Online Handwritten Mongolian Word Recognition Using MWRCNN and Position Maps

2016 15th International Conference on Frontiers in Handwriting Recognition (ICFHR) Pub Date : 2016-10-01 DOI:10.1109/ICFHR.2016.0024

Ji Liu, Long-Long Ma, Jian Wu

{"title":"Online Handwritten Mongolian Word Recognition Using MWRCNN and Position Maps","authors":"Ji Liu, Long-Long Ma, Jian Wu","doi":"10.1109/ICFHR.2016.0024","DOIUrl":null,"url":null,"abstract":"Considering the characteristic of Mongolian words where all letters of one Mongolian word are conglutinated together, the segmentation-free strategy is more suitable for Mongolian word recognition. This paper presents a novel recognition method based on MWRCNN and position maps for online handwritten Mongolian word. Firstly, the incorporation of position maps and aspect ratio is used to construct data transformation layer and enrich the Mongolian word shape information. Secondly, two feature combination methods based on MWRCNN are proposed to improve the recognition accuracy. Thirdly, by adopting multiple classification combination strategy, the accuracy of OHMWR can be further improved. We evaluated the recognition performance on online handwritten Mongolian word database with 946 classes. Experimental results show the proposed methods achieved the word-level recognition rate of 92.22% with data transformation, 92.60% with multiple feature combination and 93.24% with multiple classifier combination, respectively, which are better than the benchmarking test result 91.20% reported in the literature.","PeriodicalId":194844,"journal":{"name":"2016 15th International Conference on Frontiers in Handwriting Recognition (ICFHR)","volume":"62 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"6","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2016 15th International Conference on Frontiers in Handwriting Recognition (ICFHR)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICFHR.2016.0024","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 6

Abstract

Considering the characteristic of Mongolian words where all letters of one Mongolian word are conglutinated together, the segmentation-free strategy is more suitable for Mongolian word recognition. This paper presents a novel recognition method based on MWRCNN and position maps for online handwritten Mongolian word. Firstly, the incorporation of position maps and aspect ratio is used to construct data transformation layer and enrich the Mongolian word shape information. Secondly, two feature combination methods based on MWRCNN are proposed to improve the recognition accuracy. Thirdly, by adopting multiple classification combination strategy, the accuracy of OHMWR can be further improved. We evaluated the recognition performance on online handwritten Mongolian word database with 946 classes. Experimental results show the proposed methods achieved the word-level recognition rate of 92.22% with data transformation, 92.60% with multiple feature combination and 93.24% with multiple classifier combination, respectively, which are better than the benchmarking test result 91.20% reported in the literature.

查看原文本刊更多论文

基于MWRCNN和位置图的在线手写蒙古语词识别

考虑到蒙古语单词的所有字母都是粘合在一起的特点，无分词策略更适合蒙古语单词识别。提出了一种基于MWRCNN和位置图的在线手写蒙古语词识别方法。首先，结合位置图和纵横比构建数据转换层，丰富蒙古语词形信息;其次，提出了两种基于MWRCNN的特征组合方法，提高了识别精度。再次，采用多种分类组合策略，进一步提高了OHMWR的准确率。在946个类的在线手写体蒙古语词库上对识别性能进行了评价。实验结果表明，本文提出的方法在数据变换、多特征组合和多分类器组合下的词级识别率分别为92.22%、92.60%和93.24%，均优于文献报道的基准测试结果91.20%。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

2016 15th International Conference on Frontiers in Handwriting Recognition (ICFHR)

自引率

0.00%

发文量