JPH04138583A - Character recognizing device - Google Patents

Abstract

PURPOSE:To easily and highly accurately correct an erroneously recognized character by executing recognition processing by means of the 2nd character recognition part constituted of a detailed character feature extracting part to be used for a specific case, a reference character detailed feature storing part and a detailed character recognition processing part for executing character recognition processing. CONSTITUTION:Respective characters in a word specified by a word specification part 8 are recognized by the detailed character recognition processing part 12. In this processing, a dictionary stored in the reference character detailed feature storing part 11 is collated with data extracted by the detailed character feature extracting part 13 to converge the candidates of recognized results. The combinations of respective candidates of recognized results are furthermore converged by referring to a word dictionary 10 and the final recognized result is displayed on a recognized result display part 7. When there are two or more finally recognized results, respective candidates are successively displayed on the display part 7 and one of them is selected. Or, all the candidates are displayed and a required one is selected. Consequently, the simple and highly accurate correction of an erroneously recognized character can be executed only by specifying the word including the erroneously recognized character without reinputting the word.

Description

DETAILED DESCRIPTION OF THE INVENTION (A) Field of Industrial Application The present invention relates to a character recognition device having a function of correcting erroneously recognized characters.

(B) Conventional technology When a recognition result is incorrect in a conventional character recognition device, the correction method requires rewriting the characters in the part to be corrected and performing the recognition process again. The method used was to display candidate characters for character recognition and specify the character to be corrected.

This method requires rewriting, which forces the operator to perform complicated operations, and there is a high probability that characters that have been misrecognized due to the person's writing habits will still be misrecognized even if they are rewritten and rerecognized. This method has the disadvantage that correct recognition results are not always obtained.

In addition, some devices use a misrecognized character correction method in which characters are input from a keyboard through kana-kanji conversion processing. This requires the character recognition device to be equipped with a keyboard, which hinders miniaturization and cost reduction of the device. Furthermore, character input using kana-kanji conversion has the disadvantage that it requires skill to operate.

(c) Problems to be Solved by the Invention The present invention has been made in view of the above circumstances, and when a misrecognized character occurs, it can be input by rewriting, inputting by converting kana-kanji from the keyboard, or inputting the next candidate character. A technical problem is to easily and accurately correct misrecognized characters by eliminating the need for complicated operations such as specifying .

(d) Means for Solving the Problems In order to solve the above problems, the present invention provides a character pattern reading section that reads character patterns, and a character pattern storage section that stores character pattern data read from the character pattern reading section. a character recognition unit that recognizes the character pattern data, a display unit that displays the recognition result by the character recognition unit, and a word designation unit that designates a part of the word of the recognition result displayed on the display unit; If the character recognition result is incorrect, the word containing the error is specified by the word specification section, and each character in the specified range is recognized as a candidate by further detailed recognition processing. After selecting, the specified word is selected by referring to the word dictionary.

The character recognition unit also includes a character feature extraction unit that extracts character features of a character pattern, a standard character feature storage unit that stores character feature amounts of a standard pattern, and feature amounts obtained from the character feature extraction unit and the a first character recognition unit comprising a character recognition processing unit that collates feature quantities in a standard character feature storage unit and performs character recognition processing; a detailed character feature extraction unit that extracts character features of a character pattern in detail; a standard character detailed feature storage unit that stores character feature amounts of the standard pattern in detail;
A second character recognition section includes a detailed character recognition processing section that performs character recognition processing by comparing the feature amounts obtained from the detailed character feature extraction section with the feature amounts in the standard character detailed feature storage section; The present invention is characterized in that when recognizing a word specified by the word specifying section, the second character recognition section performs recognition processing.

(E) Effect According to the present invention, when a misrecognized character occurs in the recognition result of the character recognition device, a word containing the misrecognized character is designated and detailed recognition processing is performed, and then a word dictionary is used to perform recognition processing. By correcting the above recognition results, characters can be corrected with a simple operation and with high precision.

(F) Example A character recognition device according to the present invention will be described in detail below with reference to FIGS. 1 to 9.

First, an outline of the character recognition device of the present invention will be explained.

FIG. 1 is a block diagram showing the configuration of a character recognition device according to the present invention. In the figure, (1) is a character data reading section, which is composed of a handwritten character input tablet or an optical character reading device. (2) is a character pattern storage unit that stores data input from the character data reading unit (1).

(3) is a standard character feature storage section, which stores a reference dictionary in which character features necessary for character recognition processing are registered in advance. (4) is a character recognition processing section, which performs recognition of input data by referring to data in the standard character feature storage section (3). (5) is a character feature extraction unit which extracts character features corresponding to the reference dictionary in the standard character feature storage unit (3) from the data stored in the character pattern storage unit (2). (6) is a recognition candidate storage section, which is used to store recognition candidates obtained as a result of character recognition. (7
) is a recognition result display section, which consists of an LCD display or a CRT display. (8) is a word specification section, which specifies a part of the character string displayed on the recognition result display section (7).

(9) is a language processing unit for post-recognition processing, which determines the recognition result based on a combination of data in the recognition candidate storage unit (6). (10) is a word dictionary, which is used for reference in the language processing section (9) for post-recognition processing. (11) is a standard character detailed feature storage section, which is a reference dictionary in which character features necessary for character recognition processing are registered in advance, but it contains more information than the standard character feature storage section (3). Information indicating characteristics is stored. (12) is a detailed character recognition processing section, which recognizes input data by referring to the data in the standard character detailed storage section (11). (13) is a detailed character feature extraction unit, which extracts character features corresponding to the reference dictionary in the standard detailed character feature storage unit (11) from the data stored in the character pattern storage unit (2). (14) is a control unit that controls the processing of each of the above units.

An example of a case where the input character pattern (15) shown in FIG. 2 is input from the character data reading section (1) and this character pattern is recognized will be described with reference to the flowchart shown in FIG. 3.

At step 510, the character pattern input from the character data reading section (1) is stored in the character pattern storage section (2) at step S20. Thereafter, in order to perform character recognition processing, the character pattern stored in the character pattern storage unit (2) is sent to the character feature extraction unit (5) in step S30.
Feature extraction is performed by In step S40, the character recognition processing unit (4) performs recognition processing by comparing the character features with the standard pattern in the standard character feature storage unit (3). The recognized result is displayed on the recognition result display section (7) in step S50. In step S60, the operator judges whether or not the data displayed on the recognition result display section (7) is correct.If there is an error, 8
In step 70, a word containing an error is specified by operating the word specifying section (8). Next, in step S90, the detailed character recognition processing unit (12) recognizes each character of the specified word.
). In this process, the dictionary stored in the standard detailed character feature storage section (11) is compared with the data extracted by the detailed character feature extraction section (13) in step 880 to narrow down candidates for recognition results. Furthermore, in step 5110, each candidate combination of recognition results is added to a word dictionary (8
), the candidates are further narrowed down and displayed again on the recognition result display section (6) as the final recognition result. If there are finally two or more recognition result candidates in step 8150, select by sequentially displaying them in the recognition result display area in step 5130, or display all candidates and make a selection. . Figure 4 shows the input character pattern (15
) for the recognition result (16) by the character recognition processing unit (4)
). Figure 5 shows the recognition results (16)
FIG. 17 is a diagram showing an example (17) in which a word including a misrecognized character is specified. FIG. 6 shows recognition candidates (18) obtained as a result of detailed character recognition processing, which are stored in the recognition candidate storage section (6).

The character recognition processing unit (4) of the above-mentioned device performs recognition processing on the input character pattern (15), and the recognition result (16
) is displayed as an example, the recognition processing of the present invention will be specifically explained.

In this example, as a result of recognizing the input character string "It's sunny today", "sky" was incorrectly recognized as "o".

Standard character feature storage unit used for recognition processing at this time (
3) The standard character pattern is 4x as shown in Figure 8.
Data of a plurality of characters is stored, which is divided into four meshes and uses the blackness of each region as a feature quantity. Character feature extraction part (
In step 5), the data stored in the character pattern storage unit (2) is normalized and the blackness of each area is extracted as a feature quantity into a 4×4 mesh. In the recognition processing section (4),
The above-mentioned two feature quantities are compared to calculate the degree of similarity, and the one with the highest degree of similarity is displayed as a recognition result on the recognition result display section (7). Here, the result was that the degree of similarity to "o" was the highest.

The word "Haruo" including "Ten" is specified by the word specifying section (8). (17) is specified by the word specification section (8) as “Haruo J
This is an example of the display when specified. The character specified by the word specifying section (8) is subjected to character recognition processing using a detailed character recognition processing section (12) in more detail than the processing performed by the character recognition processing section (4). The standard character detailed feature storage unit (11) used for this processing has a dictionary that divides the standard character pattern into 8x8 meshes and uses the blackness of each area as a feature, as shown in Figure 9. is stored. In addition, the detailed character feature storage section (13) also stores a character pattern storage section (
After normalizing the data in 2), it is divided into 8×8 meshes, and the blackness of each mesh is extracted as a feature quantity. Further, the detailed character recognition processing section (12) compares both of the above feature quantities and selects a plurality of candidates in order of similarity. This detailed character recognition processing unit (12
) are stored in the recognition candidate character storage section (6). (18) is a specific example of a candidate character group obtained as a result of recognition by the detailed character recognition processing section (12). ``Heaven'' is recognized in the third place. The post-recognition processing language processing unit (9) determines whether a word in the word dictionary (10) can be composed of a combination (18) of recognition candidate characters based on the word dictionary (10). If a word can be formed by a combination of recognition candidate characters, the characters forming the word as a recognition result are used to correct misrecognition. In addition, if multiple words can be composed of a combination of recognition candidates, detailed character recognition processing (
Characters with high character recognition similarity in step 12) are given priority and displayed on the recognition result display section (7). As a result, the final result (19) shown in FIG. 7 can be obtained.

(g) Effects of the Invention As described above, according to the present invention, misrecognized characters can be easily and accurately corrected by simply specifying the word that includes the misrecognized characters without the need for re-input. This makes it possible to speed up the process, shorten the operator's work time, and reduce fatigue and stress.

Furthermore, since an input means such as a keyboard is not required, the device can be made smaller and lower in price.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of the present invention, FIG. 2 is a diagram of an input character pattern, FIG. 3 is a flowchart showing the control flow of the present invention, FIG. 4 is a diagram of recognition results, and FIG. 5 is a diagram of the input character pattern. Figure 6 is a conceptual diagram of the recognition candidate storage unit showing the recognition candidate characters obtained as a result of detailed character recognition processing, and Figure 7 is the final recognition result. The results are shown in Figure 8, which shows the character pattern divided into 4x4 meshes to obtain the data in the standard character feature storage, and Figure 9, which shows the character pattern divided into 4x4 meshes to obtain the data in the standard character detailed feature storage. It is a diagram in which the pattern is divided into 8×8 meshes. 1...Character data reading unit, 2...Character pattern storage unit, 3...Standard character feature storage unit, 4...Character recognition processing unit, 5...Character feature extraction unit, 6...Recognition Candidate storage section, 7... Recognition result display section, 8... Misrecognition Oribata incorrect specification section, 9... Language processing section for post-recognition processing, 10... Word dictionary, 11... Standard character details Feature memory section, 12...
・Detailed character recognition unit, 13...Character detailed feature extraction unit, 1
4...Control unit.

Claims (2)

[Claims] (1) A character pattern reading section that reads a character pattern, a character pattern storage section that stores the character pattern data read from the character pattern reading section, a character recognition section that recognizes the character pattern data, and the character recognition section. a display section for displaying a recognition result, a word specification section for specifying a part of the word of the recognition result displayed on the display section, and a word dictionary, and when the character recognition result is incorrect, it can detect the error. After specifying the words to be included in the word specification section, selecting recognition candidates by performing further detailed recognition processing on each character in the specified range, and selecting the specified word by referring to the word dictionary. Characteristic character recognition device. (2) The character recognition section includes a character feature extraction section that extracts the character features of the character pattern, a standard character feature storage section that stores the character features of the standard pattern, and a feature amount obtained from the character feature extraction section. a first character recognition unit comprising a character recognition processing unit that collates the feature amounts in the standard character feature storage unit and performs character recognition processing; a detailed character feature extraction unit that extracts detailed character features of the character pattern; , a standard character detailed feature storage section that stores character feature amounts of the standard pattern in detail, and a feature amount obtained from the detailed character feature extraction section and the feature amount in the standard character detailed feature storage section to perform character recognition processing. The second character recognition unit is comprised of a detailed character recognition processing unit that performs a detailed character recognition processing unit, and when recognizing a word specified by the word specification unit, the second character recognition unit performs recognition processing. A character recognition device according to claim 1.