基于字素分割的蒙古文手写识别研究

发布时间：2018-01-26 18:49

本文关键词： 蒙古文字素 HMM 手写识别　出处：《中文信息学报》2017年05期 　论文类型：期刊论文

【摘要】：隐马尔科夫模型(HMM)对序列数据有很强的建模能力,在语音和手写识别中都得到了广泛的应用。利用HMM研究蒙古文手写识别,首先需要解决的问题是手写文字的序列化。从蒙古文的构词和书写特点看,蒙古文由多个字素从上到下串联构成。选择字素集合和词的字素分割是手写识别的基础,也是影响识别效果的关键因素。该文根据蒙古文音节和编码知识确定了蒙古文字母集合,共包括1 171个字母。通过相关性处理、HMM排序筛选等手段得到长字素集合,共包括378个字素。对长字素经过人工分解,获得了50个短字素。最后利用两层映射给出了词转字素序列的算法。为了验证长短字素在手写识别中的效果,我们在HTK(hidden Markov model toolkit)环境下利用小规模字库实现了手写识别系统,实验结果表明短字素比长字素有更好的性能。文中给出的字素集合和词转字素序列的算法为后续基于HMM的蒙古文手写识别研究奠定了基础。
[Abstract]:Hmm (Hidden Markov Model) has a strong ability to model sequence data and has been widely used in speech recognition and handwritten recognition. HMM is used to study Mongolian handwritten recognition. The first problem to be solved is the serialization of handwritten characters. Mongolian characters are composed of several morphemes in series from top to bottom. The selection of morpheme sets and word morpheme segmentation is the basis of handwritten recognition. It is also the key factor that affects the recognition effect. According to the Mongolian syllable and coding knowledge, this paper determines the Mongolian alphabet set, including 1 171 letters. HMM sorting and screening methods were used to obtain the set of long word elements, which included 378characters. The long words were decomposed manually. In order to verify the effect of long and short morpheme in handwriting recognition, we obtain 50 short morphemes. Finally, we use two-layer mapping to give the algorithm of morpheme sequence. In the environment of HTK(hidden Markov model toolkit, we realized the handwriting recognition system by using a small font library. The experimental results show that the short morpheme has better performance than the long word primer. The algorithm of the morpheme set and the morpheme sequence provided in this paper lays a foundation for the further study of Mongolian handwriting recognition based on HMM.
【作者单位】：内蒙古大学计算机学院;内蒙古大学电子信息工程学院;
【基金】：内蒙古自治区自然科学基金(2016MS0603)
【分类号】：TP18;TP391.4
【正文快照】： 1引言各种语言的手写体识别是人工智能领域最具有挑战性的研究课题之一,主要包括脱机手写体识别和联机手写体识别。国外在文字识别领域较早地开展了此项工作,由于西文(如英文、法文)大多是拼音文字,字母数量少且书写规整,其印刷体和手写体识别产品早已问世。在国内许多大专院

【相似文献】