摘要
针对常见指纹特征多次提取的易变性和识别系统原始指纹特征的安全性等问题,文中基于Frobenius范数进行了安全指纹识别系统的研究与设计。首先获得了指纹图像的感兴趣区域,在感兴趣区域的多层小波分解图像中提取其Fro-benius范数特征,并采用随机阵将范数特征映射到新的特征空间。Frobenius范数特征能够反映ROI的主要细节信息并具有较好的稳定性,随机映射可以有效增加合法用户与非法用户的可区分性。与原始指纹特征作为模板存于系统数据库相比较,存储随机映射结果可以确保原始指纹特征的安全性。FVC2002数据库进行的实验表明,基于Frobenius范数的指纹识别系统从性能和安全角度分析是可行有效的,而且获得了满意的等错率。
Aiming at the variance of fingerprint feature and the security of original feature, research and design on security fingerprint rec- ognition system based on Frobenius norm is proposed in this paper. Firstly,region of interest （ROI） can be obtained in fingerprint,and Frobenius norm can be got in the multiple wavelet decomposition of ROI. The norm can be mapped into another feature space by random matrix. Frobenius norm with a better stability can show the main and detail information, which can improve the difference between legal users and illegal ones. Compared with the original fingerprint feature stored in the system, random map can assure the security of the origi- nal fingerprint feature. The experimental results show that the system is feasible and effective. A satisfactory equal error rate （EER） can be got in the proposed system.
出处
《计算机技术与发展》
2012年第10期181-183,187,共4页
Computer Technology and Development
基金
国家自然科学基金资助项目(60272022)
