藏语单字符手写识别与应用

发布时间：2018-03-06 23:11

本文选题：藏语单字符手写识别　切入点：特征提取　出处：《西安电子科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：我国幅员辽阔人口众多,是由56个民族组成的和谐大家庭,其中藏族是我国主要少数民族之一。藏语文献日积月累,除汉语之外,是我国历史最悠久、文献最丰富的语言文明遗产。目前,英语与汉语识别技术已经成熟,并且广泛地应用在各领域。相比于中英文,藏语的识别由于研究起步晚与研究人员需要熟悉藏语等原因,导致目前技术并不成熟,成果相对较少。藏语的人机交互方式还停留在键盘编码的方法上,输入方式唯一、速度慢与效率低,不能满足用户的需求。与传统键盘编码输入方式相比,手写输入与人类自然书写方式有更多相同。伴随着各种移动设备的普及,手写输入成为人机交互的一种重要方式,所以藏语手写识别不仅具有重要的社会意义,还有广阔的市场前景。本文采集并建立了1000套手写藏语字符数据库,详细介绍了藏文字符的特征,对手写藏文单字符识别进行了详细的研究,具体工作如下:1.详细介绍了藏语识别的背景、现状和研究意义。2.系统分类描述藏语字符的特征,同时对手写藏语特征具体的分析,并讲述手写藏语字符识别难点,接着介绍了目前研究领域常用的文字识别方法。3.详细说明了藏语手写字符预处理的步骤,最大程度保留手写藏语字符的原始信息,并且滤除冗余信息,便于特征提取与识别。4.介绍了藏语字符四个特征:方向线素特征、笔段结构特征、梯度特征和Gabor特征。其中方向线素特征与笔段结构特征是联机特征,梯度特征和Gabor特征是脱机特征。实验对比联机与脱机特征两者之间的性能,从中找出性能较好的特征。同时采用了两种融合方法,将两种识别方法进行融合,通过实验证明了系统的识别率较好。5.介绍Android平台的手写数据采集软件与手写按键结合式藏语输入软件。
[Abstract]:Our country has a vast population and is a harmonious family of 56 ethnic groups, among which the Tibetan nationality is one of the major ethnic groups in our country. The Tibetan language literature is accumulated day by day, besides Chinese, it is the longest history of our country. The most abundant cultural heritage of language. At present, English and Chinese recognition techniques are mature and widely used in various fields. Compared with Chinese and English, Tibetan language recognition is due to the late beginning of research and the need for researchers to be familiar with the Tibetan language. As a result, the current technology is not mature, and the achievements are relatively few. The human-computer interaction mode in Tibetan language still stays in the method of keyboard coding, the input mode is unique, the speed is slow and the efficiency is low. Compared with the traditional keyboard coding input, handwritten input is much the same as human nature. With the popularity of various mobile devices, handwritten input has become an important way of human-computer interaction. Therefore, the recognition of Tibetan handwriting not only has important social significance, but also has a broad market prospect. This paper collects and establishes 1000 sets of handwritten Tibetan characters database, and introduces the characteristics of Tibetan characters in detail. This paper makes a detailed study on handwritten Tibetan single character recognition. The specific work is as follows: 1. The background, current situation and research significance of Tibetan recognition are introduced in detail. 2. The characteristics of Tibetan characters are systematically classified and described. At the same time, the specific analysis of handwritten Tibetan characters is made. It also describes the difficulties of handwritten Tibetan character recognition, and then introduces the commonly used method of character recognition in the field of current research .3.The steps of Tibetan handwritten character preprocessing are described in detail, and the original information of handwritten Tibetan character is preserved to the maximum extent. And the redundant information is filtered out to facilitate feature extraction and recognition. 4. Four features of Tibetan characters are introduced: directional line element feature, pen segment structure feature, gradient feature and Gabor feature. Among them, directional line element feature and pen segment structure feature are on-line features. Gradient feature and Gabor feature are offline features. The experiment compares the performance between online feature and offline feature, and finds out the better feature. At the same time, two fusion methods are used to fuse the two recognition methods. The experiments show that the recognition rate of the system is good. 5. The handwritten data acquisition software based on Android platform and the Tibetan input software combined with handwritten keys are introduced.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：H214;TP391.43

【参考文献】