多生物特征融合发展现状及其展望  

Survey on Multi-feature Fusion in Biometrics and its Developments

作　　者：柳欣[1,2] 耿佳佳[1] 钟必能[1,2] 杜吉祥[1,2] 

机构地区：[1]华侨大学计算机科学与技术学院,福建厦门361021 [2]厦门市模式识别与计算机视觉重点实验室,福建厦门361021

出　　处：《小型微型计算机系统》2017年第8期1792-1799,共8页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61300138;61572205;61673185)资助;福建省自然科学基金项目(2017J01112)资助;华侨大学中青年创新人才培育项目(ZQN-309)资助

摘　　要：多生物特征融合技术利用多个可鉴别的身份信息,在一定程度上能弥补单一生物特征识别的不足,从而可以有效达到降低误识率和实现高精度鉴别系统的要求.多生物特征融合为信息化社会日益增长的保密和安全需求提供了较好的解决方案,其相关理论与方法已成为智能信息处理的一个重要研究课题.本文围绕多生物特征识别技术,选择传感器为切入点,特征结构为分支,分别从同源同构、同源异构、异源同构、异源异构四个方面介绍多生物特征融合的典型方法及其研究现状,并在此基础上介绍了深度学习在多生物特征融合中的最新应用现状,并对其发展趋势作了一定展望.Multi-biometrics fusion incorporating multiple distinguishable identity information, is able to remedy the shortcomings within the single biometric recognition system and often holds a strong ability to reduce the false accept rate. As a consequence, the high precision identification system can be realized. Evidently ,multi-feature fusion in Biometrics provides a better solution to the increasingly demand in privacy and security, and its related theory and methodology have become an important research topic in the intelligent information processing. In this paper, we select the sensor as the breakthrough point and utilize the feature type as the model branch to introduce the representative multi-feature fusion approaches and elaborate their current research states, in which the four aspects with respected to the homogenous features and heterogeneous features of single biometric source, homogenous features and heterogeneous features of multiple biometdc sources are comprehensively grouped for illustration. Finally, we provide some state-of-the-art application status for deep learning based multi-biometrics, and draw a technical conclusion in the future.

关 键 词：多生物特征融合 同构 异构 深度学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]