Handwritten Chinese/Japanese Text Recognition Using Semi-Markov Conditional Random Fields

Handwritten Chinese/Japanese Text Recognition Using Semi-Markov Conditional Random Fields
复制标题

使用半马尔可夫条件随机场的手写中文/日文文本识别

DOI:
10.1109/tpami.2013.49
复制
发表时间:
2013-10
影响因子:
23.6
通讯作者:
Nakagawa Masaki
Nakagawa Masaki
中科院分区:
计算机科学1区
文献类型:
--
作者:
Zhou Xiang-Dong;Wang Da-Han;Tian Feng;Liu Cheng-Lin;Nakagawa Masaki

文献摘要

参考文献

被引文献

相似文献

提出了一种基于半马尔可夫条件随机场(semi-Markov conditional random fields,semi-CRFs)的手写体中文/日文文本(字符串)识别方法。高阶半CRF模型定义在一个包含字符串所有可能的分割识别假设的格上,通过在特征函数中表示候选字符识别的分数以及几何和语言上下文的兼容性,优雅地融合候选字符识别的分数。基于给定的字符识别模型和兼容性模型,通过在训练字符串样本集上引入一个裕度项,最小化负对数似然损失,优化融合参数。提出了一种前后向格点剪枝算法以减少使用三元语言模型时的训练计算量,并研究了波束搜索技术以加快解码速度。我们评估所提出的方法的性能无约束的在线手写文本行的三个数据库。在CASIA-OLHWDB(中文)和TUAT Kondate(日文)数据库上的测试集上,字符级正确率分别为95.20%和95.44%,准确率分别为94.54%和94.55%。在ICDAR 2011中文手写体识别竞赛的测试集(在线手写文本)上,该方法的性能优于竞赛中的最佳系统。
This paper proposes a method for handwritten Chinese/Japanese text (character string) recognition based on semi-Markov conditional random fields (semi-CRFs). The high-order semi-CRF model is defined on a lattice containing all possible segmentation-recognition hypotheses of a string to elegantly fuse the scores of candidate character recognition and the compatibilities of geometric and linguistic contexts by representing them in the feature functions. Based on given models of character recognition and compatibilities, the fusion parameters are optimized by minimizing the negative log-likelihood loss with a margin term on a training string sample set. A forward-backward lattice pruning algorithm is proposed to reduce the computation in training when trigram language models are used, and beam search techniques are investigated to accelerate the decoding speed. We evaluate the performance of the proposed method on unconstrained online handwritten text lines of three databases. On the test sets of databases CASIA-OLHWDB (Chinese) and TUAT Kondate (Japanese), the character level correct rates are 95.20 and 95.44 percent, and the accurate rates are 94.54 and 94.55 percent, respectively. On the test set (online handwritten texts) of ICDAR 2011 Chinese handwriting recognition competition, the proposed method outperforms the best system in competition.
DOI: 10.1016/s0031-3203(02)00041-9
发表时间: 2003
期刊: Pattern Recognit.
影响因子: --
作者:
Shuyan Zhao;Z. Chi;P. Shi;Hong Yan
通讯作者: Shuyan Zhao;Z. Chi;P. Shi;Hong Yan
DOI: 10.1109/dial.2006.19
发表时间: 2006-04
期刊: Second International Conference on Document Image Analysis for Libraries (DIAL'06)
影响因子: --
作者:
Shaolei Feng;R. Manmatha;A. McCallum
通讯作者: Shaolei Feng;R. Manmatha;A. McCallum
DOI: 10.1109/icdar.2001.953776
发表时间: 2001-09
期刊: Proceedings of Sixth International Conference on Document Analysis and Recognition
影响因子: --
作者:
S. Tulyakov;V. Govindaraju
通讯作者: S. Tulyakov;V. Govindaraju
DOI: 10.1109/icdar.2005.209
发表时间: 2005-08
期刊: Eighth International Conference on Document Analysis and Recognition (ICDAR'05)
影响因子: --
作者:
Xianghui Wei;Shaoping Ma;Yijiang Jin
通讯作者: Xianghui Wei;Shaoping Ma;Yijiang Jin
DOI: --
发表时间: 2006-10
期刊: --
影响因子: --
作者:
T. Do;T. Artières
通讯作者: T. Do;T. Artières