JPH0520797B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION FIELD OF APPLICATION OF THE INVENTION The present invention relates to a character recognition device having a function of correcting misread characters, which is suitable for cases where there are many character categories such as Chinese characters.

〓Prior Art〓 Conventionally, most of the processing of various applications in government offices, etc., has been done manually. These application forms are usually written in a mixture of kanji and kana, and if the application processing task were to be automated, the input unit would need a recognition device for Japanese characters, including kanji. Currently, at the laboratory level, we have successfully conducted a principle experiment of a printed kanji recognition device with reading accuracy that is satisfactory for practical use (see, for example, Journal of the Institute of Electronics and Communication Engineers, Vol. 58-D, No. 2, p. 94). Considering that most of the various application forms mentioned above are relatively high-quality type-printed documents in Japanese type, it can be said that the environmental conditions for using the printed kanji recognition device in the above application processing work are in place.

However, if a printed kanji recognition device is to be put into practical use, a fairly high degree of recognition is required due to the nature of the application process. On the other hand, considering that kanji has an extremely large number of characters, and even if the print quality is relatively good, it is possible that an application form of relatively poor quality may be input, so the reading accuracy is quite sufficient. It can not be said.

However, it is conceivable that the rate of misrecognition can be significantly reduced by testing whether or not the recognition results are correct. Conventionally, the above idea was carried out as follows. Character recognition devices that target numbers often handle monetary amounts, so for example, the amount of each item and the total amount are written on the form, and the recognition device records the total amount of recognition results for each item, A method is used to detect errors by comparing the total recognition results. In addition, in character recognition devices that target English letters, a method called N-gram is used on the premise that each English letter constitutes one word in a limited vocabulary. It will be done.

However, the above conventional method cannot be directly applied to a character recognition device for Chinese characters.
The reason for this is that in the case of kanji, the character types are alphanumeric (at most
50 characters), it is 2000 to 4000, which is more.
For example, the storage capacity of an N-gram table becomes enormous.
It will become impossible to do so.

<Object of the Invention> Therefore, an object of the present invention is to provide a means for correcting the reading result using word information as a method suitable for cases where there are many character types, and to reduce the overall recognition error rate.

[Summary of the invention]

In order to achieve the above object, the present invention includes an input means for inputting a character pattern, and an output for outputting a reading result obtained by comparing the character pattern inputted by the input means with a standard pattern for each word. means for storing word information; means for collating the reading result for each word with the word information; If it does not match any word, the word is determined by calculating the degree of similarity between the standard character pattern corresponding to each character of the word output as a reading result and the standard character pattern corresponding to each character of the candidate word in the word information. A test means for testing whether the word can be identified; a correction means for correcting the reading result to the specified word if the word can be identified by the test means; and a correction means for changing the word that cannot be corrected to another word. The feature is that it is equipped with a display means to display the information separately.

〓Example of the invention〓 For example, consider an application with the following documents.

(Example) Type of application Registration application Purpose of registration Full transfer Cause February 2, 1972 Holder of right to purchase Name Taro Koyama Location 1-1 Kunitachi-shi, Tokyo Equity 1/3 Obligor Name Jiro Otokawa Address: 2-2 Tachikawa-shi, Tokyo Application date: March 3, 1978 The principle of the apparatus of the present invention will be briefly explained using the flowchart shown in FIG. First, at 201 and 202, characters on a form are photoelectrically converted, cut out within a certain frame, recognized line by line, and the recognition results are output for each line in the form of a character code. The recognition unit continues the above operation until all characters on the form are recognized. Everything up to this point is the same as the conventional character recognition device. Next, the recognition result verification unit extracts the character sequence of the recognition result corresponding to the key item (character string printed in a field of a predetermined length on the left side of one line), and creates a dictionary in which all key items are stored. 203 recognizes which key item this character sequence corresponds to among the following. This is called key item word recognition. Note that character recognition may be incorrect, so the word recognition method described above needs to be devised. The method will be described later.

Once you know the number of the key item, you can limit the character types that can appear in the fixed items following this key item.
204 specifies the character type. Therefore, in the character code string resulting from character recognition, the character code corresponding to the fixed item is checked to see if it is included in the above-mentioned permissible character types in step 205. At this time, if it is found that it is not included, the result of character recognition is incorrect.
Either the characters on the form were misspelled.
Therefore, in this case, the above information is output along with the recognition result. For example, reverse the sign of the character code. As a result of the test, if the character type is included in the allowable character types, it is assumed to be read correctly and the character code is output as is.

Continue the above operations until there are no more characters on the form.

Next, a procedure for recognizing a character sequence resulting from recognition as a word, which is the main point of the present invention, will be explained. Generally, in order to recognize words, it is sufficient to prepare a word dictionary (a table consisting of character code strings that make up each word) and check which dictionary item the input character sequence matches. However, in reality, not all input character sequences are read correctly, so there may be cases where a complete match cannot be achieved with any dictionary entry. Therefore, it is necessary to perform word recognition by determining the distance or equivalently the degree of similarity (to be defined later) between the input character sequence and each dictionary item, rather than determining whether or not there is a match with the dictionary item.

For example, "application date" may be obtained as a result of reading "application date," but there is clearly no dictionary entry for "application date."

If the degree of similarity between a character series and a dictionary entry is the degree of similarity between each character, then in the above example, the degree of similarity between "Monkey" and "A" is required. However, such a combination of two characters is difficult to read when the number of characters to be read is 2000.
There are 400,000 combinations, which are impossible to remember. Therefore, in the device of the present invention, when it becomes necessary to measure the similarity between different characters (in the above example, “A” and “Mon”), the device calculates the similarity between the corresponding standard patterns in the recognition device. Use that value. The similarity between the same characters is always 1.

Here, the degree of similarity is a numerical value ranging from 0 to 1, defined between two character patterns, easily calculated by a dedicated calculation circuit, and is well known, so its explanation will be omitted here.

The word recognition algorithm using the above method is
This will be explained using the flowchart shown in the figure. First, each dictionary entry consists of the number of characters N _k that make up the word and the character code string
It is expressed as W _k = {wi(k)〓i=1, 2,..., N _k }. Let K be the total number of dictionary items. k above
is an item number (word number) and takes values from l to K. In addition, the character sequence (character code string) of the character recognition result input to the word recognition unit is S = {si〓
i=1, 2,...,N}. character series S and
The degree of similarity with W _k is expressed as 〓k.

Figure 2 shows the structure of a dictionary necessary for word recognition.

The first word 501 (number D) in the dictionary is the number of key items K
, followed by a word 502 that stores addresses A ₁ , A ₂ , . . . , A _K for storing character code strings of each item.
Next follows a word that stores the character code string of each key item. For example, A ₁ address 503 holds the length (number of characters) N ₁ of the word of item number 1,
The following _N1 word 504 stores each character code.

FIG. 3 illustrates a character code string that is a target of word recognition. The character code string is stored in a working area of the memory and consists of N words.

In FIG. 4, word recognition is performed as follows. First, initialize with 101 and 102.

In step 103, it is determined whether the word length matches the input character sequence length or not, and if they do not match, the similarity 〓k remains at 0 and the next word is examined. When the word lengths match, the similarity 〓
Find k.

Initialization is performed in step 104, and in step 105, the i-th character code wi(k) of the k-th item in the dictionary and i of the input character sequence are
Check whether the th character code si matches or not,
If they do not match, 1 is added to 〓k in step 106, and if they do not match, it is checked in step 107 whether it is impossible to determine. When s _k = 0, it indicates that it is impossible to determine,
In this case, 106 is executed, and when s _i ≠ 0, in 108, wi(k) is executed using the standard pattern in the recognition device.
The similarity between the standard pattern of and the standard pattern of si is calculated and added to 〓k. With the number of characters i up to that point 〓k
109 to determine whether the divided value exceeds the threshold value
If it does not exceed 113, item k is selected from the candidates.
Excluded in If the number exceeds the number, the process advances to the next character, and when the processing in steps 105 to 111 is completed for all characters, in step 112, the degree of similarity between the character series is divided by the number of characters N to normalize it.

When it is detected in 115 that all dictionary items have been processed, the maximum value of all similarities {〓k〓k=1,2,...,K} calculated in 116 is ¹ and the next Find the value 〓 ² , compare the absolute threshold 〓 and 〓 ¹ , and then test whether there is a sufficient difference between 〓 ¹ and 〓 ² using the relative threshold 〓. outputs the word number k ^* that gives 〓 at 119. If it is not sufficient, it outputs undeterminable at 120.

Next, a method for specifying character types that can appear in fixed items following a key item will be explained. In the present invention, a flag table as shown in FIG. 5 and a bit number conversion table as shown in FIG. 6 are prepared. When the word recognition result for the key item is k ^* , first refer to the bit number conversion table and select the bit position number b (k ^* ) that indicates which bit in the flag table is to be used.
seek. Next, extract the contents of the flag table for an arbitrary character, and if the value of the b(k ^* ) bit item is 1, the same character is allowed as a character type following the same key item, and if it is 0, it is not allowed. I understand.

Therefore, using this result, the recognition result can be tested as described in the explanation of the principle.

Hereinafter, the present invention will be explained in detail with reference to Examples.

FIG. 7 is a block diagram of one embodiment of the apparatus of the present invention. An example will be described below with reference to the same figure.

In the figure, 1 is a conventional character recognition device, 3 is a character observation section for observing unknown patterns, 4 is a character recognition processing device, and 5 is a standard pattern storage device.
The above portions are well known and will not be described in detail here. The output 6 of the recognition processing device 4 is the result of recognizing characters on a form line by line, and is transferred in the form of a character code string. Here, when the character code is 0, it means that the character is unrecognizable.

The verification processing device 10 includes a memory 11, a similarity calculation circuit 30, and a microprocessor 20. The circuit 30 receives two character codes from the microprocessor 20, receives two standard patterns corresponding to the same character code from 5, calculates the similarity between the same standard patterns, and calculates the resulting similarity. 20
Transfer to. Circuit 30 is used when performing process 108 of FIG.

The memory 11 includes a section 12 for storing the flag table shown in FIG. 5, a section 13 for storing the bit number conversion table shown in FIG. 6, and a section 14 for storing the key item dictionary shown in FIG. , and further work area 1
It consists of 5.

The microprocessor 20 performs word recognition (key item recognition) using the algorithm explained in FIG. 4 according to the microprogram contained in the microprocessor 20.
The character type of the fixed item is specified using step 12, and the character code that is the recognition result of the fixed item is verified.

Next, the flow of processing as a character recognition device will be explained.

The character pattern applied on the form is photoelectrically converted by 3, cut out within a certain frame, and transferred to 4. In step 4, the degree of similarity between the unknown pattern sent from step 3 and each standard pattern in step 5 is calculated, and the character codes that give the maximum degree of similarity are summarized for one line and output as a character code string on output line 6. Output.
However, here, 4 tests whether the maximum similarity is equal to or higher than a predetermined threshold, and if it does not reach the threshold, the output code is set to 0.

Microprocessor 20 in the verification processing device 10
receives the character code string of the recognition result for each line through 6 and stores it in memory 15. First, a character code sequence corresponding to a key item is extracted from a character sequence for one line (blanks are also given one character code), and the process moves on to word recognition. An example of a character code string for one line is shown in FIG. One line consists of 25 characters, the first 8 characters 801 correspond to key items, and the latter 17 characters 802 correspond to fixed items. Character code
9999 means blank. Non-blank character codes in column 801 (in Figure 8, s ₁ , s ₂ ,
..., s ₆ ) is the character code string resulting from recognizing the characters of the key items.

Word recognition is performed by a microprogram according to the algorithm shown in FIG. However, in the same algorithm, the process 108 in FIG.
This is done by a similarity calculation circuit. That is, 20 transfers the two character codes s _i and the i-th character code wi(k) of the k-th dictionary entry (see Figure 4) to 30,
Issue a similarity calculation command to 30. 30 receives the same command, reads out two standard patterns corresponding to si and wi(k) from 5, calculates the degree of similarity 〓 ^* between the same standard patterns, and returns it to 20. The above is the process 203 in FIG.

When the microprogram word recognition is completed, the process moves to verification processing. First, processing 204 in FIG. 1 is performed.
First, when the key item number as a result of key item recognition is known, the bit number conversion table in the memory 13 is checked.
Obtain the bit number b ^* of the flag table that specifies the character type of the fixed item following the same key item. Subsequently, a verification process 205 of the fixed item recognition results is performed. 20 is the character code 802 to 1 corresponding to the fixed item in the recognition result character code string (FIG. 8) in the memory 15.
The flag table in memory 12 (fifth
b _* of flags corresponding to each character code (see figure)
Examine the bits. When the same bit is 1, it is an acceptable character type, so nothing is done at that time, but
If it is 0, the character type is not allowed, so the sign of the character code in 802 that gave the same result is inverted.
For example, if the recognition result character code with a fixed item is
500, and if the character is not allowed as a result of the verification, the same sign is reversed and it becomes -500.

Here, the character code of the fixed item recognition result is 4.
If the code is negative when sent from , no verification processing is performed for the same character code.

Regarding the character code corresponding to the key item, the character code string of the dictionary resulting from word recognition is substituted into the character code string of the key item shown in FIG. For example, if the character recognition result 801 is "Application date" but the word recognition result is a key item number corresponding to "Application date", the microprocessor 20
retrieves the character code string corresponding to "application date" instead of "application date" from the dictionary stored in memory 14 and rewrites 801, so even if there is an error in the character recognition result, it will be corrected correctly. . If the result of word recognition for a key item is undeterminable, subsequent character codes cannot be verified, so all accompanying character codes are inverted to negative values.

When the verification is completed and the character code string shown in FIG. do.

The above process is executed for each line of the form.

[Effects of the Invention] As explained above, the device of the present invention correctly corrects erroneous recognition of character recognition results and outputs them.
The misrecognition rate can be reduced.

The results of this character recognition device can be displayed, for example, as shown below, and the final judgment can be made manually. In other words, if a positive character code is output, it will be displayed normally, but if a negative character code is output, there is a high possibility of misrecognition, so please change the brightness or color of the display, or change the gutter display. Characters can be displayed with special symbols added to the side, and can be corrected manually. The features of the device of the present invention are that it does not require major changes as it can be installed after a conventional character recognition device, and that the verification process can be easily removed and the recognition unit can operate as a conventional character recognition device as is. , the main verification processing section can be handled as an option.

Another feature of word recognition is that the measure of the closeness between any two characters, which is required when searching a dictionary from a character code sequence containing errors, is obtained from the similarity between standard patterns. It is. Therefore, extensive storage for storing proximity measures is unnecessary. In the embodiment described in this specification, the similarity calculation circuit 30 is provided in 10 in FIG. 5, but since 4 originally has the similarity calculation function, 4 may be slightly modified. Accordingly, 3
0 can be included in 4 to make it more efficient overall.

[Brief explanation of the drawing]

FIG. 1 is a flow chart for explaining the principle of the present invention. FIG. 2 is a diagram of the structure of a word dictionary, and FIG. 3 is a diagram of character code strings corresponding to key items.
FIG. 4 is a flow chart for explaining the word recognition (key item recognition) algorithm, FIG. 5 is a flag table, and FIG. 6 is a bit number conversion table. FIG. 7 is a block diagram of one embodiment of the present invention. FIG. 8 is a diagram of a character code string resulting from character recognition.

Claims (1)

[Scope of Claims] 1. A character recognition device comprising an input means for inputting a character pattern, and an output means for outputting a reading result obtained by comparing the input character pattern with a standard pattern for each word. , storage means for storing word information; means for comparing the word information with the reading result for each word outputted by the output means; If the word does not match, the word is determined by calculating the similarity between the standard character pattern corresponding to each character of the word output as a reading result and the standard character pattern corresponding to each character of the candidate word in the word information. a test means for testing whether a word can be identified by the test means; a correction means for correcting the reading result to the specified word if the word can be identified by the test means; and a word that cannot be corrected by the correction means. A character recognition device comprising: display means for displaying words in a manner that distinguishes them from other words;