Handwritten Gurumukhi Character Recognition Using Neural Networks

Abstract

Today, Handwritten Character recognition is one of the challenging computational processes. Some computational fields like artificial intelligence, expert systems have provided an important role in recognition of these handwritten Gurumukhi characters. There is competition between the speed and efficiency. The human mind can easily decipher these handwritten characters easily, accurately and speedily. The human mind can do it because of the presence of densely neural network in his mind. In 1948, Wiener proposed an idea of using the non-biological neural network to solve human processes. Using that non-biological neural network, many processes can be done easily and efficiently as human mind can do where the system can take the decisions to solve the problems. In pattern recognition, artificial neural network take decision to recognize the characters. In pattern recognition, first of all the Punjabi characters are digitized using any VLSI technology. After that these characters are pre-processed for removing noisy data. After preprocessing, feature extraction is performed in which mainly neural network is needed. Feature extraction is used to differentiate between the characters. The character is written on page and scanned. These characters are stored as an image of 32 * 32 matrix size. The image can be of any type i.e. .jpg, .bmp, .gif. The end user can test the additional characters if required.