JP2001266074A - Device for recognizing character - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent increase in load on an operator, even when multiple similar words are comprised in a word dictionary. SOLUTION: A character string 13 on a slip is read by a picture input part 1 and segmented by a character segmenting part 2. A character-recognizing part 3 recognizes the character string 13 and outputs a character recognition result candidate 9. A word-collating part 4 inspects the matching rate 12 of the candidate 9, with each word in the word dictionary 6 and the word with the high matching rate 12 is outputted as a word recognition result candidate 10 which is the recognition result candidate of the character string 13. When the plurality of word recognition result candidates 10 exist and the similar word recognition result candidates 10 exist among them, a similar word determining part 5 performs selection from the similar candidates 10 to obtain a word recognition result 11, which is the recognition result of the character string 13.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device, and more particularly to a character recognition device for recognizing a character string.

[0002]

2. Description of the Related Art Heretofore, this type of character recognition apparatus has been used for reading words written on a form.

Referring to FIG. 10 which is a block diagram of this conventional character recognition device, the conventional character recognition device includes an image input section for reading a form and outputting image data, and a character cutout section for cutting out a character string from the image data. And a character recognition unit that recognizes the characters of this character string for each character by the number of characters in the character string and outputs respective character recognition result candidates corresponding to the recognized characters, and a character recognition result corresponding to the character string. Each word in the word dictionary in which candidates and a plurality of words to be entered in the form are stored in advance is checked for the degree of matching, and words with a high degree of matching are output as word recognition results, which are character string recognition results. And a word collating unit that performs Then, the word matching unit outputs a word having a high degree of matching as a word recognition result of a character string. When a word dictionary including similar words is used, a plurality of words having a high degree of matching are obtained, and as a result of the word recognition. In many cases, the similar words in the word dictionary are output. At this time, the operator refers to the image data read from the form and selects a word corresponding to the image data to obtain a word recognition result.

[0004]

In the above-mentioned conventional character recognition apparatus, a word matching section outputs a word having a high degree of coincidence as a word recognition result. A plurality of words with high degrees are obtained and the similar words are output as word recognition results, and the operator refers to the image data read from the form and outputs a word corresponding to the image data by the word matching unit. Since a word is selected from the words and the word recognition result is used, there is a problem that the load on the operator increases when the word dictionary contains many similar words.

An object of the present invention is to provide a character recognition apparatus which does not increase the burden on an operator even when a word dictionary contains many similar words in order to eliminate such conventional disadvantages.

[0006]

According to a first character recognition device of the present invention, when a character string is read and recognized, there are a plurality of recognition result candidates for the character string, and a similar one among the plurality of recognition result candidates. When the recognition result candidate is present, the recognition result candidate is selected from the similar recognition result candidates and used as the character string recognition result.

According to a second character recognition device of the present invention, when a character string indicating a word is read and recognized, there are a plurality of recognition result candidates for the character string, and the recognition result similar to the plurality of recognition result candidates. When there is a candidate, the candidate is selected from the similar recognition result candidates and used as the recognition result of the word.

According to the third character recognition device of the present invention, when a character string indicating a word described on a form is read and recognized, there are a plurality of recognition result candidates for the character string, and among the plurality of recognition result candidates, A recognition result candidate that is similar to the above-mentioned word, and selects from the similar recognition result candidates to obtain the word recognition result.

The recognition results of the first to third character recognition apparatuses of the present invention are used to determine which character is certain for a different character between the similar recognition result candidates. Is performed, and the recognition result candidate having the most likely character among the similar recognition result candidates is selected and obtained.

According to a fourth aspect of the present invention, there is provided an image input unit for reading a form and outputting image data; a character extracting unit for extracting a character string from the image data; A character recognition unit that outputs each character recognition result candidate corresponding to each character recognized and recognized by the number of characters in the character string, and a plurality of the character recognition result candidates output by the character recognition unit and written in the form. Word matching for checking the degree of matching indicating the degree of matching with each word in a word dictionary in which the word is stored in advance, and outputting the word having a high degree of matching as a word recognition result candidate that is a recognition result candidate for the character string When there are a plurality of word recognition result candidates output by the word matching unit and there is a similar word recognition result candidate among the plurality of word recognition result candidates, It is constituted by a similar word determination unit for word recognition result as the recognition result of the character string selected from the similar word recognition result candidates.

Further, the word collating unit of the fourth character recognition device of the present invention uses the words in the word dictionary having the same number of characters as the number of characters in the character string, and corresponds to the characters in the character string. Whether the character recognition result candidate and the character in the word at the same position as the position of the character in the character string are the same or not for all the characters in the character string, and determine the same number. The degree of coincidence is set.

Further, in the fourth character recognition device of the present invention, the similar word determination unit checks which character is certain for a different character between the similar word recognition result candidates. And the word recognition result candidate having the most probable character among the similar word recognition result candidates is set as the word recognition result.

Further, the first to fourth character recognition devices of the present invention may be arranged such that the “probability” of the character is determined by calculating a half value of an average of distance values of similar characters of the character. The value is calculated by adding it to the value.

[0014]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing one embodiment of the character recognition device of the present invention.

In this embodiment shown in FIG. 1, an image input unit 1 for reading a form or the like and outputting image data, a character extracting unit 2 for extracting a character string 13 from the image data, and a character of the character string 13 Each character recognition result candidate 9 corresponding to each character recognized and recognized by the number of characters in the character string 13 is output for each character (a plurality of candidates having a small distance value are usually output as the character recognition result candidates 9. Here, a plurality of candidates are output. The distance value indicates the degree of difference between the recognized character and the character recognition result candidate 9.) The character recognition unit 3, the character recognition result candidate 9 output by the character recognition unit 3, and a plurality of entries written in a form or the like. The degree of coincidence 12 indicating the degree of coincidence with each word in the word dictionary 6 in which the word is stored in advance is checked, and a word having a high degree of coincidence 12 is output as a word recognition result candidate 10 which is a recognition result candidate of the character string 13. A word collating unit 4, there are a plurality word recognition result candidates 10 the word collating unit 4 is output similar word recognition result among the plurality of word recognition result candidates 10 candidates 1
0, the similar word recognition result candidate 1
And a similar word determination unit 5 which selects a word from 0 and makes a word recognition result 11 which is a recognition result of the character string 13.

The word collating unit 4 uses words in the word dictionary 6 having the same number of characters as the number of characters in the character string 13, and a character recognition result candidate 9 corresponding to the character in the character string 13 and a character string 1.
3 is checked for all the characters in the character string 13 to determine whether the position of this character is the same as the character in the word at the same position.
The same number is set as the coincidence degree 12.

The similar word determination unit 5 conducts a "probability check" to determine which character is certain with respect to a different character between the similar word recognition result candidates 10, and determines the similar word recognition result candidate. The word recognition result candidate 10 having the most probable character in 10 is set as the word recognition result 11.

FIG. 1 also shows a similar word table 7 and a similar character table 8 used by the similar word determination section 5.

Next, the operation of the character recognition device according to the present embodiment will be described in detail with reference to FIGS.

FIG. 2 is a flowchart showing an example of the operation of the similar word determining unit.

FIG. 3 is a diagram showing an example of the word dictionary.
This indicates that words described in a form or the like are stored in the word dictionary 6 in advance in the order of word numbers.

FIG. 4 is a diagram showing an example of the similar word table. In the word group of the word dictionary 6, words having the same word length and differing only by one character (similar word 14) are similar words (similar word 14). No.), the number of characters of this word, the number of similar words 14, and the characters that differ therefrom (hereinafter referred to as
(A) indicating the differentiating character 15), and (b) indicating the position of the differentiating character 15 and the similar word No. for each word No. corresponding to the word in the word dictionary 6, and prepared in advance. Keep it.

FIG. 5 is a diagram showing an example of the similar character table. For each of the differentiating characters 15 shown in FIG. The number of characters (the number of similar characters) and the number of similar characters (similar characters) are stored for the characters to be recognized.

FIG. 6 is a diagram showing an example of a character recognition result candidate. When the character recognizing unit 3 recognizes a character written on a document (that is, a character cut out by the cutout unit), the character Are shown in descending order of similarity.

FIG. 7 is a diagram showing an example of the degree of coincidence of words in the word dictionary.

FIG. 8 is a diagram showing an example of the re-recognition result.

FIG. 9 is a diagram showing an example of a calculation process of “probability” and a result.

In FIG. 1, a form or the like is read by an image input unit 1 to output image data, and a character cutout unit 2 cuts out, for example, a character string 13 indicating a word from the image data output by the image input unit 1. The character recognition unit 3 recognizes the characters of the character string 13 for each character by the number of characters in the character string 13, and each character recognition result candidate 9 (one or more) corresponding to each recognized character, and The distance value of each of the character recognition result candidates 9 is output. The word matching unit 4 checks the degree of coincidence 12 indicating the degree of coincidence between the character recognition result candidate 9 output by the character recognition unit 3 corresponding to the character string 13 and each word in the word dictionary 6, and checks the degree of coincidence 12
Is output as a word recognition result candidate 10 which is a recognition result candidate of the character string 13. At this time, a word in the word dictionary 6 having the same number of characters as the number of characters in the character string 13 is used, and the character recognition result candidate 9 corresponding to the character in the character string 13 and the same position as the position of this character in the character string 13 Is checked for all the characters in the character string 13 to determine whether or not the characters in the word are the same.
Let it be 2. The similar word determining unit 5 receives the word recognition result candidate 10 output by the word matching unit 4, and when the word recognition result candidate 10 is one, the word recognition result candidate 10 is the word recognition result which is the recognition result of the character string 13. It is assumed to be 11. Further, when there are a plurality of word recognition result candidates 10 received from the word matching unit 4, when there is no similar word recognition result candidate 10 among the plurality of word recognition result candidates 10,
When the plurality of word recognition result candidates 10 received from the word matching unit 4 are referred to as word recognition results 11 and the similar word recognition result candidate 10 exists in the plurality of word recognition result candidates 10 received from the word matching unit 4. A word recognition result 11 which is a recognition result of the character string 13 is selected from the similar word recognition result candidates 10. That is, a “probability” check is performed to determine which character is certain for a different character between the similar word recognition result candidates 10, and the most likely character in the similar word recognition result candidate 10 is determined. Is set as the word recognition result 11.

Here, referring to FIG.
Will be described in more detail. First, the word recognition result candidate 10 is received from the word collating unit 4 (S1), and the similar word 1 between the word recognition result candidates 10 is included in the word recognition result candidate 10.
4 is checked with reference to the similar word table 7 of FIG. 4 (that is, the similar word table 7 shown in FIG. 4).
(B), the word dictionary 6 of the word recognition result candidate 10 is referred to.
The similar word Nos. Corresponding to the word Nos. Are examined for all the word recognition result candidates 10 received in step S1. For example, when the same similar word No. is obtained from at least two or more word recognition result candidates 10, Is determined to be “similar word 14” based on the similar word No. ) (S2). If there is no similar word 14, the process continues to step S8. When the similar word 14 exists, the similar word 14 corresponding to FIG.
The recognition target when recognizing the differentiating character 15 shown in (a) and the similar character (shown in FIG. 5) of this differentiating character 15 is limited, and the character at the differentiating character position in the character string 13 is changed to a character. The character is recognized again by the recognition unit 3, and as a result of character recognition, FIG.
As shown in the figure, the character recognition result candidate 9 (the above-described differentiated character 1
5 and a character value of the character recognition result candidate 9 are obtained (S3). From this result, for each distinctive character 15, the “probability” of these characters is expressed by “Expression 1” (“Expression 1”: Rc = γc
+ Σγx / (Nc + 2), where x = 1 to Nc, where Rc is “probability” of the distinguishing character 15c, γc is the distance value of the distinguishing character 15c, and γ1 to γNc are similar characters of the distinguishing character 15c. The distance value and Nc are differentiating characters 15c.
Is the number of similar characters. ). At this time,
Not only the differentiating character 15 but also the similar character corresponding to the differentiating character 15 is reflected in the calculation of the “probability”. That is, in “Equation 1”, for the character c,
The value of the average of the distance values of similar characters of the character c is added to the distance value of the character c to calculate the “certainty” of the character c. The degree of “probability” increases as the numerical value of “probability” decreases, similarly to the distance value (S4). As a result of the calculation, the “probability” of one of the differentiating characters 15 is, for example, less than a predetermined value (which can be obtained in advance by an experiment or the like) and the other of the differentiating characters 15 When the value is smaller than a predetermined value (determined in advance by experiments or the like) from “probability”, the distinguishing character 15
Is determined as a determination result, and when the above condition is not satisfied, there is no determination result (S5). Check for the judgment result (S
6) If there is a judgment result, the word recognition result candidate 10 in the similar word 14 checked in step S2 including the segmented character of the judgment result is output as the word recognition result 11 (S
7) If there is no determination result, the word recognition result candidate 10 received from the word matching unit 4 is output (S8).

Further, the operation of the similar word determination unit 5 will be described with reference to FIG. The character string 13 described in the form or the like and cut out by the character cutout unit 2 is assumed to be “Fuchu Nishi Elementary School”, and the character recognition result candidates 9 output by the character recognition unit 3 are as shown in FIG. When word matching is performed by the word matching unit 4 with reference to the character recognition result candidate 9 shown in FIG. 6, the three words "Fuchu Minami Elementary School", "Fuchu Kita Elementary School", and "Fuchu Nishi Elementary School" match. The degree 12 is the highest and becomes the word recognition result candidate 10. That is, each word in the word dictionary 6 shown in FIG. 3 is compared with the character recognition result candidate 9 shown in FIG.
First, if it is checked whether the first character “blue” matches the first characters “fu”, “bottom”, “common”, “hemp” and “eave” of the character recognition result candidate 9, the results match. Not. Similarly, as to the second character, if it is determined whether or not “leaf” matches “medium”, “shin”, “instep”, “width”, and “ya”, the result does not match. When the same is performed on the third and subsequent characters, the word No. is 1
In the case of “Aobadai Elementary School”, in the character recognition result candidates 9 corresponding to the fourth, fifth and sixth characters “small”, “study” and “school”, as shown in FIG. Since “small”, “study” and “school” exist, the number of matching characters in the word “Aobadai Elementary School” is “3”, that is, the degree of matching 12 is 3. Similarly, when the words in all the word dictionaries 6 are collated with the character recognition result candidates 9 shown in FIG.
Since the result is as shown in FIG. 7, three words “Fuchu Minami Elementary School”, “Fuchu Kita Elementary School”, and “Fuchu Nishi Elementary School” having the highest matching degree 12 are output as word recognition result candidates 10.
However, at this point, it cannot be determined which of the three words is most likely. Then, the similar word determination unit 5 receives three words of the word recognition result candidates 10 “Fuchu Minami Elementary School”, “Fuchu Kita Elementary School”, and “Fuchu Nishi Elementary School” output by the word matching unit 4 according to the flow shown in FIG. (S1) It is checked whether or not there is a similar word 14 among these three words. That is, referring to (b) of the similar word table 7 shown in FIG. 4, similar words No. 1 and No. 2 from the word No. 4 of the word recognition result candidate 10 “Fuchu Minami Elementary School”, and the word recognition result candidate 10 “Fuchu North” Since the similar word No. 2 is obtained from the word No. 5 of “elementary school” and the word No. 6 of “Fuchu west elementary school”, for example, the same similar word N is obtained from at least two or more word recognition result candidates 10.
If these word recognition result candidates 10 are determined to be “similar words 14” by the similar word No when o is obtained, it is determined that there is a similar word 14 by the similar word No2 (S2). Next, character type limited re-recognition is performed, that is, using the similar word No2 determined in step S2, the differentiating character 15 corresponding to the similar word No2 in the similar word table 7 shown in FIG. "South",
"North" and "West" and their similar characters (shown in Fig. 5)
"Quote", "ho", "rain", "ratio", "this", "rooster",
The character to be collated when recognizing “both” and “face” is limited, and the character (in this case, “west”) in the character string 13 or the character at the character position (in this case, 3) is reproduced by the character recognition unit 3. Recognized,
As a result, a re-recognition result shown in FIG. 8 is obtained (in this re-recognition, the distance value of the character recognition result candidate 9 is also acquired) (S3). Next, in the calculation of the likelihood of distinguishing characters, from this re-recognition result,
Calculate the certainty of each. The calculation progress and the result are as shown in FIG. 9 (S4). If the determination condition is that “probability” is, for example, less than 100 and smaller than any other value by, for example, 10 or more, and if the distinguished character 15 is the determination result, the “probability” of the differentiated character 15 “west” "Is" 69 ", which satisfies the condition.
5 Word recognition result candidates including "West" 10 "Fuchu Nishi Elementary School"
Is output as the word recognition result 11 (S5, S6, S
7).

[0032]

As described above, according to the character recognition apparatus of the present invention, when a character string (for example, a word) is read and recognized, if there are a plurality of recognition result candidates for the character string, the character When there is a similar recognition result candidate among the recognition result candidates, a selection is made from the similar recognition result candidates to be a recognition result of this character string (that is,
Even if a similar word is included in the word dictionary, this character string is selected and selected from similar recognition result candidates. As the result of column recognition is used, the rate of outputting multiple recognition results is reduced,
When a plurality of recognition results are output, the burden on the operator to select from these is reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one embodiment of a character recognition device of the present invention.

FIG. 2 is a flowchart illustrating an example of an operation of a similar word determination unit.

FIG. 3 is a diagram illustrating an example of a word dictionary.

FIG. 4 is a diagram showing an example of a similar word table.

FIG. 5 is a diagram illustrating an example of a similar character table.

FIG. 6 is a diagram illustrating an example of a character recognition result candidate.

FIG. 7 is a diagram showing an example of the degree of coincidence of words in a word dictionary.

FIG. 8 is a diagram illustrating an example of a re-recognition result.

FIG. 9 is a diagram illustrating an example of a calculation process of “probability” and a result.

FIG. 10 is a block diagram of a conventional character recognition device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image input part 2 Character extraction part 3 Character recognition part 4 Word collation part 5 Similar word judgment part 6 Word dictionary 7 Similar word table 8 Similar character table 9 Character recognition result candidate 10 Word recognition result candidate 11 Word recognition result 12 Matching degree 13 Character string 14 Similar words 15 Differentiating characters

Claims (8)

[Claims] 1. When a character string is read and recognized, there are a plurality of recognition result candidates of the character string, and when there are similar recognition result candidates among the plurality of recognition result candidates, the similar recognition result is obtained. A character recognizing device, wherein a character string is selected from the candidates to be used as the character string recognition result. 2. When a character string indicating a word is read and recognized, there are a plurality of recognition result candidates for the character string, and when there is a similar recognition result candidate among the plurality of recognition result candidates, the similarity is determined. A character recognition apparatus, wherein the selected character is selected from among the recognized recognition result candidates to obtain the word recognition result. 3. When a character string indicating a word written on a form is read and recognized, there are a plurality of recognition result candidates for the character string, and a similar recognition result candidate exists among the plurality of recognition result candidates. A character recognition device that selects from the similar recognition result candidates to make the recognition result of the word. 4. The recognition result is obtained by performing a “probability” check to determine which character is certain with respect to a different character between the similar recognition result candidates. 4. The character recognition device according to claim 1, wherein the recognition result candidate having the most probable character is selected and obtained. 5. An image input unit for reading a form and outputting image data, a character cutout unit for cutting out a character string from the image data, and recognizing and recognizing the characters of the character string for each character by the number of characters in the character string. A character recognition unit that outputs each character recognition result candidate corresponding to each of the characters, and a word dictionary in which the character recognition result candidates output by the character recognition unit and a plurality of words to be written in the form are stored in advance. A word matching unit that checks a degree of matching indicating a degree of matching with each word of the word, and outputs the word having a high degree of matching as a word recognition result candidate that is a recognition result candidate of the character string; When there are a plurality of candidate word recognition results, and the plurality of candidate word recognition results include the similar candidate word recognition result, the similar candidate word recognition result is determined. Character recognition apparatus characterized by comprising a, a similar word determining unit that the word recognition result is a recognition result of the character string by selecting from within. 6. The word matching unit uses the words in the word dictionary having the same number of characters as the number of characters in the character string, and selects the character recognition result candidate corresponding to the character in the character string and the character string. Whether the character in the word at the same position as the character in the word is the same or not for all the characters in the character string, and set the same number as the matching degree. The character recognition device according to claim 5, wherein 7. The similar word recognition unit performs a search of “probability” for checking which character is certain for a different character between the similar word recognition result candidates, and performs the similar word recognition. 7. The character recognition device according to claim 5, wherein the word recognition result candidate having the most probable character in the result candidates is set as the word recognition result. 8. The “probability” for the character
The character recognition device according to claim 4, wherein the value is calculated by adding a half value of an average of distance values of similar characters to the character to the distance value of the character.