Font Size: a A A

Research On The Key Technologies Of Online Handwritten Uyghur Word Recognition

Posted on:2012-04-02Degree:MasterType:Thesis
Country:ChinaCandidate:N G L D G T ReFull Text:PDF
GTID:2178330335485859Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
In this paper, based on the Uighur writing rules and structure characteristics, research the technology of online handwriting Uyghur word recognition,puts forward a good solution for online handwriting Uyghur word recognition,and initially solved the two most important problems of online handwriting Uyghur word recognition,namely segmentation problem and characters recognition problem.On the word segmentation, this paper adopted the two levels of segmentation scheme,which first the word segmentation into conjoined section,then the conjoined section cut into characters.We put forward conjoined section segmentation algorithm and characters segmentation algorithm for conjoined section segmentation problem and characters segmentation problem.The accuracy of conjoined section segmentation algorithm and characters segmentation algorithm reached 79.54% and 58.68% for unlimited online handwritten samples.On the characters recognition, we put forward a characters recognition scheme which accord with the features of Uyghur characters.And based on this scheme,we made four characters recognition system for four different writing form of Uyghur characters,namely independent form characters recognition system,even form before characters recognition system,even form after characters recognition system and duplex form characters recognition system.After using this scheme,the accuracy of independent form characters recognition system,even form before characters recognition system,even form after characters recognition system and duplex form characters recognition system raised 6.78%,9.59%,7.31% and 9.21% separately.
Keywords/Search Tags:Uygur, online handwriting word recognition, conjoined section segmentation, characters segmentation, characters recognition
PDF Full Text Request
Related items