基于ALCSL&LS 核函数的支持向量机签名认证

提出一种名为ALCSL&LS的核函数来描述签名动态信息之间的相似性. ALCSL&LS核函数利用自适应最长共同子序列的长度,不仅描述了签名动态信息整体上的趋势相似性,而且选择性的描述了某些重要的特征局部在时间弯曲校正后的相似性,并将两者融合生成核矩阵. 该核函数具有相似性描述直观、对噪音具有的鲁棒性和易于实现等优点,它核化的支持向量机实现了签名原空间上的非线性划分. 基于此方法的签名认证实验获得了比较满意的结果.

Signature Verification by Support Vector Machine with ALCSL&LS Kernel

It proposes a new kernel, which is named ALCSL&LS kernel, to measure the similarity between signing dynamic signals. ALCSL&LS kernel measures not only the tendency similarity of the signing dynamic signals through the Adaptive Longest Common Subsequence Length (ALCSL), but also the Local Similarity (LS) of the important parts of the signing dynamic signals after time warping rectification. It has some advantages, including robust to noise and distortion, full of intuition, and easy realization. ALCSL&LS kernel is introduced into the Support Vector Machine (SVM) to realize the nonlinear classification of signatures in the input space. The signature verification experiment using this method yields satisfactory results.