生物识别系统中基于匹配信息积累的模板自动选择方法

Information accumulation methods for automatic biometric template selection

为了解决生物识别系统因样本类内差异引起的识别性能下降,定义了用于评估模板价值的可代表性和不可替代性概念,并提出了基于认证信息最大化和不可替代权重最大化两种模板自动选择方法。在不同生物特征数据库中进行了实验,各对比实验的等错误率结果验证了方法的优越性。所提方法可以通过降低错误拒绝率显著改善多种类型生物识别系统的识别性能。

For biometric systems,intra-class variations may lead to degradation of recognition performance.This paper defined representativity and irreplaceability to evaluate the value of templates,and proposed two biometric template selection methods.The first algorithm was to maximize the pass-fail information,another one was to maximize the irreplaceable weights.According to the indicator of equal error rate,the experimental results on datasets from different biometric traits demonstrate the superiority.The proposed approaches can improve the performance of biometric systems by decreasing the false rejection rate.