Please use this identifier to cite or link to this item:
Title: Handwritten character recognition using hybrid techniques
Researcher: Jagadeesh Kannan R
Guide(s): Suresh R M
Keywords: Structural Hidden Markov Model
Hidden Markov Model
Character recognition system
Tamil character
Upload Date: 11-Jul-2013
University: Anna University
Completed Date: 01/04/2011
Abstract: The primary objective of this research work is to develop a character recognition system, which can further be developed to recognize all printed and handwritten Tamil characters. Such a system will be extremely useful to digitize innumerable ancient documents available both on paper and palm leaves and are dated few hundred years back. The problem considered is to develop a recognition system for Printed and Handwritten Tamil and Numeral Characters. An overview of the research work is presented below. The first component of this research work focuses on technique to recognize handwritten Tamil characters using a Hidden Markov Model (HMM) approach, for a subset of the Tamil alphabet. The second component of this research work also focuses on technique to recognize handwritten Tamil characters using an Octal Graph approach. The third component introduces a methodology to recognize handwritten characters using Structural Hidden Markov Models (SHMM). The proposed approach is motivated by the need to model complex structures which are encountered in many areas such as speech/handwriting recognition, content-based information retrieval etc. The observations considered are strings that produce the structures. These observations are related in the sense they all contribute to produce a particular structure. The recognition efficiency of the system is 96.5%. The results reported in this component shows that the Structural Hidden Markov Model (SHMM) produces better recognition than the Hidden Markov Model. A character recognition system was successfully developed for subset of Tamil characters and numerals and found to perform reasonably well with sufficient accuracy. With this preliminary study of character recognition systems it is planned to continue the work in future. A more complete OCR will evolve in public domain with this work as the starting point, so that the primary source of motivation will benefit from the efforts.
Pagination: xvii, 114p.
Appears in Departments:Faculty of Information and Communication Engineering

Files in This Item:
File Description SizeFormat 
01_title.pdfAttached File50.33 kBAdobe PDFView/Open
02_certificates.pdf568.08 kBAdobe PDFView/Open
03_abstract.pdf20.88 kBAdobe PDFView/Open
04_acknowledgement.pdf17.44 kBAdobe PDFView/Open
05_contents.pdf38.17 kBAdobe PDFView/Open
06_chapter 1.pdf166 kBAdobe PDFView/Open
07_chapter 2.pdf93.88 kBAdobe PDFView/Open
08_chapter 3.pdf197.29 kBAdobe PDFView/Open
09_chapter 4.pdf184.54 kBAdobe PDFView/Open
10_chapter 5.pdf313.07 kBAdobe PDFView/Open
11_chapter 6.pdf40.96 kBAdobe PDFView/Open
12_references.pdf88.15 kBAdobe PDFView/Open
13_publications.pdf17.01 kBAdobe PDFView/Open
14_vitae.pdf15.12 kBAdobe PDFView/Open

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

Altmetric Badge: