| 基于HMM的英文印刷体识别建模研究 |
| |
| 引用本文： | 冀爽,哈力木拉提·买买提.基于HMM的英文印刷体识别建模研究[J].现代计算机,2011(8):14-16. |
| |
| 作者姓名： | 冀爽 哈力木拉提·买买提 |
| |
| 作者单位： | 新疆大学信息科学与工程学院 |
| |
| 摘 要： | 讨论一种基于HMM(隐马尔可夫模型)的英文印刷体识别方法。先将整篇文本图像切分成字母级别,提取出字母轮廓的8方向特征,之后把特征向量进行矢量量化并送入HMM训练识别。根据切分中出现的错误特点,对矢量量化过程和训练算法提出一些改进方法,提高识别率。
|
| 关 键 词： | 特征提取 矢量量化 隐马尔可夫模型 |
Research on Modeling of Printing English Words Recognition Based on HMM |
| |
| Affiliation: | JI Shuang,Halmurat Mamat (College of Information Science and Engineering,Xinjiang University,Urumqi 830046) |
| |
| Abstract: | Discusses a printing English words recognize method based on Hidden Markov Model.The entire text image is segmented into letters,and extracts each letter’s 8-direction features,then adopts a vector quantization method and sends the feature vectors to HMM for training or recognition.According to segmentation,proposes some improvements in VQ method and training algorithm,so as to improve the recognition rate consequently. |
| |
| Keywords: | Feature Extraction Vector Quantization Hidden Markov Model |
| 本文献已被 CNKI 等数据库收录！ |
|
