基于DWT的二次特征提取脱机中文签名鉴定

Off-line Chinese signature verification based on the second feature extraction by DWT

提出了一种基于投影特征和离散小波变换(DWT)的脱机中文签名鉴定的新方案.该方案利用小波分解的高频系数具有较好时间分辨率的特点,对调整后的投影特征矢量选用各层高频系数进行单支重构提取二次特征,此二次特征在增强了签名者的个人特性的同时也增加了签名特征隶属于真签名的不确定性,由此引入模糊网络来表征这种不确定性进行鉴定.实验中对中文签名数据库取得了大约11%的平均错误率,较现有文献的实验结果减少了大约6%.

A novel approach to off-line Chinese signature verification is proposed based on projection profiles and discrete wavelet transform （DWT）. Since the high frequency coefficients of wavelet decomposition represent the better time-resolution features, the reconstructed signals are the second features of the adjusted projection profiles. These features can enhance the signature characteristics, while with an increase of uncertainty to the true signatures. Fuzzy nets are introduced to describe the variation and to employ verification. The average error rate is about 11% obtained from a Chinese signatures database, which is about 6 % less than that of the existing research.