基于跳变点能量的DTW匹配在线手写签名验证算法  

On-line handwriting signature verification algorithm based on DTW energy feature matching of sharp trajectory change point

作　　者：徐陶祎[1] 郑建彬[2] 

机构地区：[1]武汉理工大学信息工程学院,武汉430070 [2]华中科技大学电子与信息工程系,武汉430074

出　　处：《计算机应用研究》2007年第8期84-86,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(69672014)

摘　　要：提出一种以签名能量为特征的在线手写签名验证算法,侧重签名能量特征提取和匹配判决的研究。对签名波形进行小波分解,提取签名波形在跳变点处的能量,从中提取若干个能量值作为特征矢量;在基于动态时间规整的特征匹配用改进的动态时间规整方法将测试特征序列和模式特征序列进行匹配的基础上,计算最小匹配距离,得出匹配路径。实验表明,动态时间规整算法在签名验证识别中获得了良好性能,对于随机伪造签名,误拒率为0时,误纳率为6.86%。The paper proposed an on-line handwriting signature verification algorithm as feature based on signature energy, emphasized the study of signature energy features extraction method and matching and decision-making process. Extracted the energy of sharp trajectory change point in the signature wave, by means of wavelet decomposition of signature. Then, chosen some points with most dominant energies as characteristic value. Matching the pattern character and test character with dynam- ic time warping enables the researchers to compare the distortions between them so as to obtain the criterion for identification, calculated the minimum marching distanee, got matching route. According to the experiment, the algorithm of DTW performed well in the handwriting signature verification, Resu/ts show that false rejection rate is 0 percent while false acceptance rate is 6. 86 percent for random forgery .

关 键 词：签名验证 能量特征 动态时间规整 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]