首页> 外文期刊>情報処理学会論文誌 >On-line Chinese Handwriting Character Recognition: Comparison with Japanese Kanji Recognition and Improvement of Input Efficiency
【24h】

On-line Chinese Handwriting Character Recognition: Comparison with Japanese Kanji Recognition and Improvement of Input Efficiency

机译:在线中文手写字符识别:与日语汉字识别的比较和输入效率的提高

获取原文
获取原文并翻译 | 示例
           

摘要

Most Chinese characters are thought to have the same shape as their Japanese Kanji counterparts. Although this is true for most printed characters, it is not true for handwriting characters. This paper first describes the differences between handwriting character patterns in Chinese and Japanese based on an investigation of 8,063 categories of Chinese character patterns collected in Shanghai. 3,942 categories of the basic set were written by 400 people and 4,121 categories of the optional set by 50 people. Second, it is shown that Chinese characters are difficult to be recognized by adapting our original on-line Kanji recognition algorithm, because of the variety in number of strokes, stroke order and shape. Only 30/100 of the characters were written in the correct number of strokes (compared with 60/100 for Kanji), and only 80/100 of the correct samples were written in the correct stroke order. The original recognition algorithm was modified and improved to handle cursive style handwriting. The experimental results show a recognition rate of 89/100 and a 98/100 recognition rate is achieved for the 10th candidate. For practical Chinese character input, much higher recognition rates seem necessary and it is shown that a Chinese word dictionary can be used effectively together with the character recognition algorithm.
机译:人们认为大多数汉字的形状与日文汉字相同。尽管对于大多数印刷字符而言都是如此,但对于手写字符而言并非如此。本文首先对上海收集的8063种汉字样式进行了调查,描述了中文和日文手写字符样式之间的差异。基本集的3,942个类别由400人编写,可选集的4,121个类别由50人编写。其次,表明由于笔画数量,笔画顺序和形状的多样性,采用我们的原始在线汉字识别算法难以识别汉字。只有30/100个字符以正确的笔划数书写(汉字为60/100个字符),并且只有80/100个正确的样本以正确的笔划顺序书写。对原始识别算法进行了修改和改进,以处理草书风格的手写体。实验结果表明,第10位候选者的识别率是89/100,而识别率是98/100。对于实际的汉字输入,似乎需要更高的识别率,并且表明汉字词典可以与字符识别算法一起有效地使用。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号