Main Article Content
Abstract
Achievement rates of the Optical Character Recognition (OCR) algorithms of published Gurmukhi Script documents are fairly remarkable with state-of-art, levels of accuracy ranging from 85 to 95 percent for specific. Nonetheless, additional enhancement of this level of accuracy is required to support actual applications. One of the bottle necks is known as similar looking characters in further improving the accuracy. In this paper, we outline the similar looking characters in Gurmukhi Script and creation of a specialized classifier for these closely matching characters. State-of-art OCR's output is used and the similar looking character set groups are further provided to an advanced classifier to improve precision. Support vector machine (SVM) algorithm is used to make this classifier and it uses feature vectors taken from near matching characters' spectral coefficients of projection histogram signals.