| 摘要 |
PURPOSE:To approximate the features of an individual in a dictionary to the actual one corresponding to the number of registration time and to improve a recognition ratio by initially recognizing a character based on an ordinary dictionary and recognizing it based on a personal dictionary as the number of recognition times increases. CONSTITUTION:Input character data is inputted to a recognition part 12 through an input part 11. The recognition part 12 recognizes the input character and outputs it to an output part 13 and a decision part 14. It decides whether the recognized result is rejected or not. If the result is accepted, a dictionary vector calculated by a vector calculation part 15 is registered in the personal dictionary 16. In that case the dictionary vector in the ordinary dictionary 18 is initially written in the dictionary 16, and the initial recognition is executed based on the dictionary vector in the dictionary 18. As the number of registration times in the dictionary 16 increases, recognition is made according to the dictionary 16. With the small number of recognition times, recognition is mainly made according to the dictionary 18, and as said number increases, recognition is made according to the dictionary 16 with many features. Therefore the recognition ratio can be improved.
|
