Journal of South China University of Technology (Natural Science Edition) ›› 2014, Vol. 42 ›› Issue (1): 72-76,83.doi: 10.3969/j.issn.1000-565X.2014.01.013

• Electronics, Communication & Automation Technology • Previous Articles     Next Articles

Recognition of Similar Handwritten Chinese Characters Based on CNN and Ｒandom Elastic Deformation

Gao Xue Wang You- wang   

  1. School of Electronic and Information Engineering,South China University of Technology,Guangzhou 510640,Guangdong,China
  • Received:2013-06-15 Revised:2013-09-29 Online:2014-01-25 Published:2013-12-01
  • Contact: 高学(1967-)，男，博士，副教授，主要从事图像处理、模式识别与智能信息处理、手写汉字识别研究． E-mail:xuegao@scut.edu.cn
  • About author:高学(1967-)，男，博士，副教授，主要从事图像处理、模式识别与智能信息处理、手写汉字识别研究．
  • Supported by:

    国家自然科学基金资助项目(61271314);国家科技支撑计划项目(2013BAH65F01 －2013BAH65F04)

PDF

1953

Abstract:

In order to recognize similar handwritten Chinese characters effectively,a convolutional neural network(CNN) model is proposed,and the topology of the network model is presented.Then,the sample set is extendedby introducing a stochastic elastic deformation to enhance the generalization performance of the model.Experimen-tal results indicate that the recognition accuracy of the proposed CNN model is 1.66% higher than that of the tradi-tional CNN model,especially,for distorted handwritten Chinese characters,the recognition accuracy increases by12.85%; moreover,as compared with the traditional recognition methods,the proposed CNN model reduces therecognition error rate by 36.47%.It is thus concluded that the proposed method is effective.

Key words: character recognition, deep learning, convolutional neural network, elastic deformation

CLC Number: