基于HMM的脱机手写字符识别方法研究  

Method of off-line handwritten character recognition based on HMM

作　　者：葛照君[1] 葛世龙[2] 盛磊[1] 

机构地区：[1]杭州师范大学钱江学院,浙江杭州310012 [2]南京航空航天大学经济与管理学院,江苏南京210016

出　　处：《江苏科技大学学报（自然科学版）》2008年第6期57-61,共5页Journal of Jiangsu University of Science and Technology:Natural Science Edition

摘　　要：为了对手写字符进行识别,对隐马尔可夫模型算法及训练方法进行了探讨.首先简要地描述了字符识别的预处理和字符分割,然后利用Haar基函数提取字符特征,给出了解决隐马尔可夫模型的三个关键问题的算法,尤其是利用高斯混合模型对样本进行训练,建立了结合高斯混合模型的隐马尔可夫模型.实验结果表明,该方法有效可行,与已有研究结果进行对比,该方法具有一定的优越性.Training and computing methods of the HMM （ Hidden Markov Model） are studied in order to recognize handwritten character. The new recognition method including a detailed description of the preprocessing and segmentation is introduced; and then the feature extraction is described by using Haar basis function. Algorithm for solving three key problems of HMM is proposed, and HMM combined with Gauss mixture model is founded. The simulation result shows that the new recognition method is useful. Also it is superior to others.

关 键 词：手写字符识别 Haar基函数 高斯混合模型 期望最大化算法 隐马尔可夫模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]