基于Android的联机手写维吾尔文识别研究

发布时间：2018-07-24 09:02

【摘要】：维吾尔文是新疆维吾尔自治区少数名族的主要交流文字,为了便利当地人们交流,开展有关维吾尔文文字处理技术的研究是很有必要。在当今社会,常用电子信息设备都已经进入了平常百姓家,特别是手机移动终端设备,已经成为了人们日常生活之中不可缺少的通信工具,开展有关手机移动终端上的维吾尔文文字信息处理技术已显得迫切需要。目前,针对维吾尔文字的识别技术主要有印刷体识别和联机手写文字识别。在印刷体文字识别方面,已经取得了较大发展,但有关联机手写文字识别方面的研究还比较少并且大部分研究都是以字母为基本识别单位的。以字母为基本识别单位,其优点就是需要建立的字符模型少,识别效率快。但其缺点也较为明显,首先要面临的就是字母切分困难问题。因为维吾尔文属于黏着性语言文字,其大部分字母都可以粘连书写,具有天然的草书特点,并且字母变体多,如何切分出正确的字母一直都是研究难点。再就是针对基本字母的识别无法满足人们在触屏手机上进行单词输入的实际应用需求。为了解决此问题,本文主要针对基于Android移动终端上的手写维吾尔文字识别方法进行了研究。首先研究了在Android手机上针对维吾尔文单词进行采样的方法,其次,以采集的样本信息为基础,对维吾尔文单词的预处理和特征提取方法进行了研究。同时为了满足可以识别维吾尔文整词的需求,本文中根据维吾尔文单词都是由多个连体段组成的特点,选取以连体段为基本单位进行特征的提取,然后通过拼接连体段特征构成一个单词的特征向量。最后,通过本文中提出的特征压缩方法将已提取的单词特征向量转化为离散隐马尔可夫模型可以读取的观察值序列,接下来使用隐马尔科夫模型完成对单词的建模和识别测试。本文主要研究了在安卓移动设备上对手写维吾尔文单词进行识别的方法。通过在手机上完成维吾尔文单词采样后,前期研究工作主要是在电脑端完成的,后期再将识别程序移植到手机设备,初步在安卓移动终端上完成对手写维吾尔文单词的识别。
[Abstract]:Uygur language is the main communication character of a few ethnic groups in Xinjiang Uygur Autonomous region. In order to facilitate the exchange of local people, it is necessary to carry out research on Uygur character processing technology. In today's society, common electronic information devices have entered the homes of ordinary people, especially mobile terminal devices, and have become an indispensable communication tool in people's daily lives. It is urgent to develop Uighur character information processing technology on mobile terminal. At present, Uygur character recognition technology mainly includes print recognition and online handwriting recognition. In the field of printed character recognition, great progress has been made, but there are few researches on on-line handwritten character recognition and most of the researches are based on letters. Taking letters as the basic recognition unit, it has the advantages of less character models and faster recognition efficiency. But its shortcomings are obvious, the first is the problem of alphabetic segmentation. Because Uyghur belongs to adherent language and characters, most of its letters can be stuck and written, and has the characteristics of natural cursive script, and there are many variations of letters. How to cut out the correct letters has always been a research difficulty. Second, the recognition of basic letters can not meet the practical needs of word input on touch-screen mobile phones. In order to solve this problem, this paper mainly focuses on handwritten Uighur character recognition based on Android mobile terminal. Firstly, the method of sampling Uygur words on Android mobile phone is studied. Secondly, based on the sample information collected, the preprocessing and feature extraction methods of Uygur words are studied. At the same time, in order to meet the need to recognize Uygur integer words, according to the characteristics of Uygur words are composed of multiple conjoined segments, we select the conjoined segment as the basic unit for feature extraction. Then the feature vectors of a word are constructed by splicing the joint segment features. Finally, the extracted word feature vector is transformed into a sequence of observation values that can be read by discrete hidden Markov model through the proposed feature compression method in this paper. Then, the hidden Markov model is used to complete the word modeling and recognition test. This paper mainly studies the recognition method of handwritten Uygur words on Android mobile devices. After completing the Uygur word sampling on the mobile phone, the previous research work is mainly completed on the computer, and the recognition program is transplanted to the mobile phone device at the later stage, and the recognition of the handwritten Uygur words is preliminarily completed on the Android mobile terminal.
【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.43;TP316

【参考文献】