匹配层多生物特征融合识别理论研究.doc

文档介绍：
匹配层多生物特征融合识别理论研究#
李秀艳1,2,刘铁根2,袁臣虎1**
(1. 天津工业大学电子与信息工程学院,天津,300387;
5
10
15
20
25
30
35
40
2. 天津大学精仪学院,光电信息技术科学教育部重点实验室,天津,300072;)
摘要:多生物特征融合识别分为特征层融合、匹配层融合和决策层融合三个层次,其中匹配
层融合策略既能融合若干个单模态生物特征各自的信息量,其融合实现难度相对较小,且融
合效果显著的特点。本文研究了均值融合方法和最大值融合方法的匹配层融合识别理论,推
导了多生物特征融合系统中降低识别错误拒绝率(FRR)、错误接受率(FAR)和等错率(ERR)
识别性能的约束条件,并基于手背静脉、虹膜和指纹三种生物特征进行了融合实验验证,对
于多生物特征识别应用具有指导意义。
关键词:图像处理;生物特征;融合;身份识别;理论研究;约束条件
中图分类号:
Study on Multi-Modal Biometric Theory at Matching Score
Level
LI Xiuyan1,2, LIU Tiegen2, YUAN Chenhu1
(1. School of Electronics and Information Engineering,Tianjin Polytechnic
university,Tianjin,300387;
2. Key Laboratory of Opto-electronics Information and Technical Science of MOE, College of
Precision Instrument and Opto-electronics Engineering, Tianjin University, Tianjin, 300072;)
Abstract: The data fusion in a multi-modal biometric system can be realized at feature extraction level,
matching score level and decision level. At the matching score level, each individual system provides a
matching score and those scores bined to affirm the authenticity of the claimed identity, which
is effective to improve the recognition performance and easy to realize. The data fusion theory based
on Average and Max methods at matching score level was studied in the paper. The constraint
conditions on improving recognition performance including FRR, FAR and