QQ客服 官方微博 反馈留言

题名	手写中文速记符的自动识别
作者	陈小苹
学位类别	工学博士
答辩日期	1997-10-01
授予单位	中国科学院自动化研究所
授予地点	中国科学院自动化研究所
导师	戴汝为 ; 乔谊正
其他题名	Automatic Recognition of Handwritten Chinese Shorthand
学位专业	模式识别与智能系统
中文摘要	在当前信息时代里，高速信息交换作为计算机的主要应用方面越来越受到人们的重视， 然而，计算机本身信息处理的高速性与信息输入(尤其是中文信息输入)缓慢的矛盾大大限 制了计算机资源的充分利用。为了充分发挥计算机的高速处理性能，如何有效地解决信息输 入缓慢的问题已成为信息科学的重要研究方向。 手写中文速记符作为中文信息的计算机输入手段在自然性、简单性、快速性等方面具有 其它输入方案无可比拟的优越性，也是一种很有潜力的高速中文输入界面。 本文以手写中文速记体系中的“人群速记”作为研究对象，对手写人群速记符自动识别 系统ATSRS(Automatic Transcription System of Renqun Shorthand)的设计原理和研制过程进 行了详细描述。 本论文共分九章。第一章介绍了现有中文信息的计算机输入方法，在对各种计算机输入 方法优缺点分析的基础上，提出了手写中文速记符识别输入方案，并阐述了手写中文速记符 的一般识别方法。 第二章介绍了“人群速记”体系的一些基本知识，在对国内外手写速记符自动识别技术 总结的基础上，给出了联机手写人群速记符的两种识别方案。 第三章主要讲述了研制速记符自动识别系统所采用的输入装置和数据采集过程。 第四章在借鉴已有字符预处理技术基础上，对速记符采样数据的预处理进行了详细描 述，其中包括剔除飞点、空间采样、剔除虚假抬笔点、滤波、特征点保持、位置和点数的规 格化、局部滤波、矢量化等八个方面。在以上几种预处理中，前五个方面侧重于特征抽取， 后三个方面侧重于识别。 第五章重点讨论了速记符的基元切分算法。本章从连笔速记符的结构特征及形状特征入 手，在GT启发式搜索算法的基础上，结合问题分层规划思想提出了一种适于切分问题的算 法，我们称之为GTAS算法(Generate and Test Algorithm for Segmentation)。 第六章介绍了速记符的基元选择及特征抽取过程，在此基础上，利用模糊分类原理以及 多级分类策略对各基元进行了分类。 第七章给出了速记符的具体识别方法，其中有基于逻辑推理的人工智能法、基于类似度 的形状匹配法、基于方向码序列的动态规划模板匹配法以及基于速记符形状特征和结构特征 的三级识别方法。 第八章描述了对基本音符、略符和连笔符的识别测试情况，对基本音符和略符，分别采 用了逻辑推理法、类似度形状匹配法、动态规划模板匹配法，对基本音符，还采用了三级识 别方法进行了测试。对连
英文摘要	As a main application of computer, rapid information exchange has been receiving more and more attention in current information age. However, the contradiction between rapid processing and slow entry of information into computer (especially for Chinese information entry) greatly restricts the full use of computer resources. In order to give full play to the rapid information processing of computer, how to solve the slow information entry problem efficiently has become an important research area of information science. Handwritten Chinese shorthand has the advantage of other input means at naturalness, simplicity and rapidity as a_ means of Chinese information entry into computer. Therefore it is very potential in becoming a high speed Chinese information entry medium. In this thesis, Renqun shorthand -- a handwritten Chinese shorthand system is taken as the research object, and the design principle and development process for the ATSRS (Automatic Transcription System of Renqun shorthand) are described in detail. This thesis can be divided into 9 chapters. Chapter 1 gives a brief review of alternative inputs for Chinese information input to computers. Based on the analysis of several main input means, handwritten Chinese shorthand recognition scheme is presented and general recognition approach for shorthand is also given. Chapter 2 describes the basics of Renqun shorthand. After an overall review of handwritten shorthand recognition technologies, two strategies for the on-line automatic recognition of Renqun shorthand are put forward. Chapter 3 mainly describes the input device and data collection process for the shorthand recognition system. Chapter 4 focuses on the preprocessing processes of shorthand sampling data, which include inflection points removal, space sampling, false pen-lift removal, filtering, characteristic points keeping, position and number normalization, partial fikering and vectorization. Among these preprocessing processes, the former five aspects are aimed at feature extraction, and the latter three are directed at recognition. Chapter 5 describes the primitive segmentation of Renqun shorthand. A generate and test algorithm for segmentation (GTAS) is presented based on the structral and shape characteristics of linking shorthands and also the theory of Generate and Test heuristic search algorithm and hierarchical planning. Chapter 6 introduces the primitive selection and feature extraction process for Renqun shorthand. The primitives selected and extracted are classified by using fuzzy classification theory and multistage classification strategy. Chapter 7 gives the concrete recognition approaches, which include artificial intelligence method based on logical reasoning, shape matching method based on the similarity between patterns, dynamic programming template matching technique based on the directional codes and three-stage recognition method based
语种	中文
其他标识符	409
内容类型	学位论文
源URL	[http://ir.ia.ac.cn/handle/173211/5677]
专题	毕业生_博士学位论文
推荐引用方式 GB/T 7714	陈小苹. 手写中文速记符的自动识别[D]. 中国科学院自动化研究所. 中国科学院自动化研究所. 1997.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

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。