| 摘要 |
PURPOSE:To make accurate read possible, by adopting Katakana (square form of Japanese syllabary) read besides distinction of Kanji (Chinese character) in the character read system of the OCR and so on. CONSTITUTION:Unknown pattern 12 of the Katakana part corresponding to one character of Kanji on the form is read and is subjected to Katakana distinction 13, and one character (a) of Kanji 14 presumed from Katakana is brought into the register of distinction 18. Next, unknown pattern 15 of one character of Kanji is read and is subjected to Kanji distinction 16, and one character (b) out of resembling Kanji 17 is brought into another register. The resemblance degree is calculated between these two characters (a) and (b) by the fix point sampling method and so on. This operation is performed for all of combinations of two Kanji, and Kanji is distinguished 19 finally on a basis of numerics of resemblance degrees. |
