摘要
文章提出了一种基于多重隐马尔可夫模型和区域投影变换的手写体汉字识别新方法。该方法对每个汉字建立4个HMM,通过等比重综合方法将4个分类器的计算结果进行综合,从而得到识别结果。该方法对于识别应用问题中的隐马尔可夫建模具有普适性。实验证明,该识别方法具有较高的识别准确率。
A new approach for handwritten Chinese character recognition based on multiple hidden Markov model classifiers and sub-region projection transform is proposed in the thesis. The object oriented program framework is constructed by 4 HMMs for each Chinese character, and synthesizes the results of the 4 HMMs by equal proportion method, then gets the final recognition result. Experimental results show the efficiency of the proposed method.
出处
《信息安全与通信保密》
2007年第2期75-77,共3页
Information Security and Communications Privacy
关键词
汉字识别
特征提取
多重隐马尔可夫模型
边界链编码
Chinese character recognition
Feature extraction
Multiple Hidden Markov Model
Contour chain code
