Analysis of Handwriting Variation and Development of High Performance Computer-Based System to Support Forensic Examination of Japanese Documents

该研究项目已进行了两年多（2003-2004年）。本研究的目的是建立笔迹的个性，并开发一个高性能的司法文书检验支持系统。我们能够达到这个研究项目的目的。主要研究成果如下：（1）在传统的笔迹提取方法的基础上，加入颜色信息，改进了一种笔迹提取方法。通过实验证明，该方法能够准确地从各种类型的文档中提取出笔迹和印鉴，即使笔迹和印鉴相互叠加也能准确地提取出来。(2)法医笔迹鉴定是基于这样的假设，即每个人都有一致的笔迹，与其他人的笔迹不同。通过许多案例研究，人们主观上接受了这一假设。然而，由于笔迹的个体性还没有得到科学的、定量的严格证明，其作为司法证据的可采性还存在疑问。提出了一种基于变异熵的笔迹个性识别方法，并进行了基于模式匹配的笔迹鉴别实验。通过实验，证实了变异熵作为笔迹整体形状变异的评价指标是合适的。最后给出了将变异熵与笔迹鉴别结果相结合进行笔迹个性鉴别的前景。(3)我们开发了一个手写体笔画结构的测量工具。我们目前正利用统计分析的方法来研究笔迹的笔画结构的个体性，希望将来能结合以上研究结果，发展出一套完整的系统来支援司法文件的检验。

This research project has been carried out over two years (2003-2004). The aims of this research project are to establish individuality of handwriting and to develop a high performance support system for forensic document examination. We were able to achieve the aim of this research project. The summary of results obtained by the research project are as follows :(1) We have improved a handwriting extraction method by adding color information to the conventional extraction method. Through an experiment, it confirmed that this extraction method can extract precisely handwriting and seal imprint from various types of documents, even if handwriting and seal imprint are superimposed each other.(2) Forensic handwriting examination is based on the hypothesis that each individual person has consistent handwriting that is distinct from the handwriting of other individuals. This hypothesis has accepted subjectively through many case studies. However the individuality of handwriting has not been established with scientific and quantitative rigor, therefore its admissibility as forensic evidence can be questioned. We proposed a method for validating handwriting individuality using a "variation entropy" and a writer verification experiment based on the pattern matching method. Through the experiments, it was confirmed that the variation entropy was appropriate for evaluating the variation of handwriting as whole shape. We got a prospect to be possible of validating handwriting individuality by combining the variation entropy and writer verification results.(3) We have developed a measurement tool of handwriting stroke structure. We are now investigating on the individuality of stroke structures of handwriting by using a statistical analysis method.We aim to develop a complete system to support forensic document examination by combining above results in the future.