JP3548372B2 - Character recognition device - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a character recognition device for recognizing characters written on a form.
In particular, the present invention relates to a form character recognition device in which a character area to be written on a form is specified, and a voiced point and a semi-voiced point are written as one character.
[0002]
[Prior art]
Conventionally, character recognition devices that use half-width characters and full-width characters are different. In addition, the rules for entering characters are different for half-width characters such as kana. It is common to fill in the character box. This is because the EBCDIC code (half-width code) has no kana code with a voiced point and a voiced point.
[0003]
A conventional technique will be described with reference to FIG.
In FIG.
Reference numeral 111 denotes a character such as a handwritten character entered in the form, and shows an example in which Yamada is entered in the form. In the first box, "Y" and in the second box are entered, and in the third box, "D" and "Dakuten" are erroneously entered in the same box. Reference numeral 111 'denotes an example in which a correct entry is made, in which the third box is filled with a letter and the fourth box is filled with a cloud point.
[0004]
Reference numeral 112 denotes the result of character recognition of the entered characters one by one. The character recognition for each character is performed by reading characters written in a form as image data, extracting one character, extracting a characteristic of each character extracted, and comparing the extracted character with a character recognition dictionary 120 having a characteristic for each character. , Character recognition is performed for each character. Reference numeral 112 denotes the result obtained in such a manner, and indicates that the character in the first frame is recognized as "Y", the character in the second frame is recognized as "", and the character in the third frame is recognized as "1". The character recognition dictionary has the characteristics of characters with voiced and semi-voiced dots, and can recognize them as one character. For example, if the character and the voiced dot “゛” are entered in the same box, one character Character recognition).
[0005]
Reference numeral 113 denotes a knowledge process for searching for a word representing the meaning of a character string whose character has been recognized by referring to the knowledge dictionary 114 for the character recognition result 112 for each character. In the case of the example of FIG. 12, since the input area of the input characters 111 is the input area of the first and last names, it is determined that the meaning of the character recognition result 112 is for the first and last names, and the knowledge dictionary 114 for the first and last names is referred to. , Recognize the first and last names of the recognized character string.
[0006]
Reference numeral 114 denotes a knowledge dictionary, which holds words representing meanings for each abstraction of meanings represented by character strings. FIG. 12 shows a knowledge dictionary which holds first and last names for each number of characters. , Etc., indicating that the name has Yamata @, etc. as a four-character first and last name.) For a half-width character, the knowledge dictionary regards, for example, “ta” and “濁” as one character, and looks at two characters as “゛”. There is no single character with a dot.
[0007]
An output process 117 outputs a recognition result of the knowledge process.
Reference numeral 120 denotes a character recognition dictionary, which has a characteristic for each character.
Reference numeral 121 indicates that each character has a characteristic.
[0008]
In the case of FIG. 12, since the character and the cloud point are entered in the same box, the character recognition device determines the entered character as a character with a single character and a character recognition result of ya, ma, and da. Is determined to be three characters. Therefore, the character recognition device refers to the first and last name knowledge dictionary 114 for the three-character Yamada since the entered area is an area for the first and last names. Then, the three-letter words such as Yamano and Yamata in the knowledge dictionary are compared with the character recognition result 112. However, since there is no Yamada in the three-letter name dictionary, a correct result (Yamada) cannot be output.
[0009]
[Problems to be solved by the invention]
When a general user fills out a form in which a single character entry area is defined by a frame or the like, it is natural to write the voiced and semi-voiced dots and the characters related thereto in the same frame. Therefore, the rules of the characters specified by the character recognition device are often not observed. Since the erroneously entered characters are not in the words in the knowledge dictionary, the knowledge processing is completely ineffective, and a correct character recognition result cannot be output.
[0010]
Further, even if the information is correctly entered as indicated by 111 'in FIG. 12, in the half-width display, the character "ta" and the turbid point "@" are treated as different characters and displayed as four characters such as "ya", "ma", "ta" and "@". Therefore, there is a problem that the operation is troublesome when the cursor operation is performed to correct the displayed recognition result.
[0011]
SUMMARY OF THE INVENTION It is an object of the present invention to provide a character recognition device that can correctly perform knowledge processing even when a character having a clouded point or a semi-voiced mark is written in the same frame. It is another object of the present invention to provide a character recognition device capable of efficiently performing an operation even when correcting a recognition result.
[0012]
[Means for Solving the Problems]
The present invention provides a character reading unit that reads characters of a form, a character number determining unit that determines the number of characters in a character string read by the character reading unit, a knowledge dictionary that treats each of voiced and semi-voiced characters as one character, and The first word is stored in the first table with the number of characters of the read character string determined by the character number determination unit, and the number of characters of the read character string is calculated by adding 1 to the number of characters of the read character string. A knowledge dictionary reference unit that stores words having a number of characters up to the doubled number of characters in a second table; and a reading unit that has determined that the character number determination unit has determined that voiced or semi-voiced dots are to be integrated with reference to the second table. An integration unit that extracts words having the number of characters of the extracted character string and stores words in which voiced or semi-voiced voices are integrated in the first table, and a word stored in the first table. Compared to the character recognition result for the word, and so provided most things distances close in, or order and a display unit that displays output with a candidate ranked close in distance.
[0013]
FIG. 1 shows the basic configuration of the present invention.
In FIG.
Reference numeral 1 denotes a character recognition unit, which is a character written on a form or the like in which a character writing area is defined, for example, a character recognition result for each character (the number of characters in the character recognition result is n).
[0014]
Reference numeral 2 denotes a knowledge processing unit which performs knowledge processing for determining the meaning of the character recognition result of the character recognition unit 1 with reference to the knowledge dictionary 10 and recognizes the meaning of the character string of the character recognition result for each character. Things.
[0015]
3 is a process for determining a concept (first name, last name, address, etc.) representing the meaning represented by the character recognition result recognized by the character recognition unit 1.
Reference numeral 4 denotes processing for determining the number of characters in the character string as a result of the character recognition by the character recognition unit 1.
Reference numeral 5 denotes a process of searching and acquiring a word of n characters from the knowledge dictionary 10 having a word representing the meaning of the character string.
[0016]
6 searches the knowledge dictionary 10 having a word representing the meaning of the character string for a word of a character string of (n + 1) characters, a word of a character string of (n + 2) characters,..., A word of n characters of 2n characters This is the process of acquiring.
[0017]
Reference numeral 7 denotes a process for integrating the voiced and semi-voiced dots, which is a process for integrating the voiced and semi-voiced characters and the characters to which they are based into one character. For example, this is a process in which the four characters Yamata タ obtained from the knowledge dictionary and the voiced point “゛” are integrated into one character, and are regarded as da to become three characters.
[0018]
Reference numeral 8 denotes a process of comparing the recognized character string recognized by the character, the n characters obtained from the knowledge dictionary, and the character string obtained by integrating the n characters among (n + 1) to 2n characters.
[0019]
10 is a knowledge dictionary.
An output unit 11 outputs a recognition result having a meaning obtained as a result of the knowledge processing.
[0020]
FIG. 2 is a diagram illustrating the operation of the basic configuration of the present invention.
In FIG.
Reference numeral 20 denotes a Yamada character obtained by character-recognizing a character string Yamada in which a letter and a cloud point “点” are erroneously entered in the same frame on a form one by one (the letter and the cloud point “゛” are recognized as one character). Indicating that).
[0021]
In the knowledge processing unit 2,
Reference numeral 27 denotes an integration unit, which indicates that the four-character Yamata $ 22 of the knowledge dictionary 10 for the first and last names is integrated into a three-character Yamada 28.
[0022]
In the knowledge dictionary 10,
Reference numeral 22 denotes a four-character first and last name Yamata.
Reference numeral 23 denotes a three-character first and last name Yamaki.
[0023]
In the output unit 8,
Reference numeral 31 denotes Yamada of the recognition result obtained by the knowledge processing.
The operation of the basic configuration of the present invention in FIG. 1 will be described with reference to an example in which the input character (character recognition result) is Yamada (see FIG. 2).
[0024]
It is assumed that Yamada 20 is obtained as a result of character recognition for each character, and that the abstract concept of the meaning represented by the character string in the entry area represents the first and last names. Since the character recognition result is for the first and last names and the number of characters in the character string is three, the knowledge processing unit 2 searches for and obtains three-letter words such as first and last names using the first and last name knowledge dictionary 10. .
[0025]
Further, since the character recognition result is three characters, words of a character string from four characters to six characters twice as large are acquired from the knowledge dictionary 10. For example, it is Yamata $ 22 of four characters.
[0026]
For a word having a character string of 4 to 6 characters, the integration unit 27 integrates the voiced and semi-voiced characters. Combined into a three-character Yamada. The knowledge processing unit 2 compares the character recognition result Yamada 20 for each character with three characters (Yamaki or the like) obtained from the knowledge dictionary 10 and a word (Yamada or the like) integrated into three characters. Then, for example, the similarity is examined by distance calculation or the like, and the most similar word (the closest one, etc.), in the case of FIG. 2, Yamada 31 is output to the output unit 11 as a recognition result based on the knowledge processing.
[0027]
According to the present invention, it is possible to obtain a correct recognition result in the knowledge processing even when a voiced point or a semi-voiced point and characters related thereto are written in the same frame. Further, also when displaying the recognition result, processing is performed so that the voiced and semi-voiced dots are integrated into the displayed character and can be handled as one character (the details of this point will be described later).
[0028]
BEST MODE FOR CARRYING OUT THE INVENTION
FIG. 3 is a block diagram of the character recognition device of the present invention.
In FIG.
Reference numeral 40 denotes a form, in which a character entry area and a meaning indicating the content are defined in the entry area (for example, entry areas such as first and last name, first name, and address are defined).
[0029]
Reference numeral 41 denotes a character reading unit which reads characters such as handwritten characters written on a form by a scanner and performs photoelectric conversion.
Reference numeral 42 denotes a feature extracting unit for extracting one character based on the image data of the form obtained by the character reading unit 41 and extracting a feature required for character recognition for each character.
[0030]
A collating unit 43 compares the feature obtained by the feature extracting unit 42 with the feature stored in the dictionary of the character recognition dictionary 44, and performs character recognition for each character.
Reference numeral 44 denotes a character recognition dictionary for storing the characteristics of each character.
[0031]
A knowledge processing unit 45 determines an abstract concept (first name, last name, address, etc.) of the meaning represented by the character from the character recognition result of each character, and generates a knowledge dictionary 46 having a word representing the meaning of the character string. By referring to this, the meaning of the character string of the character recognition result is recognized.
[0032]
Reference numeral 46 denotes a knowledge dictionary which holds, for each abstract concept of a meaning represented by a word, for example, an address, first and last name, name, company name, general name (school, library, etc.), words included in the abstract concept. .
[0033]
Reference numeral 47 denotes a display unit, such as a display for outputting a recognition result of the knowledge processing unit 45.
Reference numeral 48 denotes a correction unit for confirming and correcting the recognition result of the knowledge processing.
[0034]
Reference numeral 49 denotes an input unit for inputting user confirmation, correction data, and the like.
The operation of the configuration of FIG. 3 will be described.
[0035]
The character reading unit 41 reads a form as an image. The feature extracting unit 42 cuts out characters for each character based on the image data and extracts features. Then, the matching unit 43 refers to the character recognition dictionary 44, compares the characteristics of each character, and obtains similar characters. Then, the similar characters are ranked according to the degree of similarity, and given to the knowledge processing unit 45.
[0036]
The knowledge processing unit 45 determines the meaning of the character string (attribute such as address, name, company name, etc.) according to the character recognition result obtained by the matching unit 43. Then, the knowledge dictionary 46 having a word representing the meaning is referred to. At this time, words having a clouded point, a semi-voiced point, etc. among the words included in the knowledge dictionary 46 are integrated to reduce the number of words, and the words of the character string of the character recognition result are considered in consideration of the candidate order of the character recognition result. (Knowledge processing of the present invention will be described in detail with reference to FIG. 6). Then, the obtained recognition result is transferred to the display unit 47 (recognition results to be output are only those having the first recognition result obtained by the knowledge processing, or a plurality of recognition results are output according to the need. Selected).
[0037]
The display unit 47 displays the recognition results obtained by the knowledge processing unit 45 on the screen in the order of candidates.
The user uses the input unit 49 to check the recognition result or input correction data according to the recognition result displayed on the display unit 47. The correction unit 48 performs processing such as confirmation and correction of the recognition result based on data such as confirmation and correction.
[0038]
FIG. 4 shows an embodiment of the knowledge dictionary according to the present invention, which shows a dictionary holding words representing first and last names in katakana.
Reference numeral 51 denotes an area having a character number n of words registered in the dictionary and a pointer to a word storage area.
[0039]
An area 52 holds the first and last names for each number of characters.
There are no registered words (first and last names) of n = 1, registered words of n = 2 are back, registered words of n = 3 are Sasaki, Imai, etc., and n = 4 are Yamata ゛, Aoshima, Takahashi, etc.
[0040]
FIG. 5 is a block diagram of an embodiment of the knowledge processing unit of the present invention.
In FIG.
In the knowledge processing unit 2,
Reference numeral 61 denotes a character number determination unit which receives a character recognition result for each character and determines the character number n of the character string.
[0041]
Reference numeral 61 'denotes an abstract concept determination unit for the meaning represented by the character string, which determines the abstract concept (address, first and last name, name, company name, general name, etc.) of the meaning represented by the input character string.
[0042]
Reference numeral 62 denotes a knowledge dictionary reference unit which, according to the number of characters in the character string as a result of the character recognition and the abstraction of the meaning, retrieves n characters of the knowledge dictionary 10 having the word of the abstraction and words of (n + 1) to 2n characters. Search and ask.
[0043]
Reference numeral 63 denotes a table 1, which holds words of n characters obtained by searching the knowledge dictionary 10 and words having a cloud point and a semi-voice point integrated into n characters by the integration unit 65.
[0044]
Reference numeral 64 denotes a table 2, which holds words of (n + 1) characters to 2n characters obtained by searching the knowledge dictionary 10.
Numeral 65 denotes an integration unit, of the words having (n + 1) to 2n characters held in the table 2 (64), for a word having a voiced point and a voiced voice, the voiced voice and the voiced voice and a character to which the voiced voice is applied Is performed.
[0045]
Reference numeral 66 denotes a similarity determination unit that integrates the character recognition result for each character with the n-character word and the voiced and semi-voiced characters stored in the table 1 (63) to determine the similarity between the word and the word determined to be n characters. Is what you do. The similarity is determined, for example, by calculating the distance between the character of the character recognition result for each character and the character stored in the table 1 (63) (this point will be specifically described with reference to FIG. 6). .
[0046]
In the knowledge dictionary 10,
Reference numeral 71 denotes a first name and last name dictionary that holds words representing the last name of each name for each number of characters.
[0047]
Reference numeral 72 denotes a name dictionary which holds words representing names of names for each number of characters.
An address dictionary 73 holds words such as place names representing addresses for each number of characters.
[0048]
74 is a company name dictionary.
75 is a general name dictionary.
FIG. 6 is an explanatory diagram of character recognition by the knowledge processing of the present invention. An example is shown in which the first candidate of the character recognition result for each character is Yamada, the second candidate is Inugu, and the third candidate is aster. FIG. 6A shows the case where the distance is calculated by comparing the first and last name Yamada extracted from the table 1 (63) with the character recognition result, and FIG. The distance calculation method in the case of Yamada is shown.
[0049]
In FIG. 6A,
63 is a table 1.
Reference numeral 64 denotes a table 2.
[0050]
Reference numeral 65 denotes an integration unit.
66 is a similarity determination unit.
Reference numeral 71 denotes a dictionary representing n first and last names of the knowledge dictionary.
[0051]
Reference numeral 71 'denotes a dictionary representing the first and last names of (n + 1) to 2n characters in the knowledge dictionary.
Reference numeral 82 denotes a distance obtained by comparing the character string extracted from the table 1 (63) with the character string of the character recognition result 91 (a method of calculating the distance will be described later).
[0052]
Reference numeral 91 denotes a character recognition result for each character. The example shows that the first candidate is Yamada, the second candidate is Inu, and the third candidate is aster.
The case of FIG. 6A will be described.
[0053]
A method of acquiring words to be extracted from the knowledge dictionaries 71 and 71 'to the table 1 (63) will be described.
Since the character recognition result 91 is three characters, all three-character words having the first character of the first candidate of the character recognition result 91 as the first character are extracted and stored in the table 1 (63). Next, all three-character words having the first character A of the second candidate of the character recognition result 91 as the first character are extracted and stored in the table 1 (63). Further, all three-character words having the first character a of the third candidate of the character recognition result 91 as the first character are extracted and stored in the table 1 (63).
[0054]
Next, since the character recognition result 91 is three characters, all words having four, five, and six character strings are similarly extracted and stored in the table 2 (64). For example, in the case of four characters, all four-character words whose first character is Y, four characters whose first character is A, and four characters whose first character is A are extracted, and Table 2 (64) ). This operation is repeated for words of 5 or 6 characters, and stored in Table 2 (64).
[0055]
Next, the integrating unit 65 integrates the voiced and semi-voiced words for the words stored in the table 2 (64). Then, the word (Yamada or the like) which has become three characters in the integration processing is moved to the table 1 (63).
[0056]
The similarity judging unit 66 compares the three-character words stored in the table 1 (63) and the words that have been combined into three characters with the character recognition result to determine the distance.
[0057]
It is assumed that the first and last names of three characters Yamada are extracted from the table 1 (63). The similarity determination unit 66 is stored in the table 1 (63). Is The first character of the character recognition result that matches the first character is searched. As a result, the distance between the first character and the first candidate is set to zero. Next, the second character of the character recognition result that matches the second character of the Amada 81 is obtained. As a result, since the value matches the first candidate, the distance of the second character is set to zero. Next, the third character of the character recognition result that matches the third character of Yamada 81 is obtained. As a result, the distance coincides with the first candidate, so the distance is set to 0. Then, the sum 0 of the distances of the characters of Yamada 81 is obtained as the total distance.
[0058]
As shown in FIG. 6B, the distance when Amada is taken out to the table 1 (63) is obtained as follows.
First, the words obtained from Table 1 (63) are compared with the first characters of the character recognition result. Since the first character A of Amada matches the first character A of the third candidate, the distance is set to 2. Next, the second characters are compared. Since the character of the second character of Amada matches the first candidate, the distance is set to zero. Next, the third characters are compared. As a result, since the third character of Amada matches the first candidate, the distance is set to 2. Therefore, the total distance of Amada is 2.
[0059]
In this way, the distances are calculated for all the words stored in the table 1 (63), the one with the smallest total distance is output as the first candidate of the recognition result of the knowledge processing, and the candidates are ranked in ascending order of distance. Output. Only the first candidate may be output as a result.
[0060]
The user sees and confirms the recognition result of the knowledge processing displayed on the display of the output unit, and if necessary, inputs necessary correction data to correct the recognition result.
FIG. 7 is a flowchart of an embodiment of the knowledge processing unit of the present invention.
[0061]
The operation of the configuration of FIG. 5 will be described according to the illustrated step numbers (see FIG. 5).
S1 The character number determination unit 61 sets the candidate number koho (koho = 3 in FIG. 6).
[0062]
S2 The character number determination section 61 sets the input character number n (n = 3 in FIG. 6). S3, S4 Loop from i = 1 to koho. In the loop, assuming that the first i-th candidate character of the input character is S, all words of length n starting with S are taken out of the dictionary and put into table 1 (table1). In the case of FIG. 6, since the first character of the first candidate Yamada of the character recognition result is "YA", all three-letter words (first and last names) starting from "YA" are extracted. Next, since the first character of the second candidate imada is a, all three-letter words starting with a are extracted. Similarly, since the first character of the third candidate aster is "A", all three words starting with "A" are searched for and found.
[0063]
S5, S6 Loop from k = n + 1 to k = 2n. Further, in that loop, the process loops from i = 1 to i = koho.
In the process of S7 loop, when the i-th candidate character of the first character of the input (character recognition result) is S, all words having a length k starting with S are extracted from the dictionary and stored in the table 2. For example, since n = 3, for a four-character word, a four-character word in which the first character of the first candidate of the character recognition result is the first character is stored in the table 2. Next, a four-character word having the first character of the second candidate of the character recognition result as the first character is stored in the table 2. Further, a four-character word having the first character a of the third candidate as the first character is stored in the table 2. Similarly, for words of 5 characters and 6 characters, all words having the first character of ya, i, and a are obtained and stored in the table 2 (64).
[0064]
S8 The integration unit searches the table 2 and adds a word having a length of n characters as a result of integrating the voiced and semi-voiced dots to the table 1 (for example, integrating the four-character Yamata タ and the voiced point “゛” 3 characters and added to Table 1 (63)).
[0065]
S9 All the words held in the table 1 are collated with the input character candidates, the distance is calculated, and the word with the shortest distance is output as a recognition result of the knowledge processing.
FIG. 8 shows an embodiment of the display method of the character recognition device of the present invention.
[0066]
FIG. 8A shows an example of a method of displaying the recognition result of the knowledge processing according to the present invention.
FIG. 8A shows a display when the recognition result of the knowledge processing is Nozaki.
Since half-width fonts do not have character fonts with voiced dots, even if voiced characters are integrated and recognized as three characters according to the present invention, each character and voiced dot ゛ are displayed as single-width characters. For example, when a Nozaki is displayed, four characters are displayed with one voiced character as shown in the figure. Conventionally, when the cursor is moved from the first character to the fourth character in such a display, it is necessary to perform the cursor movement operation four times, assuming that the voiced dot is also one character. According to the present invention, in the case of displaying a half-width character, a voiced point and a voiced halftone point are integrated into one character, and the integrated portion is operated by one cursor operation.
[0067]
Reference numeral 92 denotes a state where the cursor is located at the first character. Reference numeral 93 denotes a state in which the second character "S" is integrated with the clouded point "@" and the cursor is located below the clouded point "@". Reference numeral 94 denotes a state where the cursor is located at the fourth character key. The cursor can be moved from the first character to the fourth character by performing the cursor processing three times. In addition, even when deleting and changing the cloud point “゛”, the operation can be performed by one operation.
[0068]
FIG. 8B is a block diagram of display control of half-width characters according to the present invention.
In FIG. 8B,
Reference numeral 95 denotes a recognition result of the knowledge processing, which is obtained by recognizing Yamada (three characters) by integrating the character "ta" and the turbid point "@".
[0069]
Reference numeral 96 denotes a display control unit for performing display control.
Reference numeral 97 denotes a display integration processing unit, which performs processing for integrating half-width characters.
[0070]
Reference numeral 98 denotes a font holding unit for displaying single-byte characters.
99 is a display.
When a recognition result of a half-width character including a voiced character such as Yamada 95 is input, the display control unit 96 extracts the half-width character “Y”, “ma”, “ta”, and the voiced character “゛” from the display half-width character font storage unit 98. . Then, the display integration processing unit 97 performs the integration processing so that the character and the character are handled as one character, and performs a cursor operation, a character deletion, a change, and the like in one operation.
[0071]
FIG. 9 is a first embodiment of the flowchart of the recognition result display control unit of the present invention.
S1 Input the recognition result.
S2: It is determined whether the character of the recognition result is a half-width character or not. If it is a double-byte character, display processing of the double-byte character is performed in S5. If it is a half-width character, the process proceeds to S3.
[0072]
S3 Determine whether there is a cloud point or a semi-voice point. If there is no voiced point or semi-voiced point, display processing of half-width characters is performed in S5. If there is a voiced point and a half-voiced point, the integration processing is performed so that the voiced-point and the half-voiced point are treated as one character, and the voiced point and the half-voiced point integrated in S5 are integrated into the preceding half-width character and treated as one character to perform the display processing. I do.
[0073]
FIG. 10 is a second embodiment of the flowchart of the recognition result display control unit of the present invention. The half-width characters with a voiced point and a half-voiced point are handled as full-width characters with a voiced point and a half-voiced point.
[0074]
S1 Input the recognition result.
S2 Determine whether the character is a half-width character or not. If it is not a half-width character, display processing of a full-width character is performed in S5. If it is a half-width character, the process of S3 is performed.
[0075]
S3 It is determined whether there is a cloud point or a semi-voice point. If there is no voiced point or semi-voiced point, display processing of half-width characters is performed in S5. If there is a cloud point or a semi-voice point, the process of S4 is performed.
S4: Characters having a voiced point and a semi-voiced point are handled as full-width characters, and in S5, a double-width character with a voiced and a half-voiced point is displayed.
[0076]
A double-byte character with a voiced point and a half-voiced character displayed with the method of the flowchart of FIG. 10 before and after the half-width character without a voiced and a half-voiced voice is greatly unnatural. Full-width characters are designed to use small fonts.
[0077]
S1 Input the recognition result.
S2 Determine whether the character is a half-width character or not. If it is not a half-width character, a full-width character display process is performed in S4. If it is a half-width character, the process of S3 is performed.
[0078]
S3 It is determined whether there is a cloud point or a semi-voice point. If there is no voiced point or semi-voiced point, display processing of half-width characters is performed in S4. If there is a cloud point or a semi-voice point, the process of S4 is performed.
S4: Merges characters with clouded and semi-voiced dots into one character.
[0079]
S5 The integrated characters are displayed as small double-byte characters having a voiced and semi-voiced point.
[0080]
【The invention's effect】
According to the present invention, it is possible to obtain a correct recognition result by knowledge processing even when a character having a clouded point or a semi-voiced mark is written in the same frame. In addition, since the display of half-width characters with a cloud point and a half-voice point can be treated as one character, the recognition result can be corrected and changed efficiently.
[Brief description of the drawings]
FIG. 1 is a diagram showing a basic configuration of the present invention.
FIG. 2 is a diagram illustrating the operation of the basic configuration of the present invention.
FIG. 3 is a block diagram of the character recognition device of the present invention.
FIG. 4 is a diagram showing an example of a knowledge dictionary.
FIG. 5 is a block configuration diagram of an embodiment of a knowledge processing unit.
FIG. 6 is an explanatory diagram of the operation of the knowledge processing of the present invention.
FIG. 7 is a diagram showing a flowchart of an embodiment of a knowledge processing unit of the present invention.
FIG. 8 is a diagram showing an embodiment of a method of displaying a recognition result of knowledge processing according to the present invention.
FIG. 9 is a diagram showing Example 1 of the flowchart of the display control unit of the present invention.
FIG. 10 is a diagram showing Embodiment 2 of the flowchart of the display control unit of the present invention.
FIG. 11 is a diagram illustrating a third embodiment of the flowchart of the display control unit of the present invention.
FIG. 12 is an explanatory diagram of a conventional technique.
[Explanation of symbols]
1: Character recognition unit
2: Knowledge processing unit
10: Knowledge dictionary
11: output section

Claims (6)

A character reading unit for reading the characters of the form,
A character number determining unit for determining the number of characters of the character string read by the character reading unit;
A knowledge dictionary that handles each of the voiced and semi-voiced characters as one character,
A character string read from the number of characters obtained by adding 1 to the number of characters of the read character string, storing a word having the number of characters of the read character string determined by the number-of-characters determining unit with reference to the knowledge dictionary. A knowledge dictionary reference unit that stores words having a number of characters up to the number of characters that is twice the number of characters in the second table,
With reference to the second table, a word having the number of characters of the read character string, which is determined by the character count determining unit to integrate the voiced or semi-voiced voice, is extracted. Integration unit to store in the table of
A display unit that compares the words stored in the first table with the character recognition results, and displays the words in the shortest distance or in the candidate order in the order of the shortest distance, and
A character recognition device comprising: A feature extraction unit for extracting the characteristics of each character of the characters entered in the form, a character recognition dictionary having the characteristics of the characters for each character, and comparing the extracted characteristics of each character with the character recognition dictionary The character recognition dictionary is provided with a character recognizing unit, and the character recognition dictionary has a feature that a character with a voiced or semi-voiced character cut out into one character is viewed as one character.
2. The character recognition method of claim 1, wherein in the character recognition for each character, a character having a voiced or semi-voiced point and the character written in the same entry box is also recognized as a character with a voiced or semi-voiced point. The character recognition device according to 1. The character recognition device according to claim 1, wherein the voiced point, the semi-voiced point, and such characters are single-byte characters. In the character display of the recognition result of a half-width character having a voiced point or a half-voiced point, a character having a voiced point or a half-voiced point is integrated with the voiced point or the half-voiced point and such a character, and is displayed as one character. The character recognition device according to claim 1. In the character display of the recognition result of half-width characters with voiced or semi-voiced characters, the characters with voiced or semi-voiced voices are considered as one character by integrating the voiced or semi-voiced voice and such a character, and the character is voiced or half-voiced. 2. The character recognition device according to claim 1, wherein the characters are displayed in full-width characters. 6. The character recognition device according to claim 5, wherein full-width characters with a voiced or semi-voiced character are displayed as characters smaller than other characters.

Images