基于退化HMM的印刷体文字识别研究

Study on printed Chinese character recognition based on DHMM

提出一种新的基于退化隐马尔柯夫模型的印刷体文字识别方法.此方法按照一定规则提取文字的一维笔段序列特征,然后将该特征输入到设计好的分类器中进行分类.在分类器的设计上摒弃传统的左右型结构模型,采用了遍历型结构模型.实验证明此方法能够更好地完成文字分类任务,识别率可以达到99%以上.

A novel printed character recognition method based on degraded Hidden Markov Model（DHMM） is proposed, which can be applied to engineering. We extract 1D stroke sequence as the feature of the character under some rules and then put it into the classifier we designed. For the design of classifier, we adopt the ergodic model instead of the traditional left-right model. Experimental results showed the recognition rate of classifier with ergodic model is 99 %.