JPH0554192A - Recognition character correction method - Google Patents

Abstract

PURPOSE:To provide a recognition character correction method capable of surely applying corrections when an error exists in the character code of the character size of a recognition character in a character recognition device. CONSTITUTION:A character is segmented by a character segment part 2 from a document pattern recognized by a character input part 1. The character is recognized by a character recognition part 3 and each maximum value of the longitudinal and lateral sizes is calculated by a maximum side calculation part 7. This maximum value and the length of the corresponding side of a character which is to be judged as a large character or a small character are compared and a character size decision part 4 decides that the character is a large character or a small character. When the contents of the character is different from the contents recognized in the character recognition part 3, it is corrected to the decided contents by a correction part 5.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a recognized character correction method for correcting an error relating to the character size of a recognized character at the time of character recognition in which a character pattern is read by a character recognition device and a character is recognized from the character pattern. It is about.

[0002]

2. Description of the Related Art In recent years, a character recognition device has been widely introduced as one of input terminal devices for various devices, and is being put to practical use. In the conventional recognition character size correction method in this character recognition device, the vertical or horizontal length of the character pattern is used as a reference. Was.

[0003]

However, the above-mentioned conventional method is effective to some extent when the target typeface or font is limited, but it is difficult to correct it when the target is widened. Had. This will be described using a specific example.

Although the typeface and the typeface are different from each other in FIG. 6, six kinds of kana characters (a, ya, u, you,
Yo, yo) and four types of kanji (smoke, swallow, monkey, rim). The same font is used in the column direction. In the figure, for example, c1 is a lower case "U" and c2 is an upper case "Y". In this case, the vertical length is lowercase c1 and uppercase c
It is larger than 2 and it can be seen that the conventional size-only method is difficult to correct even if there is an error in the character code, such as lowercase or uppercase, during recognition.

The present invention has been made in view of the above problems, and an object of the present invention is to provide a recognized character correction method which can surely correct an error in a character code relating to the character size at the time of character recognition.

[0006]

In order to achieve the above object, the recognized character correction method according to the present invention, when recognizing a character from a document pattern to be recognized, has a vertical size among characters of a recognized character string. The first maximum value of the maximum side length and the second maximum value of the maximum side length of the horizontal size are calculated, and among the recognized characters that have both uppercase and lowercase letters,
By comparing the length of the longer side in the vertical direction or the horizontal direction with the corresponding first or second maximum value described above, it is determined whether the character code relating to the character size is uppercase or lowercase.

[0007]

According to the recognized character correction method of the present invention by the above-mentioned procedure, the first and second maximum values, which are the maximum values of the vertical and horizontal sizes of the characters in the character string, are calculated, and the upper and lower case letters By comparing with the longer side of the kana character that has both, it is possible to accurately distinguish uppercase letters and lowercase letters by making the best use of the characteristics of the typeface or font of the character.

The principle of the present invention will be described below with reference to the drawings. Figure 6 has different typefaces and fonts, but the number of points is the same. Kana characters (ya, ya, yu, yu, yo, yo) that have both uppercase and lowercase letters and four kinds of kanji (smoke, swallow) , Monkey, rim). The column direction has the same typeface or font, that is, the same font. From this, the following can be said.

(1) In the same font, kana capital letters are always smaller than kana lower case letters. (2) In the same font, the ratio of the vertical and horizontal lengths of kana characters in uppercase and lowercase (hereinafter, the aspect ratio) is substantially equal.

(3) In the same font, the long sides of Kana and Kanji are almost equal. In the present invention, for characters with different typefaces and fonts,
Effective classification is performed by using the properties of (1), (2), and (3). That is, if it is known that the input character is a kana character that can be in both uppercase and lowercase (for example, "tsu" and "tsu"), the kana character and the kana character estimated to be the kana character respectively Uppercase and lowercase letters are determined by the area of the input characters based on the area of the uppercase letters estimated from the obtained vertical and horizontal lengths. That is, the maximum values of the vertical and horizontal lengths of all the character patterns existing in the character string to which the input character belongs, including the kanji, are obtained from the fact that the aspect ratio is almost the same as that of Kana capital letters.

When the vertical and horizontal lengths of a kanji character and an input kana character are M, N, a, and b, respectively, the vertical and horizontal lengths of the kanji character also have the maximum values. The maximum value is M, and the second maximum value is N. Next, in the case of uppercase / lowercase determination of kana characters, if a <b, the area of the kana capitalization is estimated to be a value obtained by multiplying the area kb of the input kana character by (N / b) * (N / b). .. In this method, the ratio of the long sides is calculated so that the character font is the same, and a value equal to the area ratio of the character size can be obtained. Therefore, the ratio of the area of input kana characters to kana capital letters is (b / N) · (b /
N). Similarly, when a ≧ b, (a / M)
(A / M). From this value, it is possible to determine whether it is uppercase or lowercase.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a functional block diagram of a character recognition apparatus in one embodiment using the recognized character correction method of the present invention. As shown in FIG. 1, 1 is a character input unit for photoelectrically converting recognition data of a document pattern to be recognized and storing it as binary data in a document pattern memory as shown in FIG. The character slicing unit 3 for outputting the character code that outputs the character code corresponding to the character pattern from the character slicing unit 2 is a character recognizing unit 4 for recognizing the character from the character recognizing unit 3 that has both uppercase and lowercase letters. , The maximum values of the vertical and horizontal lengths of the character pattern in the same character string given from the character cutout unit 2 are obtained, and the character pattern is capitalized or lowercased from this value and the vertical and horizontal lengths of the input pattern. The character size determination unit 5 that determines which of the two is used when the character size of the recognized character from the character recognition unit 3 is incorrect according to the result of the character size determination unit 4. Is a display unit for outputting the result of the correction unit 5, and 7 is a maximum value of each of the vertical and horizontal lengths of all the characters in the character string obtained from the character cutting unit 2 described above. It is a maximum side calculating unit that has a function of obtaining a vertical first maximum value and a horizontal second maximum value.

FIG. 2 is a block diagram showing the configuration of the character recognition device of this embodiment. Reference numeral 21 is a scanner for reading a document to be recognized, and outputs the read document as bit data. Reference numeral 22 denotes a readable / writable memory (hereinafter abbreviated as RAM), which is included in a document pattern area 23 that stores bit data from the scanner 21 and a character string cut out from the document pattern in the document pattern area 23. A character size area 27 for storing the size of a character pattern to be stored, a maximum side area 25 for storing the maximum values of the vertical and horizontal lengths of the character size area 27,
A character pattern area 24 for storing a character pattern, a character code area 26 for storing a character code obtained by recognizing the character pattern, and a register area 2 used for processing.
8 and. 29 is a read-only memory (hereinafter R
OM) and has a character size dictionary area 30 in which upper and lower case character codes and threshold values unique to the character codes are stored, and a program storage area 31 in which a control program according to the flowchart shown in FIG. 3 is stored. .. Three
Reference numeral 2 denotes a processing circuit that performs processing according to the control program stored in the program storage area 31. Reference numeral 33 is a keyboard for inputting data, and 34 is a character pattern area 2
4 is a display unit for displaying the character pattern in 4 and the character code in the character code area 26, or the character font corresponding thereto.

FIG. 3 shows the recognized character correction method of the present invention using the character recognition apparatus of the present embodiment having the above-described structure.
It will be described according to the flowchart of First, step S
1, the characters are cut out from the document pattern read from the document pattern to be recognized in the character input unit 1 by the processing of the character cutout unit 2, and the vertical and horizontal lengths of all the character patterns existing in the same character string. Character size area 2
Write to 7. In the next step S2, the maximum side calculation unit 7 searches for the maximum value, the first maximum value M, and the second maximum value N, from the vertical and horizontal lengths in the character size area 27, and the result is the maximum side area. Save to 25. Then step S
In step 3, character recognition processing is performed by the character recognition unit 3, and the recognized character and character code are stored in the character code area 26. In step S4, it is determined whether the recognized character in the character code area 26 is registered in the character size dictionary area 30. If registered, this character has a lower case character, so the character size determination process is performed in step S4, but if not, the process ends. In step S5, first, the size relationship between the length a and the length b of the input character pattern is checked. If a ≧ b, the process proceeds to step S6, and if a <b, the process proceeds to step S7. In step S6, the square of (a / M) is calculated, and in step S7, the square of (b / N) is calculated. In step S8, when the result of step S6 or S7 is smaller than the predetermined value registered in the character size dictionary area 30, it is determined to be a small letter, and when it is large, it is determined to be an upper case. Finally, in step S9, the correction unit 6 corrects the content of the character code area 26 when the determination result is different from the character code of the recognized character in the character recognition unit 3 and the character size of the recognized character is incorrect.

Next, the operation of the recognition target example character string pattern "National Electric" will be specifically described below. First, the character input unit 1 inputs the recognized character string pattern as a binary image and stores it in the document pattern area 23. Next, after the character string is cut out by the character cutting unit 2, the character cutting unit 2 separates the characters one by one. At this time, the vertical and horizontal lengths of all the character patterns existing in the same character string are set to the character size area. Write in 27. The character size area 27 at this point is shown in FIG. Maximum side calculator 7
According to the above, the maximum vertical and horizontal lengths are "na" and "ru", and M = 60 and N = 122 are obtained, respectively. Character recognition unit 3 recognizes the example character string "National Electric"
Is recognized as "Nasiyonarelektritsk". The character size determination unit 4 first determines whether or not the recognized character has both uppercase and lowercase letters. Here, "Yo",
Three of "e" and "tsu" correspond. Since "Yo" has a longer vertical length than horizontal, (a / M) squared, and since "E" and "Ts" have a long horizontal length, (b / N) squared is calculated as shown in Fig. 4. become. On the other hand, since the character size dictionary area 30 stores uppercase letters, lowercase letters, and predetermined values as shown in FIG. 5, if the value of the input character is larger than the value of the corresponding character, it is determined to be uppercase, and if smaller, it is determined to be lowercase. From this, it is determined that "yo" is lowercase, "e" is uppercase, and "tsu" is lowercase, and a recognition character as shown in the bottom row of FIG. 4 is obtained. Here, the correction unit 6 corrects “yo”, “e”, and “tsu” assuming that the character size of the recognized character is incorrect.

As described above, according to this embodiment, by providing the maximum side calculation unit, the character size determination unit and the correction unit,
When correcting the error in the character size among the recognized characters output from the character recognition unit, it is possible to correct even double-width characters with high accuracy.

Although the kanji and kana characters have been described in the present embodiment, it is needless to say that the same effect can be obtained for alphabets and the like whose typeface can be determined.

[0018]

As is apparent from the above description, the recognized character correction method of the present invention is a method of correcting a horizontally-oriented character by comparing the character sizes based on the maximum lengths of the vertical and horizontal sides of the recognized character. With regard to the special typefaces and fonts of, it is possible to accurately correct errors related to the character size of the recognized characters.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of a recognized character correction method of the present invention by function.

FIG. 2 is a block diagram of a character recognition device showing the same embodiment as a hardware configuration.

FIG. 3 is a flowchart showing the procedure of the same embodiment.

FIG. 4 is an explanatory view showing a procedure using a specific character string of the same embodiment.

FIG. 5 is a layout view of stored contents stored in a character size dictionary area of the same embodiment.

FIG. 6 is a character pattern diagram for explaining a conventional recognized character correction method.

[Explanation of symbols]

 1 character input part 2 character cutout part 3 character recognition part 4 character size judgment part 5 correction part 6 display part 7 maximum side calculation part

Claims (2)

[Claims] 1. A document pattern to be recognized is read, characters are cut out from the document pattern, and a first maximum value and a horizontal size that are the maximum vertical size of the character pattern in the character string including the cut out characters. The second maximum value that is the maximum of is calculated, the character code is read from each character pattern, the possibility that the character code has both uppercase and lowercase letters is determined, and if it has the possibility, the character pattern is determined. The vertical length and the horizontal length are compared, and if the vertical length is long, the upper and lower case letters are determined by the vertical length and the value calculated from the first maximum value, and the horizontal length is long. In this case, the uppercase and lowercase letters are determined based on the horizontal length and the value calculated from the second maximum value, and if different from the character code, the character code is corrected. 2. The method for correcting a recognized character according to claim 1, wherein when deciding the uppercase letter and the lowercase letter, it is decided whether it is the uppercase letter or the lowercase letter based on the area ratio with the character pattern having the first or second maximum value.