藏文字OCR特征分析和识别算法研究
首发时间:2016-12-28
摘要:为提高印刷体藏文OCR识别的精度和速度,提出将极限学习机(Extreme Learning Machine ELM)算法应用到藏文OCR(Optical Character Recognition)过程中,并对比了传统的单隐含层BP神经网络(Back Propagation Neural Networks)算法和支持向量机(Support Vector Machine SVM)算法。此外,特征提取阶段分别采用了三种不同的特征,分别是映射特征,网格特征以及像素特征。实验分析了识别率以及识别时间,其结果表明ELM算法取得了较高的识别率以及较短的识别时间。
For information in English, please click here
Tibetan OCR Feature Analysis and Recognition Algorithm
Abstract:To improve the accuracy and speed of painted Tibetan OCR(Optical Character Recognition) , the Extreme Learning Machine algorithm is applied to the process of Tibetan OCR, and compared with the traditional single-hidden layer BP neural network algorithm and support vector machine(SVM) algorithm. In addition, in the feature extraction phase, three different features are used for Tibetan OCR, namely mapping feature, grid feature and pixel feature. Experiments are conducted by considering the recognition rate and recognition time are analyzed. Our experimental results show that the ELM performs better than SVM and BP in terms of time and accuracy.
Keywords: Tibetan OCR Feature extraction Extreme Learning Machine
论文图表:
引用
No.4713155117394914****
同行评议
共计0人参与
勘误表
藏文字OCR特征分析和识别算法研究
评论
全部评论0/1000