Java character recognition using Hopfield network

3rd Annual Applied Science and Engineering Conference

工业技术;自然科学

Widodo, W.^1 ; Priambodo, R.A.^1 ; Adhi, B.P.^1

Department of Informatics Education, Universitas Negeri Jakarta, Jl.Rawamangun Muka, Jakarta, Indonesia^1

关键词: Associative memory;    Bitmap formats;    Character images;    Hopfield Networks;    Indonesians;    Java language;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/434/1/012034/pdf
DOI  :  10.1088/1757-899X/434/1/012034

来源: IOP

【 摘 要 】

Java character is a legacy in Indonesian culture. It is a character to express Java language for Java people. Java characters have high level of its complexity because it consists of three types, basic characters (carakan), numbers, and complementary characters (sandhangan). This research performed an experiment to recognize Java character using Hopfield network. Firstly, Java character image will be transformed into bitmap format, then we process the data using Hopfield network as an associative memory to recognize. The experimental result denoted that average accuracy is 88%. It is better than some other neural networks method used by other researchers.