PURPOSE: To make the control of recognition processing technically easy, to improve a speed for the processing and to miniaturize a character recognizing device by executing weighing to the similarity of a local area and selecting a result, for which difference between the most similar area and secondary similar area is made maximum, from plural discriminated results.
CONSTITUTION: A feature plane for executing pattern matching concerning the character of a recognizing object is divided into several areas P1 - P9 and the similarity is calculated for the respective areas P1 - P9. When the similarity of the first candidate is closed to the similarity of the second candidate in the recognized results using the similarity of the whole character, the similarity for each area calculated in advance is weighed and added and the whole similarity is calculated. Out of the similarities obtained by repeating the same operation while changing the several kinds of the areas to be weighed, the candidate, for which the difference between the similarity of the first candidate and the similarity of the second candidate is made maximum, is selected and the new recognized result is obtained. Thus, the control of the recognition processing is made easy and the processing can be executed at high speed. Then, the device can be miniaturized.
WO/2022/120997 | DISTRIBUTED SLAM SYSTEM AND LEARNING METHOD THEREFOR |
JPH0535924 | AUTOMATIC MUSICAL SCORE RECOGNIZING DEVICE |
JPS6068487 | IDENTIFICATION METHOD AND APPARATUS |
FUJI FACOM CORP
JPS6486286A | 1989-03-30 | |||
JPS5380922A | 1978-07-17 |