Paper Title :Comparison Of SVM And K-Nn Classifiers For Recognizing Degraded Printed Gurmukhi Numerals
Author :Seema, Nishu Goyal
Article Citation :Seema ,Nishu Goyal ,
(2015 ) " Comparison Of SVM And K-Nn Classifiers For Recognizing Degraded Printed Gurmukhi Numerals " ,
International Journal of Advance Computational Engineering and Networking (IJACEN) ,
pp. 44-48,
Volume-3, Issue-2
Abstract : Character recognition is very important areas in the field of document analysis and recognition. Character
recognition can be performed on fine printed, handwritten or typewritten text. The accuracy and performance of optical
character recognition system depends on the printing quality of the text document. Various researchers have done prominent
work in recognition of printed and handwritten text using OCR. Limited authors have worked for recognition of degraded
Gurmukhi Numerals using OCR. In present paper, the main focus is to recognize printed degraded Gurmukhi Numerals using
OCR. Binarization technique was applied to recognize degraded printed Gurmukhi numerals. Different types of printed
degradations such as broken characters, background noise problem heavily printed and shape variant characters were
considered during recognition of degraded Gurmukhi numerals. Various structural and statistical features e.g. zoning,
transition features, distance profile features and neighbor pixel zone, were used for generating feature sets to recognize
printed Gurmukhi numerals using support vector machine (SVM) and k-nearest neighbors’ (K-NN) method. Normalized
accuracy and error are calculated to evaluate the performance of various recognition techniques and their combination in
various training/tested case study formats using SVM and K-NN classifiers.
Keywords: OCR, Degraded Gurmukhi numerals, Feature extraction, Zoning, Classifer, Support vector machine, K-nearest
neighbors.
Type : Research paper
Published : Volume-3, Issue-2
DOIONLINE NO - IJACEN-IRAJ-DOIONLINE-1680
View Here
Copyright: © Institute of Research and Journals
|
|
| |
|
PDF |
| |
Viewed - 81 |
| |
Published on 2015-01-30 |
|