Recognition and Translation of the Myanmar Printed Text Based on Hopfield Neural Network

• Published in: 6th Asia-Pacific Symposium on Information and Telecommunication Technologies pp. 99 - 104
• Main Authors: Swe, T., Tin, P.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 2005
• Subjects: Character recognition; Hopfield neural networks; Java; MATLAB; Microcomputers; Myanmar Printed Character; Neural Network; Optical character recognition software; Read-write memory; recognition and translation; region based segmentation; skew detection; Text recognition
• ISBN: 4885522161; 9784885522161
• DOI: 10.1109/APSITT.2005.203638

Characters are printed in various languages in the world. Printed character recognition is one of the oldest fields of research since the advent of computers. However it is an open field for researchers due to the challenging nature of perception and recognition. Myanmar language is the main language and Myanmar printed characters are widely used by over 85% of 49.5 million populations in Myanmar. Although there are more than 8 different languages used in Myanmar. Myanmar printed characters are mostly used in the common Myanmar application forms. The objective of this paper is to develop an optical character recognition for the Myanmar printed characters and the translation of Myanmar printed text for many other people (who don't understand the Myanmar language) in Myanmar to understand Myanmar words. In our experimental system, we will use Myanmar standard application forms. They are scanned and the reading system is capable of extracting the printed text from standard student application forms, the segmenting of the Myanmar printed characters, recognizing character system to identify Myanmar printed words, and translating them into user's own language. Our algorithms, Myanmar character recognition algorithm is extended from the Hopfield neural network method. Our system can detect about 97.56% of Myanmar printed characters. The main program of character recognition is written in MATLAB version 6.5 and Java on IBM compatible personal computers (512 MB of RAM, 3.04 GHz) and was successfully implemented and used Cannon N340P/n640P scanner