基于联机笔迹的身份鉴别
文献类型:学位论文
作者 | 于昆 |
学位类别 | 工学硕士 |
答辩日期 | 2004-06-01 |
授予单位 | 中国科学院研究生院 |
授予地点 | 中国科学院自动化研究所 |
导师 | 王蕴红 ; 谭铁牛 |
关键词 | 笔迹鉴别 文字识别 文本独立 笔划 笔划段 抬笔段 网络 矩 Handwriting Authentication Handwriting Recognition Text-independent Stroke Stroke Segment Interstroke Lattice Moment |
其他题名 | HANDWRITING AUTHENTICATION BASED ON ON-LINE INFORMATION |
学位专业 | 模式识别与智能系统 |
中文摘要 | 依靠笔迹进行身份鉴别长期以来都是人们研究的焦点。迄今为止,针对于离线 笔迹鉴别的研究已经比较广泛,但是在对在线笔迹鉴别的研究中,仍有许多关键性问 题有待解决,尤其是针对内容独立的笔迹鉴别问题的深入探索。本文中总结了传统的 文本独立和文本相关的笔迹鉴别方法和技术,并在此基础上针对文本独立的联机笔迹 鉴别问题从笔迹预处理、特征提取以及特征匹配等方面进行了研究,提出了基于笔划 分割的书写人鉴别方法。本论文内容安排如下: 第一章介绍了笔迹分析相关背景,尤其是笔迹鉴别技术的背景知识以及有关概 念,并对于目前笔迹鉴别所面临的主要问题作了简要分析: 第二章介绍了联机笔迹数据的采集方法,在分析了传统的手写笔迹预处理方法 的基础上,提出了针对文本独立的联机笔迹所进行的基于笔划的逐次分割预处理方 法: 第三章总结了对于书写笔迹特征的全局和局部提取方法,定义了用于书写人识 别的汉字笔划段,并提出基于笔划段分割和高斯模型来对笔划动态特征建模的方法, 以此来提取笔迹动态信息: 第四章实现并比较了多种分类方法,提出了基于笔划的加权最小距离分类方法, 并根据测试时笔划段出现的频率给出了加权准则。实验证明,这种措施可以有效地提 高笔迹鉴别效果; 第五章简要介绍了联机笔迹鉴别方法在系统上实现的流程,并分析了系统的可 行性问题以及对于笔迹鉴别效果产生影响的主要因素; 第六章对于本论文的工作进行了总结,并对于笔迹鉴别技术的发展作出展望。 总的来说,对于文本独立的联机笔迹鉴别问题,本文提出了一种基于笔划段的 方法,并以此为基础对于书写机理等问题作了探索性的研究。 |
英文摘要 | Biometric authentication based on handwriting has been the focus of research all along. So far people have made some achievements on off-line handwriting analysis. There still exist many significant problems on on-line handwriting analysis unsolved, especially those pertinent to text-independent handwriting authentication. The traditional methods for handwriting analysis concerning off-line and on-line cases are summarized in the thesis, and based on the comparison of the existing methods, a stroke-based handwriting authentication method is proposed. The pertinent work includes research on handwriting collection and preprocessing, feature extraction and matching, and a system of handwriting authentication is realized. The thesis is organized as follows: Chapter 1 introduces some basic concepts concerning handwriting analysis, the popular methods and techniques, and background of application and difficulties. Chapter 2 presents techniques for handwriting sample collection and traditional handwriting preprocessing methods. A stroke-based stepwise handwriting segmentation method is proposed. Chapter 3 summarizes the traditional methods for global and local feature extraction of handwriting samples. A stroke segmentation method is employed after some important strokes are defined. The proposed feature extraction method is to use Gaussian models to represent the statistical characteristics of strokes. Chapter 4 compares different classifiers, The rule of weights distribution is proposed, and the experimental results indicate that the proposed method is valid to writer identification. Chapter 5 gives a brief introduction of the working procedure on the system, and the feasibility of the system is further discussed. Some factors concerning the effect of the method are also discussed. Chapter 6 summarizes the thesis and delivers the prospect of handwriting analysis. |
语种 | 中文 |
其他标识符 | 774 |
源URL | [http://ir.ia.ac.cn/handle/173211/6772] ![]() |
专题 | 毕业生_硕士学位论文 |
推荐引用方式 GB/T 7714 | 于昆. 基于联机笔迹的身份鉴别[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2004. |
入库方式: OAI收割
来源:自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。