Implementation of web-based Japanese digital handwriting OCR using chain code and manhattan distance

Hiragana is the original writing system from Japan. The Japanese language itself has spread to many other countries, including Indonesia. However, Japanese is not an international language, so it is not easy to learn. In order to help people who try to learn Japanese, we need a system that can detect digital-based handwriting in Japanese, especially Hiragana. This research will be focus on designing a digital-based handwriting recognition application using feature extraction of chain code and manhattan distance to calculate the distance between 2 or more Hiragana. In building this web-based application, the PHP programming language and database system integration are used. There are 5 stages process in this system : pre-processing of image, image segmentation, linear spatial filtering to find the contour of the image, feature extraction of chain code, and then find the distance with manhattan. The distance of the tested Hiragana will be compared with trained Hiragana in the database to find the smallest distance. The results of the tests showed that 60.26% of Hiragana writings were successfully recognized and 39.74% of failure. The failure has happened due to the lack of object size variations in the trained data because each user writes with different sizes, therefore the resulting feature values are also different.