| 摘要 |
PURPOSE:To select a proper candidate item by averaging the distance of each candidate item when >=2 candidate items appear at equal distance, and selecting a final candidate item from the means value. CONSTITUTION:A feature quantity extracted from an input pattern is compared with feature quantities in a previously supplied dictionary to find distance. Then this found distance is compared with a reference threshold level and items which remain in a candidate group are registered. Then, the candidate item having the shortest distance is selected among the registered candidate items. When plural items having the shortest distance appear, an equal-distance flag is set and the candidates are registered. When all the registered candidates are processed, it is decided whether or not the equal-distance flag is set; when not, the minimum one is determined as a candidate item and when the flag is set, the mean distance values of the respective candidate items having the equal distance are found and the candidate having the smallest means value is selected as the final candidate item.
|
