基于DCT和SVDD的在线手写签名认证方法①-计算机系统应用.PDFVIP

计 算 机 系 统 应 用 2016 年 第 25 卷 第 7 期 ① 基于DCT 和 SVDD 的在线手写签名认证方法 李成华, 刘 磊, 龚良慧, 郭 珩 ( 中南民族大学 电子信息工程学院, 武汉 430074) 摘 要: 为使在线手写签名认证的使用更具实用性, 提出了一种基于离散余弦变换(DCT)频域分析和支持向量数 据描述(SVDD)的在线手写签名认证方法. 依托自制的 Android 手机软件采集签名数据, 采用了基于 DCT 频域特 征分析和奇异值分解(SVD)的特征融合方法提取签名特征, 根据 SVDD 分类器适用于有限样本、一类分类方法建 模的优势, 建立了基于 SVDD 的认证模型进行在线签名认证, 并采用了网格搜索法对核函数参数进行优化选择. 实验结果表明, 该方法算法复杂度低, 快速有效, 提取的签名特征区分明显, 使用少量的一类真实签名作为训练 样本, 取得了较好的认证识别效果. 关键词: 在线手写签名; 离散余弦变换; 支持向量数据描述; 奇异值分解 Online Handwritten Signature Verification Method Based on DCT and SVDD LI Cheng-Hua, LIU Lei, GONG Liang-Hui, GUO Heng (College of Electronics and Information Engineering, South-central University for Nationalities, Wuhan 430074, China) Abstract: In order to make online handwritten signature verification more practical, this paper presents a method based on Discrete Cosine Transform (DCT) frequency analysis and Support Vector Data Description (SVDD). Signature data is collected by Android mobile phone software, and then extracts signature features based on DCT frequency analysis and Singular Value Decomposition (SVD). The online signature verification is established based on SVDD, for which is suitable for the small amount sample and one-class classification method, and uses the grid search method to optimize the kernel function parameters. The experimental results show that this method is efficient, has a low complexity, can extract the signature features clearly, and obtains a better recognition effect with a small amount of a real signature as the training sample. Key words: online handwritten signature; discrete Cosine transform; support vector data description; singular value decomposition 随着移动互联网和移动设备的广泛普及,