Handwritten Malayalam Character Recognition System using Artificial Neural Networks

Abstract

This paper presents a system that recognizes Malayalam in handwritten form using artificial neural networks. A system for recognizing handwritten text (HCR) is a technique that is used for recognizing human handwritten text in any language. HCR is one of the research areas of recognition of patterns, which is still very challenging as perfect solutions have not yet been found. For certain foreign languages like English, Japanese, Chinese, etc, HCRs have been developed, which are reasonably good. But it is still premature for languages in India, especially for languages in south India. Because of the large character set, compound characters, presence of modifiers, and the curvature of characters in these languages, the task is quite complicated. This project aims to convert the photograph containing handwritten script into corresponding text. In this approach a trained ANN is used to identify the handwritten characters. The recognition system has been developed in python. The OpenCV library is used for performing different operations on the input image.This paper pertains to the first part of a work where individual characters alone are recognized. The continuation of the work which is ongoing, is to recognize complete sentences.