| 摘要 |
PURPOSE:To efficiently prepare a dictionary which can be applied to the reading processing of a hand-written character by counting the similarity between respective input patterns, preparing the pattern set, further, counting the similarity between respective sets and preparing a standard pattern set. CONSTITUTION:One pattern is selected from an input pattern set storing part 21a by a control command C1 from a control part 20 and outputted to a similarity counting part 25. On the other hand, one pattern set is selected from a coarse classifying pattern set storing part 22 by a control command C3 and outputted to the counting part 25. By a control command C2, the counting part 25 executes the similarity counting and determines the set of an input pattern from the storing part 21a. On the other hand, the updated set pattern vector is stored to the storing part 22 and executed to all input patterns. Thereafter, in such a way, similarity counting parts 26-28 executes the similarity counting by commands C2 and C3. These classifying results are given to the control part 20 and to a base classifying pattern set storing part 23, always, a new classifying pattern vector is stored. Thus, the dictionary is prepared.
|
