| 摘要 |
<p>Disclosed is a character recognizing system for machine-printed characters such as alphanumerics, in which the characters are recognized through their comparison with a plurality of standard patterns of the characters. In a conventional character recognizing system, the character recognition is achieved by comparison of an input pattern with a plurality of predetermined standard patterns. For this purpose, a great number of standard patterns are provided, for example, at least 10 kinds of standard patterns for numerals, and at least 26 kinds for letters. Further, in case the characters to be recognized have various kinds of character-types or fonts, more standard patterns are necessary so as to recognize the input patterns with minimum errors. Therefore, the conventional system requires a memory means of great capacity. This makes the system more expensive as a whole. The system disclosed provides improved character recognizing system requiring a smaller capacity of the memory means. In the present system a great number of standard patterns are classified into a plurality of groups each having one common standard patterns turn having a pattern arrangement in common with the standard pattern of each group, and a plurality of non-common standard patterns representing differences between the standard patterns and the common standard pattern. An input pattern read out by scanning a printed character is first compared with the common standard pattern of any group. Only when the difference from the common standard pattern is less than a predetermined value is the input pattern compared with the non-common standard patterns in the same group.</p> |
