A cursive on-line Hangul recognition system based on the combination of line segments

A cursive online Hangul (Korean script) recognition system called Jokjipke based on structural character information is presented. A segmentation method that observes the variations of the graphemes type is proposed. The cursive writing between graphemes is divided into line segments using the segmentation technique. Each line segment becomes a primitive of the recognition system. From the relational structure of line segments, the feature vectors are calculated by comparing the current line segment to the other line segments in the same grapheme. The recognition is then accomplished with the best-first search process. This research calculates the line segment that has the least amount of difference between the candidate grapheme string and the reference grapheme models. After learning each grapheme by 10 different person, the system reaches a recognition rate of 81.65% of the cursive Hangul character over the 10,396 characters. That is, more than 400 of the most frequently used characters are written by 25 different persons in order to show the writing style independence