will utilize confidence interval for significance level of resuts.
It has been shown in result section that proposed feature combination method will act much better than other methods. These results has been inferred of the ROC diagram and also FRR value in constant FAR. Finally we will examine proposed methods in comparison with state of the art methds show its better
