摘要
生物特征认证技术作为一门新兴技术现已应用于多个领域,然而单一模态的生物特征认证技术易受外界因素影响,为了提高身份认证的鲁棒性和安全性,文中提出一种基于最小最大概率机的多生物特征融合算法。该方法首先对人脸和声纹分别进行特征提取及匹配,然后在匹配层对两种单生物特征的匹配得分值进行了融合。通过在Vidtimit多模态库上的实验结果表明,融合系统的等错误率降低到0.97%,证明了该融合算法的有效性。
As a new technology,biometric authentication technology has been applied in many areas. However,the uni-modal biometric authentication technology is susceptible to external factors. In order to improve the robustness and security of identity authentication,we proposes a multimodal biometric fusion algorithm based on minimax probability machine. Firstly,the feature extraction and matching of the face and the voiceprint are performed respectively,and then we fuse the matching score of the two uni-modal biometric on matching layer. The results on the Vidtimit multimodal database show that the equal error rate of the fusion system is reduced to 0. 97%,Which proves the effectiveness of this fusion algorithm.
作者
张瑛杰
彭亚雄
ZHANG Yingjie;PENG Yaxiong(School of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China)
出处
《电子科技》
2018年第5期40-43,共4页
Electronic Science and Technology
基金
国家科技支撑计划(2015BAK28B00)
关键词
生物特征认证
最小最大概率机
融合
匹配层
等错误率
biometric authentication
minimax probability machine
fusion
matching level
equal enor rate
