Image processing method in implementation of handwriting identification for Japanese katakana characters

Abstract

Japanese is one of the popular languages and is spoken in international world. Japan ranks fourth out of ten commonly used languages in the world. Pattern recognition techniques have developed over the time and often used to solve problems. Pattern recognition technique is used for identification of handwriting, images, etc. Japanese Katakana handwriting with all the complexity turns out to have strict rules in writing. In its application, there is an inaccuracy in the writing of Katakana letters. This is caused by the many variations and procedures of writing Katakana. The procedure of writing Katakana letters has its own rules especially regarding the number of scratches. Therefore, in this research, authors implemented Recurrent Neural Network to identify the words based on the Katakana handwriting.