模糊保险箱算法的模板校准参数优化研究

Research on Parameters Optimization for Fuzzy Vault Algorithm of Template Alignment

在模糊保险箱(fuzzy vault)算法的具体实现中,几何哈希法是一种用于生物特征模板自动校准的常见技术。针对算法实现时的参数取值模糊问题,研究了影响Fuzzy Vault模板匹配精度的3个参数:图片像素大小、哈希表基点数和哈希表量化参数(α和β)。通过设计单因素实验方法,得到了这3个参数的最优取值范围,并改进了Fuzzy Vault算法细节点的提取范围和基点距离的选取规则,最后基于FVC指纹数据库对算法优化前后的匹配精度进行对比实验。结果表明,优化后算法的拒真率(false rejection rate,FRR)至少降低了9.84%,认假率(false acceptance rate,FAR)至少降低了7.12%,说明该优化方案提高了算法的匹配精度,具有一定的鲁棒性和实用性。

In the concrete implementation of fuzzy vault algorithm,geometric Hashing is a kind of common technology for the biometric template automatic alignment.To solve the fuzzy problem of parameters selection in the algorithm implementation,this paper studies three parameters which affect the matching accuracy:image pixels size,the number of basic points and quantitative parameters of a Hash table(αandβ).The optimal ranges of three parameters are obtained by carrying out the factor experiment analysis.Then,the extracting range of the minutiae algorithm and the rule of selecting the distance of basis points can be further optimized.Finally,the matching accuracy before and after optimization is compared and validated by the fingerprint picture based on the FVC databases.The experimental results show that the proposed optimization scheme can improve the matching accuracy of the algorithm because the FRR(false rejection rate)is reduced by9.84%,at least FAR(false acceptance rate)is reduced by7.12%,and has a certain robustness and practicality.