A Lightweight Chinese Character Recognition Model for Elementary Level Hanzi Learning Application

Abstract

The Chinese language is widely spoken and written by a quarter of the earth's population. Its usage is recently increased due to the rise of China as a new world power in trade and economy. This attracts new learners of Chinese and Chinese is often taught as early as elementary school in countries such as Indonesia, which regard Chinese as a new foreign trade and social language. However, without proper and continuous exercise, mastering Chinese, especially the written, is a big challenge. Previous studies has proposed and affirmed the use of information technology as a learning aid to study Chinese. They show positive results, but has left out the writing exercise section. This research proposes a modest Optical Character Recognition (OCR) model applicable to aid learning of writing Chinese characters, also known as Hanzi, for elementary education level. The goal aimed is not just the functionality, but due to its modesty, it should be able to be applied to a broader condition; in wider range of devices and by wider level of programmers. Experiment results shown that for the defined environment, the model give an acceptable accuracy of 95% in recognising handwritten Chinese characters. However, if it is planned to be applied using a more complex set of characters and writing styles, the statistical features used should be replaced and improved.