ONLINE NON-CURSIVE JAWI CHARACTER RECOGNITION SYSTEM

Abstract

Jawi script is an important Malay heritage that has been in general, replaced by the Roman script drastically. From a dominant writing in Malay world, the usage of Jawi is confined mostly in Islamic religious context nowadays. As an initiative to encourage the learning of Jawi, this research proposed an online non-cursive Jawi Character Recognition system using a Single Layer with Competitive Network and Supervised Learning method. The aim of this research is to develop software that able to assist Jawi illiterate user in learning basic Jawi, and able to recognize the mouse-drawn non-cursive Jawi character. “Online” in this context means that the system recognized the character as soon as the character is drawn on a writing surface, while “Offline” means that the system received the input from an image scanner after the writing is completed. To improve the recognition of the mouse-drawn character, the system uses neural network training algorithm called Supervised Learning to receive new character pattern in order to strengthen the weight of the pixels.