摘要
在分析了语音识别的一维马尔可夫模型（１Ｄ－ＨＭＭ）方法的基础上，采用二维马尔可夫模型（２Ｄ－ＨＭＭ）方法识别工程图纸中的各类印刷体汉字、较规范的手写仿宋体汉字、英文及阿拉伯数字，适用于多字体、倾斜字符识别等情形，抗噪声能力强，取得了较高的识别率．
In this paper, after the analysis of the HMM approach in speech recognition, a new character recognition method for the engineering drawings based on 2D HMM is presented. This method can recognize the printed and handwriting Simulant Song Font Chinese character, English character and digit extracted from engineering drawings with high recognition accuracy. It is also suited for the case of multi type fonts, slant character recognition with high tolerance to distortion and blur.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
1999年第5期403-406,共4页
Journal of Computer-Aided Design & Computer Graphics
基金
"八六三"智能计算机主题资助
关键词
工程图纸
字符识别
维马尔可夫模型
工程设计
HMM
vector quantization
optical character recognition
feature selection
