On-line Japanese character recognition system for visually disabled persons

Pen-based interface systems for online recognition allow effective Japanese character entry. We propose such a system for visually disabled persons. Characters written by visually disabled persons have significant shape distortions. This paper describes a basic composition of a pen-based entry system and a character recognition method for visually disabled persons. Instead of a keyboard, an electronic tablet and a hand controller are used. So is a voice synthesizer device which indicates a selected character. As a result of distortion analysis of characters written by visually disabled persons, we propose a new method which is able to recognize such characters, using 2 algorithms. One algorithm is used for characters with more than 5 strokes. It is based on the evaluation of relative position of the stroke representative points between two neighboring strokes. The other algorithm is used for the characters having less than 4 strokes. It based on the precise analysis of stroke shapes. As a result, average recognition rate of 12 persons is 96.4%, for Kanji characters having more than 5 strokes, and 89.7% for characters having less than 4 strokes which include Kanji, Hiragana, Katakana, numerals, and symbols