文章基本信息

标题：Offline Handwritten Numeral Recognition using Combinational Feature Extraction Approach
本地全文：下载
作者：Nishu Rastogi ; Maitreyee Dutta ; Ajay Indian 等
期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2017
卷号：5
期号：4
页码：7844
DOI：10.15680/IJIRCCE.2017.05040238
出版社：S&S Publications
摘要：Handwritten Character Recognition has become a very challenging area of research, the challenge isintroduced just because of the fact that the handwriting of a person differs very much from another person, even thehandwriting of a person varies depending upon the mood and environment. The conclusion drawn from existing workis that recognition accuracy depends on how and what features are utilized to formulate the feature vector from thesample set. A combinational feature vector contributes towards increased recognition rate. In this paper, we haveproposed and implemented combinational approach of feature extraction which includes direction gradient as well as 8– directional Quadrant Based Chain Code Histogram (QBCCH) to elicit structural components from the samples andthe results are calculated by formulating feature vector individually as well as with their combinations. For training andtesting purpose, samples of Gujarati numerals from 0 to 9 are considered as the pattern of them is a blend of both Hindiand English numerals. A feature vector of length nine using directional gradient histogram (DGH), eight using chaincoding (CCH) and thirty two using chain code by dividing image into four parts (QBCCH) is passed through the BackpropagationNeural Network with Levenberg-Marquartdt training function and recognition rate of approx. 72%, 82%and 98% respectively is attained. With combinational feature vector using DGH with CCH and DGH with QBCCHresulted in much higher recognition rates for isolated Gujarati Numerals.
关键词：Handwritten Character Recognition; Gradient; Freeman’s Chain Code; Back-propagation Neural;Network.