Please use this identifier to cite or link to this item:
http://hdl.handle.net/10603/10438
Title: | Some studies on recognition of handwritten Telugu characters |
Researcher: | Sita Mahalakshmi, T |
Guide(s): | Vinaya Babu, A |
Keywords: | Computer Science Neural Networks Support Vector Machines Kernel Trick |
Upload Date: | 7-Aug-2013 |
University: | Acharya Nagarjuna University |
Completed Date: | 2011 |
Abstract: | Computers are all pervasive essentialities in today s world and technologies that improve our interactions with them are growing at a never before rate. One particular kind of techniques that enable us to interact with them better are the recognition methods, which include character or speech recognition. Character recognition is of two types - off-line and on-line. This thesis predominantly focuses on off-line character recognition- an application of pattern recognition. A typical pattern recognition system consists of three phases namely data acquisition, feature extraction and classification. For classification, knowledge acquisition about the domain is an important factor. If the domain knowledge is well defined then the characters can be classified with certainty. This work is focused on off-line recognition of Telugu characters. The data set for Telugu alphabet are neither available on on-line resources nor are they available commercially. As a result, it became necessary to acquire the character images from different people for this research. Due to the wide range of variations in the handwriting, the images are pre-processed and a set of 41 features are extracted from the images. As the generalization ability of the classifier depends on the number of attributes, dimensionality reduction has been performed using factor analysis with SPSS 16.0 and validation and consistency of the factors has been confirmed with confirmatory factor analysis. The second stage of pattern recognition is classification. In this work, the classification has been performed with neural networks, support vector machines and genetic algorithms. Two types of neural networks are used - radial basis function (RBF) and probabilistic neural networks (PNN) - due to the advantage of universal approximation property and good generalization ability. The learning task in the neural networks depends on the number of training samples while, support vector machines which work on simple geometric interpretation. |
Pagination: | xiii, 152p. |
URI: | http://hdl.handle.net/10603/10438 |
Appears in Departments: | Department of Computer Science & Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
01_title.pdf | Attached File | 12.19 kB | Adobe PDF | View/Open |
02_declaration.pdf | 9.92 kB | Adobe PDF | View/Open | |
03_certificate.pdf | 32.36 kB | Adobe PDF | View/Open | |
04_dedication.pdf | 29.6 kB | Adobe PDF | View/Open | |
05_acknowledgements.pdf | 19.61 kB | Adobe PDF | View/Open | |
06_contents.pdf | 30.62 kB | Adobe PDF | View/Open | |
07_list of figures.pdf | 21.41 kB | Adobe PDF | View/Open | |
08_list of tables.pdf | 25.23 kB | Adobe PDF | View/Open | |
09_abstract.pdf | 20.12 kB | Adobe PDF | View/Open | |
10_chapter 1.pdf | 132.24 kB | Adobe PDF | View/Open | |
11_chapter 2.pdf | 168.12 kB | Adobe PDF | View/Open | |
12_chapter 3.pdf | 320.77 kB | Adobe PDF | View/Open | |
14_chapter 4.pdf | 353.7 kB | Adobe PDF | View/Open | |
15_chapter 5.pdf | 527.36 kB | Adobe PDF | View/Open | |
16_chapter 6.pdf | 419.24 kB | Adobe PDF | View/Open | |
17_summary.pdf | 84.34 kB | Adobe PDF | View/Open | |
18_references.pdf | 94.74 kB | Adobe PDF | View/Open | |
19_appendix.pdf | 873.77 kB | Adobe PDF | View/Open |
Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Altmetric Badge: