Font Size: a A A

Research On On-Line Uyghur Character Recognition Technology Based On Features Combination

Posted on:2014-02-18Degree:MasterType:Thesis
Country:ChinaCandidate:L F Y K H E ZuFull Text:PDF
GTID:2248330398467712Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
With the popularity of PDA, Smart phone, Learning machine and other portablehandheld devices, pen and paper has been increasingly replaced by screen andkeyboard, and Handwriting input is finding wider and wider application.Implementation of the online handwritten character recognition system that free towrite, a recognition, high recognition rate has been the dream of people. As a popularlanguage in Uyghur and other ethnic minority areas, study its handwritten recognitionmethod is very conducive to promote the development of software of other people, ishelpful to improve the level of information processing technology of Xinjiangminority nationality. Also has positive significance to promote the development ofmulti-language information processing.This article systematically conducted a theoretical and experimental research onon-line handwritten Uyghur character recognition, including preprocessing, featureextraction, classification and so on. In preprocessing, we used a combination methodof linear normalization and nonlinear normalization. In feature extraction, extractedperipheral contour feature, shape feature of additional strokes, stroke number feature,additional part’s location feature, grid direction feature and feature of bottom-up andleft-right density ratio for32independent forms of Uyghur characters. Througheffective combination of features get the feature set. In feature classificationprocedure, we adopted the traditional classifier based on Minimum distance and KNN,and in which there were introduced the pre-classification mechanism, so that furtherenhanced the classification speed indeed. In this paper, the combination feature ofgrid direction feature, shape feature of additional strokes and feature of bottom-up andleft-right density ratio using KNN classifier is the highest recognition rate, reaching78.1%. This shows that the effective combination of different features is feasible, andhas the potential to continue study. The thought and method taken by this paper, has acertain generality to others inAltaic language family, has certain reference value.
Keywords/Search Tags:Uyghur characters, On-line handwritten recognition, Feature extraction, Minimum distance classifier, KNN classifier
PDF Full Text Request
Related items