摘要 |
PURPOSE:To improve an operability in character recognition by directly displaying input KANA (Japanese syllabary) rows in case the input characters are less than a predetermined value. CONSTITUTION:A phoneme identifying part 1 performs phoneme identification of input voice and outputs the results of identification and probabilities thereof. A character buffer 2 stores the results of identification until there is an input by a final decision key 12. A sentence candidate producing part 3 produces sentence candidates in the order of the lesser value of the sum total of the probabilities from the combinations of the obtained phoneme matrices. A grammar collating part 5 collates each of the sentence candidates with a grammar dictionary 4 and the coincident ones are outputted as identification results in the form of KANJI (Chinese characters) -KANA (Japanese syllabary) mixed code. A character counter 7 counts the number of characters inputted to the buffer 2. A comparator part 8 sets a changeover switch 9 to A side when the value of the counter 7 is less than a preset value of N and to B side when the value is more than it. A display part 12 displays the first candidate group of the buffer 2 in the form of KANA letters and outputs KANJI-KANA sentence when the switch is set to B side.
|