JPH0282380A - Character recognizing method - Google Patents

Abstract

PURPOSE:To increase the using efficiency of a memory and to improve the processing speed of character recognition by unifying collating patterns of characters, whose patterns are similar, to a single common pattern, and grouping the collating patterns. CONSTITUTION:When HIRAGANA (Japanese syllabary 'Ku' is read, the read character is outputted to a pattern collating part 2 as the character pattern, and the pattern collating part 2 collates the character pattern with the collating pattern in a pattern dictionary 1. Further, groups corresponding to the collating pattern to a certain extent are successively selected with the group at the highest corresponding degree as a first candidate. Next, a character code HIRAGANA 'Ku' and an arithmetic mark '<' belonging to the first candidate group are outputted to a display part, etc., as the read character candidates, and the character corresponding to the character codes are displayed. Out of the displayed characters, an operator selects the necessary character, and the character read by a reader is recognized. Thus, the using efficiency of memory can be increased, and the processing speed of the character recognition can be improved.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention provides a method for recognizing read characters by comparing pre-stored character patterns with read handwritten characters in a character reading device or the like. Regarding.

[Prior Art and Problems to be Solved by the Invention] Conventional character recognition methods generally involve storing matching patterns for each character-character-character in a memory as a pattern dictionary. Also, for a character reading device, for example, the number "1", the lowercase English letter "l", the uppercase English letter "I"
It is difficult to recognize characters with similar character patterns, such as , by simply comparing them with matching patterns in a pattern dictionary. Therefore, after selecting multiple candidate characters by comparing the read characters with the character pattern corresponding to each character, the read characters are recognized by narrowing down to a single character through post-processing that takes into account the surrounding context. ing.

In other words, although it is impossible to narrow down characters with similar character patterns to one just by matching the character patterns, a matching pattern corresponding to each of these similar characters is memorized for each character. , and memory usage efficiency is low. Furthermore, if extra data is stored in the memory, there is a problem that the processing speed for character recognition becomes slow.

The present invention has been made to solve these problems, and aims to provide a character recognition method that efficiently utilizes memory and increases the processing speed of character recognition.

[Means to solve the problem]

The character recognition method of the present invention selects a character pattern corresponding to the read character by comparing a stored character pattern with a read character, and then selects a character pattern corresponding to the read character, and then selects a character pattern based on a stored character code associated with the character pattern. In this method, a character pattern that is common to multiple characters is stored as a common matching pattern for these characters, and when the matching pattern of a read character is this common pattern, it is associated with the common pattern. It is characterized by recognizing characters based on a plurality of character codes stored in the memory.

[Effect]

The character recognition method of the present invention selects a matching pattern corresponding to the read character by comparing the read character with a stored matching pattern, and stores the stored character in association with the selected matching pattern. Recognize characters based on codes. If a plurality of character codes are stored in association with the matching pattern, the operator selects the desired character code from among them.

〔Example〕

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below based on drawings showing embodiments thereof. FIG. 1 is a block diagram showing the main structure of a character reading device that implements the method of the present invention.

1 in the figure is a pattern dictionary that stores matching patterns, and the pattern matching unit 2 matches the read character pattern with the matching pattern read from the pattern dictionary 1 to find the corresponding character code. Output to display unit, etc.

Next, a method of character recognition will be explained. FIG. 2 is a conceptual diagram illustrating six groups of matching patterns that are stored in the pattern dictionary 1 after being classified into groups such as grouping together characters with similar character patterns. In the figure, the upper half of each group shows the image of the matching pattern, and the lower half shows the character code that makes this the matching pattern. Group 1 is the Arabic numeral ``1'', 2 lowercase letters ``1'' and uppercase English letter ``■'', group 2 is the Chinese numeral ``-'', and the arithmetic symbol ``-''.
and the English symbol "- (hyphen)", group 3 is the kanji "chi" and katakana "chi", group 4 is the Arabic numeral "0"
, uppercase English letter “0”, lowercase English letter “0” and symbol “○ (circle)”
", group 5 is divided into hiragana "<" and arithmetic symbol "<", and group 6 is divided into "a" alone.

When the hiragana "ri" is read, the read character is output as a character pattern to the pattern matching section 2, and the pattern matching section 2 matches this character pattern with the matching pattern in the Tumaturn dictionary book 1, and then outputs it to the i-tan (for matching). Groups that match to some extent are selected in order, with group 5 having the highest matching degree being selected as the first candidate.

Next, the character code “<” belonging to the first candidate group 5
and "<" are output to a display unit etc. as read character candidates, and characters corresponding to these character codes are displayed. An operator selects a desired character from among the displayed characters, and the character read by the reading device is recognized.

〔Effect of the invention〕

The character recognition method of the present invention improves memory usage efficiency by combining matching patterns of characters with similar character patterns into one common pattern and grouping the matching patterns, and also improves memory usage efficiency by combining matching patterns of characters with similar character patterns into one common pattern. This has the excellent effect of shortening the time required for matching with the original pattern and improving the processing speed of character recognition.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the main structure of a character reading device that implements the method of the present invention, and FIG. 2 is a conceptual diagram showing an example of grouped matching patterns. 1...Pattern dictionary 2...Patent for pattern matching unit Applicant: Sanyo Electric Co., Ltd. Agent Patent attorney: Noboru Kono

Claims (1)

[Claims] 1. Select a character pattern corresponding to the read character by comparing the stored character pattern with the read character, and select the character pattern based on the stored character code in association with the character pattern. In this method, a character pattern that is common to multiple characters is stored as a common matching pattern for these characters, and when the matching pattern for a read character is this common pattern, it is associated with the common pattern. A character recognition method characterized by recognizing characters based on a plurality of character codes stored in a computer.