期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2009
卷号：9
期号：6
页码：154-158
出版社：International Journal of Computer Science and Network Security
摘要：Recognition of Bengali characters by neural network has received much attention already. Better preprocessing and feature extraction may fail to give better accuracy if the recognition process (includes training and testing) is incompetent. In this paper, some heuristics are proposed to make the back-propagation training algorithm perform better. Finally, character recognition accuracy results obtained from different algorithms are compared.
