摘要 |
PURPOSE:To heighten the accuracy of recognition of Chinese characters and to discriminate Chinese characters from non-Chinese characters, by discriminating a Chinese character and non-Chinese character of the same shape utilizing Japanese phonetic syllabaries which easier to recognize than Chinese characters. CONSTITUTION:Character (Japanese phonetic syllabaries 12 and text 11) on a medium 1 to be read are scanned and read by a photoelectric conversion device 2 such as CCD etc., and inputted to a recognition pretreatment circuit 3. The circuit 3 performs characteristic extraction of each character and selects candidates of each character and outputs them. Candidate output from the circuit 3 is distributed according to the Japanese syllabaries and the text to buffers 4, 5 that stores the Japanese syllabaries and the text. Japanese syllabary and Chinese character correspondence output read out and text candidate output from the buffer 5 are collated successively, and candidates of Chinese character in the text are selected and reduced to 2-3 candidates from candidaea category of 20 for each character. Character candidate output from a candidate buffer 8 in which selected character candidates are stored is inputted to a selector circuit 9. Here, character candidates are arranged sideways to constitute a sentence and an ultimately correct category is selected. |