ABSTRACT
ABSTRACT: Optical Character Recognition (OCR) converts images containing handwritten or printed characters into an editable format. OCR systems have wide applications such as processing bank cheques, or document conversion of legal papers. One major concern in developing a character recognition system is the selection of efficient features. Major challenges in Malayalam handwritten character recognition are varying writing styles, the presence of compound characters and of similarly shaped characters. Mixing up old and new styles of writing adds additional complexity in HCR systems. In this paper, an alternative that allows the use of histogram of oriented gradients is presented. The proposal consists of using HOG for feature extraction, then using SVM as the classifier, Malayalam character recognition can be carried out with more quality than that obtained in the existing methods.
1 INTRODUCTION
