Handwritten Character- Recognition Through Feature Exttraction Using Artificial Neural Networks

Abstract

Today’s computer can be fast, accurate and knowledgeable. But they are far from being intelligent; computers are still unable to communicate with human beings in natural forms like written languages, speeches, pictures and images. With the rapid development in the computer facilities users have now turned their attention to interact with computers in their local languages, so it is more user friendly. In Karnataka, HINDI is the local and official language. it is very convenient and faster to enter a HINDI document into computer by handwriting rather than by typing using existing HINDI converted English keyboard, where key combinations have to be used. This necessitates the development of an efficient online system to recognize HINDI handwritten characters or words. The concept in the online character recognition system is to capture a character as a sequence of [x(t),y(t)] points while the character is being written on a computer screen. Trajectory curvature, shape, size etc …., are extracted from the characters. Each point of the input sequence is coded as a set of these features for recognition.