ABSTRACT

Multimodality is a promising trend to gain the reliability for biometrics. This paper presents a novel multimodal biometric recognition with information fusion at feature level, called eigen-based feature-level fusion (E-FLF) scheme, for personal authentication. The kernel idea of the proposed E-FLF scheme is to perform the eigen analysis for finding an optimum projection on the feature's cross-energy space by maximizing cross-energy ratio in the new projection space. Simple local and global features extracted from multiple biometric modalities, such as iris, palmprint and face, are considered in this fusion scheme. Different modes of fusion have been implemented to verify the validation of the proposed method. Experimental results reveal that the proposed fusion scheme for fusing features from multimodal biometric traits may improve the recognition performance significantly.

摘要

多模态是获得生物识别可靠性的有前途的趋势。本文提出了一种新的多模态生物特征识别，在特征级进行信息融合，称为基于特征的特征级融合 (E-FLF) 方案，用于个人身份验证。所提出的 E-FLF 方案的核心思想是通过在新投影空间中最大化交叉能量比来执行特征分析以找到特征交叉能量空间上的最佳投影。该融合方案考虑了从虹膜、掌纹和面部等多种生物识别模态中提取的简单局部和全局特征。已经实施了不同的融合模式来验证所提出方法的有效性。