您的位置:首页  > 论文页面

藏文字OCR特征分析和识别算法研究

发表时间:2017-08-31  浏览量:39  下载量:4
全部作者: 李旺平,李琳,谢忠伟
作者单位: 武汉理工大学计算机科学与技术学院
摘 要: 为提高印刷体藏文识别(optical character recognition,OCR)的精度和速度,提出将极限学习机(extreme learning machine,ELM)算法应用到藏文OCR过程中,并与传统的单隐含层BP神经网络(back propagation neural network)算法和支持向量机(support vector machine,SVM)算法进行对比。此外,特征提取阶段分别采用3种不同的特征,分别为映射特征、网格特征及像素特征。实验分析识别率及识别时间,结果表明,与SVM和BP神经网络算法相比,ELM算法取得了较高的识别率及较短的识别时间。
关 键 词: 模式识别;藏文OCR;特征提取;极限学习机
Title: Tibetan OCR feature analysis and recognition algorithm study
Author: LI Wangping, LI Lin, XIE Zhongwei
Organization: School of Computer Science and Technology, Wuhan University of Technology
Abstract: To improve the accuracy and speed of painted Tibetan optical character recognition (OCR), the extreme learning machine (ELM) algorithm is applied to the process of Tibetan OCR, and it is compared with the traditional single-hidden layer BP neural network algorithm and support vector machine (SVM) algorithm. In addition, 3 different features are used for Tibetan OCR in the feature extraction phase, namely mapping feature, grid feature and pixel feature. Experiments are conducted by considering the recognition rate and time. Experimental results show that the ELM algorithm performs better than SVM and BP neural network algorithms in terms of recognition rate and time.
Key words: pattern recognition; Tibetan OCR; feature extraction; extreme learning machine
发表期数: 2017年8月第16期
引用格式: 李旺平,李琳,谢忠伟. 藏文字OCR特征分析和识别算法研究[J]. 中国科技论文在线精品论文,2017,10(16):1834-1841.
 
0 评论数 0
暂无评论
友情链接