Segmentation, Recognition and Synthesis of Tamil Characters for Robotic Writing

Abstract

Currently, there are many ongoing researches that are targeted at making robots more human-like. In this paper, we are presenting a flexible algorithm that can allow a robot to write. This novel methodology includes two important phases namely, recognition of characters and writing of characters.Recognition is done via polynomial fitting. From a given text, each character is segmented, divided into five parts and decimated . These decimated samples are fitted to a polynomial of higher order and the coefficients of the polynomial are extracted as features of the character. These features are then compared with the standard template and are recognized. Writing is done via parametric equations, where the main idea lies in developing equations for each character. Each character is probed in a graphical sense and equations are obtained. Basic figures like cycloid, circle, spiral, ellipse, etc are studied and their features are utilised in developing equations. Thus, we obtain a set of primary equations for each character. By tracing this equation in a three dimensional environment, we get the required character.