Please use this identifier to cite or link to this item:
Title: Off-Line Handwritten Character Recognition of Manipuri Script
Authors: Singh, Th. Thokchom
Supervisor: Bawa, Seema
Bansal, P.K.
Vig, Renu
Keywords: Manipuri script;Offline handwritten character recognition;Meetei
Issue Date: 7-Aug-2017
Abstract: As computers become increasingly integrated into everyday life and since we want computers to be genuinely intelligent and to interact naturally with us, therefore the benefits of automatic recognition of handwritten digits and characters are obvious. This thesis initially presents a brief report on linguistic survey of Manipuri script along with historical background and revival movement of the Script. Many aspects of handwritten digit and character recognition research for English as well as some Indian scripts have also been reviewed. In this thesis, Handwritten Character Recognition System of Manipuri Script, HCRMS, has been presented for segmenting lines, words, non-touching characters and isolated digits and recognizing the handwritten isolated digits and non-touching characters. Analysis of different strategies for segmenting non-touching characters of handwritten word in different zones with the analysis of vertical, horizontal projection profiles and connected component analysis are given. After size normalization of the extracted component, probability features based on the size and slant invariant signatures features are extracted. The handwriting recognition results for digits and characters using K-L divergence with probabilistic features are presented. Fuzzy feature extraction technique from the resized component with zoning is also presented. Then Hybrid feature is proposed by combining the two feature sets for the better recognition rates of the characters using feed forward backpropagation neural network. The experimental results show that the choice of the features affects the performance of the classifier and the proposed hybrid feature set gives better recognition rate. The generalization of the recognition process has been improved with the size and slant invariant signatures features of the probabilistic feature method. Experimental results indicate that the proposed recognition system performs well and is robust to the writing variations that exist between persons and for a single person at different instances, thus being promising for user independent character recognition and tolerant to random noise degradations of the characters.
Description: PhD thesis
Appears in Collections:Doctoral Theses@CSED

Files in This Item:
File Description SizeFormat 
4579.pdf7.34 MBAdobe PDFThumbnail

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.