| 基于改进微结构特征的笔迹鉴别 |
| |
| 引用本文： | 李昕,丁晓青.基于改进微结构特征的笔迹鉴别[J].清华大学学报(自然科学版),2010(4). |
| |
| 作者姓名： | 李昕 丁晓青 |
| |
| 作者单位： | 清华大学电子工程系; |
| |
| 基金项目： | 国家自然科学基金资助项目(60772049,60872086); 国家“九七三”重点基础研究项目(2007CB311004) |
| |
| 摘 要： | 该文作者已提出的基于网格微结构特征的笔迹鉴别方法,是一种能适用于多文种笔迹的文本无关方法。为了对笔画轨迹进行更加细致的描述,该文对微结构特征提取方法予以改进,修改了局部微结构的生成条件,并引入了加权Manhattan距离的相似度度量方法。改进方法在中文笔迹库上有效提高了鉴别正确率。在包含240人的HIT-MW笔迹库上,首选鉴别正确率为95.4%,前20选正确率达到100%。通过实验,进一步研究了实际应用中笔迹样本的文本内容相关度对鉴别性能的影响程度。
|
| 关 键 词： | 模式识别 文本无关笔迹鉴别 改进微结构特征 文本内容相关度 |
Writer identification based on improved microstructure features |
| |
| LIXin,DINGXiaoqing.Writer identification based on improved microstructure features[J].Journal of Tsinghua University(Science and Technology),2010(4). |
| |
| Authors: | LIXin DINGXiaoqing |
| |
| Institution: | LI+Xin,DING+Xiaoqing(Department+of+Electronic+Engineering,Tsinghua+University,Beijing+100084,China) |
| |
| Abstract: | The method previously proposed by the authors for writer identification based on the grid microstructure feature is a well-performance text-independent method for multilingual handwritings. Microstructure feature extraction was improved in this paper with the weighted Manhattan distance as the similarity measurement to describe finer writing characteristics. The method improves the identification performance on Chinese handwritings with the top-1 identification accuracy of 95.4% and the top-20 accuracy of 1... |
| |
| Keywords: | pattern recognition text-independent writer identification improved microstructure features text content dependency |
| 本文献已被 CNKI 等数据库收录！ |
