Hiragana learning support system with on-line stroke detection

In this paper, we describe a Hiragana learning support system with on-line stroke detection. We introduce new methods of detecting stroke ends such as "Tome, Hane and Harai". Our system makes use of detection results to score and guide written characters. Our system supports three learning styles. The first style is watching animation of Hiragana writing. The second style is tracing thin mark of characters. The last style is writing characters without thin mark. The system evaluates the characters based on the recognition results of the characters and the detection results of Tome, Hane and Harai.