Institutional Repository
| 联机藏文手写识别特征提取方法研究 | |
| Alternative Title | The Study on Feature Extraction of Online Handwritten Tibetan Characters Recognition |
| 刘瀚猛 | |
| Major | 计算机软件与理论 |
| Supervisor | 张健 |
| 2008-06-03 | |
| Degree Grantor | 中国科学院研究生院 |
| Degree Level | 博士 |
| Place of Degree Grantor | 中国科学院软件研究所 |
| Keyword | 特征提取 网格特征 方向分解 Gaussian滤波器 |
| Classification | 暂无 |
| English Abstract | 近年来，藏文信息处理倍受国家重视，发展迅速，但是针对联机藏文手写识别的研究却处于起步阶段。半个世纪以来，关于英文、中文、日文等的联机手写识别的研究已经发展的相当成熟，大量应用于各种联机终端中，给大众带来了巨大的便利。根据联机藏文手写字符的特点，在大量的已有理论中选取适合的方案，对选中的成熟理论进行合理的改进，是一条快捷的发展联机藏文手写识别的路线。本文研究了联机藏文手写识别中的预处理和特征提取部分。预处理是保证特征提取设计方案有效性的前提，包括线性归一化、非线性归一化、等距离重采样和笔画平滑等操作。对这些操作，本文参考其他文种联机手写识别中成熟的方法，针对联机藏文手写字符的特点进行了筛选和改进。本文主要使用网格方向特征来对特征提取阶段进行设计，它是一种目前大量使用的手写体特征。网格方向特征提取方案主要包括方向分解、投影子分量图像、网格划分、确定权值分配器等操作。本文针对藏文手写字符的特点，在大量实验的基础上，确定了四方向拆分、均匀网格划分和使用Gaussian滤波器作为权值分配器的网格方向特征提取方案。在对藏文手写识别预处理和特征提取研究的基础上，本文提出了基于网格方向特征的联机藏文手写识别特征提取方案。 |
| Call Number | 暂无 |
| Abstract | Recently, the Chinese government has attached great importance to the development of Tibetan Information Processing, which has made an impressive achievement. However, for online handwritten Tibetan character recognition, the study is in its infancy. For many other languages, the study on online handwritten characters recognition have grown maturely by now, which has brought a huge convenience to the populace. According to the features of online handwritten Tibetan characters, make the reasonable improvement to the suitable theories selected from massive existing ones, is a shortcut for online handwritten Tibetan characters recognition. This paper designs the preprocessing and the feature-extraction for online Tibetan handwriting recognition. Preprocessing is the guarantee for the feature-extraction's validity, which consists of linear normalization, nonlinear shape normalization, equidistance resample and smoothing. According to the features of online handwritten Tibetan characters, all the components of the preprocessing are carefully screened and improved, referenced to the mature approaches applied in other kinds of online handwriting recognition. The meshing directional feature, which is considered efficient and practical widely, is used to design the feature-extraction in this article. The meshing directional feature extraction contains many steps, such as direction extraction, generating directional image, meshing division and weight allocator calculation. According to the features of online handwritten Tibetan characters and in the massive experiment's foundation, the meshing directional feature for online Tibetan handwriting is extracted with four-direction extracting, even meshing division and Gaussian filter. The Feature Extraction of Online Handwritten Tibetan Characters Recognition Based on Meshing Directional Feature is put forward. |
| Department | 基础软件国家工程研究中心 |
| Content Type | 学位论文 |
| URI | http://ir.iscas.ac.cn/handle/311060/7252 |
| Collection | 基础软件国家工程研究中心 |
| Recommended Citation GB/T 7714 | 刘瀚猛. 联机藏文手写识别特征提取方法研究[D]. 中国科学院软件研究所. 中国科学院研究生院,2008. |
| Files in This Item: | ||||||
| File Name/Size | DocType | Version | Access | License | ||
| 10001_20052801502905（2564KB） | 开放获取 | -- | Application Full Text | |||
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment