JPH0816730A - Character recognition system - Google Patents

Abstract

PURPOSE:To improve character recognition processing efficiency and to reduce the burden of an operator at the time of correction or the like by performing word recognition processing by rearranging the order of candidate characters selected by a character recognition part while considering frequency in use. CONSTITUTION:A use frequency table 123 showing the frequency in use corresponding to respective characters up to the last time and a rearranging part 213 is provided with a function for rearranging the order of respective candidate characters selected by a character recognition part 232 while using the data of the use frequency table stored in a use frequency table storage part 222 and for outputting the candidate characters in that rearranged order to a word recognition processor 100. Then, the data of the use frequency table 123 are transferred from the word recognition processor 100 to a character recognition device 200 and on the side of the character recognition device 200, the order of candidate characters is rearranged while considering the contents of this use frequency table 123.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition system for reading characters on a form and performing character recognition processing on the characters.

[0002]

2. Description of the Related Art As a character recognition system for optically reading characters on a form and recognizing the contents of the characters, a character recognition device and information processing for performing knowledge processing of Chinese characters based on the character recognition result of the character recognition device. A configuration including a device and the like is known. In such a character recognition system, the character recognition device is a device that reads characters on a form and performs character recognition processing for each character, and the information processing device is a character recognition result of such a character recognition device. Based on
It performs processing such as word recognition processing and correction processing of each character.

[0003]

However, in the above-described conventional character recognition system, the character recognition processing is performed independently for each form. For example, the character recognition processing results up to the previous time are used as the current character recognition processing. It was not configured to reflect it. Therefore, for example, when a character recognition error occurs in the character recognition device, in general, the word recognition process recognizes a correct character. There was a problem of reduced efficiency. From such a point, it has been desired to realize a character recognition system that can improve the efficiency of the character recognition processing by making the most of the character recognition processing results up to the previous time.

[0004]

In order to solve the above-mentioned problems, the character recognition system of the present invention is provided with a usage frequency table showing the usage frequency of each character up to the last time. And
A rearrangement unit is provided for rearranging the ranks of the candidate characters of the character recognition unit that selects and ranks a plurality of candidate characters for each character based on the frequency of use in the frequency-of-use table. Furthermore,
It is characterized in that a word recognition processing unit for recognizing a word of each character based on the rank of the candidate characters rearranged by the rearranging unit is provided.

[0005]

In the character recognition system of the present invention, the character recognition unit performs character recognition for each character and ranks and selects a plurality of candidate characters for each character. The sorting section
The order of the candidate characters selected by the character recognition unit is rearranged based on the usage frequency of the usage frequency table. The word recognition processing unit performs word recognition composed of each character based on the rank of the rearranged candidate characters.

[0006]

Embodiments of the present invention will now be described in detail with reference to the drawings. << Embodiment 1 >> FIG. 1 is a block diagram showing Embodiment 1 of the character recognition system of the present invention. The system shown in the figure comprises a word recognition processing device 100 and a character recognition device 200. The word recognition processing device 100 includes an information processing device such as a personal computer or a workstation, and includes a control unit 110, a storage unit 120, a display 131, a keyboard 132, and a mouse 133.

The control unit 110 includes a word recognition processing device 100.
Has the function of performing word recognition processing and control of each part as
A word recognition processing unit 111, a table data management unit 112,
A table data transfer unit 113 is provided.

The word recognition processing unit 111 is composed of a word collation processing program and a data correction processing program, recognizes characters in word units based on the character recognition result received from the character recognition device 200, and recognizes the characters. The result is displayed on the display 131 or the keyboard 132
And a function of correcting the recognition result of each character according to the input result from the mouse 133.

The table data management unit 112 is a storage unit 1.
20 is used to manage the data of a usage frequency table 123, which will be described later, stored in the memory 20. If there is, it has a function of increasing the counter value of the usage frequency counter. The table data transfer unit 113 has a function of controlling transfer of the contents of the usage frequency table 123 to the character recognition device 200.

The storage unit 120 is composed of a storage device such as a magnetic disk device or a semiconductor memory, and has a form format data 121, a dictionary file 122 and a use frequency table 123.

The form format data 121 is the form data of a form for which character recognition is performed by the character recognition device 200. That is, in each form, the area to be recognized (position of the entry column for handwriting or printing of characters) is predetermined for each form together with the shape of the form, and the format is the area to be recognized ( It is data such as fields) and the size of the form, and the character set table 221 used in each entry field. The character set table 221 will be described later.

The dictionary file 122 is dictionary data for word recognition for each area on the form. That is, in each form, the entry place is determined, and the content to be entered in each column is also determined to some extent. For example, since the characters entered in the address column are prefecture names or place names, these word dictionaries are required. Therefore, in the character recognition system, a dictionary file that collects words to be written in each field is created in advance.

The usage frequency table 123 is a table for indicating the usage frequency corresponding to each character up to the last time, and is configured as follows. FIG. 2 is an explanatory diagram of the usage frequency table 123. That is, the usage frequency table 123 includes the character codes of the characters that have appeared in the recognition process up to that point, the usage dictionary information indicating in which dictionary in the dictionary file 122 each character is used, and the information for each dictionary. It is composed of the frequency counter information indicating whether or not it has been used once.

For example, in the case of the character code x, the use dictionary information stores information indicating that the character code x is used in the dictionaries A, B, ... And the frequency counter and the dictionary in the dictionary A are stored. The frequency counter for B indicates the frequency of use in each dictionary A, B, ... For the character code x.

Further, the display 131 to the mouse 133
Is an input / output device used when the operator corrects data or gives a work instruction.

The character recognition device 200 is an optical character reading device (OCR) that optically reads a handwritten character entered on a form and performs character recognition for each character. It is composed of a storage unit 220. The character recognition processing unit 210 has a function of performing recognition processing as the character recognition device 200 and control of each unit.
11, a character recognition unit 212 and a rearrangement unit 213. The storage unit 220 also includes a character set table 221 and a usage frequency table storage unit 222.

The reading unit 211 is composed of an image reader or the like, and optically reads characters written on a form,
It has a function to output as image data. The character recognition unit 212 uses the reading result of the reading unit 211 as the form format data 12 from the word recognition processing device 100.
Based on 1, the entered characters are analyzed and recognized one by one, and a plurality of candidate characters are ranked and selected from the character set table 221 for each character. The rearrangement unit 213 determines the rank of each candidate character selected by the character recognition unit 212 from the use frequency table storage unit 22.
It has a function of performing rearrangement using the data of the frequency-of-use table stored in No. 2 and outputting the candidate character of the rearranged rank to the word recognition processing device 100.

The character set table 221 of the storage unit 220 is a table in which character codes of all the characters existing in each dictionary are recorded. That is, in the character set table 221, the character code of each character,
A dictionary file number indicating which dictionary each character is used is stored. Further, the usage frequency table storage unit 222 stores the usage frequency table 12 transferred by the table data transfer unit 113 of the word recognition processing device 100.
This is an area for storing the data of No. 3.

Next, the operation of the character recognition system having the above configuration will be described. First, before explaining the operation of the character recognition system of the present embodiment, the basic operation of the character recognition system excluding the rearrangement processing by the use frequency table will be described.

<< Basic Operation in Character Recognition System >> First, when the reading of a form is started in the character recognition device 200, the identification data (for example, form number) of the form described in a predetermined area is displayed. Read. As a result, the character recognition unit 212 requests the word recognition processing device 100 for the form identification data as well as the form identification data. Word recognition processing device 10
0 transmits the format data of the corresponding form to the character recognition device 200 based on the identification data of the form transmitted.

The character recognition device 200 recognizes the field on the form to be character-recognized and what kind of data is to be written in the field based on the format data, and performs the character recognition process. That is, the character recognition unit 2
12 is a character set table 2 for each field
Characters that are considered to be similar to the character registered in 21 are ranked in order of high possibility based on the characteristics of the glyph, and are selected as candidate characters.

For example, when performing character recognition processing for a certain field, the dictionary file (for example, "A") used in the field is known based on the format data. Of the characters registered in the character set table 221, only the characters having the dictionary file = "A" are selected as candidates for the candidate character. Then, these candidate characters are transferred to the word recognition processing device 100.

The word recognition processing section 111 of the word recognition processing apparatus 100 considers the order of the transferred candidate characters,
Character recognition processing is performed by matching the dictionary files 122 and selecting the optimum word. For example, the character recognition result of the prefecture name from the character recognition device 200 is “Saio prefecture”.
It was supposed to be. That is, the rank of the second candidate character is 1. King 2. This is the case when it was a ball.

In such a case, the word recognition processing unit 111
Performs word recognition based on the dictionary corresponding to that field. That is, there is no "Saio prefecture" in the dictionary, and 2
Since there is “Saitama prefecture” when the character is “tama”, which has the second rank of candidate characters, this is adopted. Then, the word recognition processing unit 111 displays the word recognition result on the display 131 together with the image data of the characters written on the form, and prompts the operator to confirm. When the operator wants to make corrections to the characters displayed on the display 131, the character recognition device 20
A list of candidate characters sent from 0 is displayed and selected, and a character which seems to be the correct answer is selected from among these by the keyboard 132 and the mouse 133.

However, in such an operation, the character recognition device 200 only selects a character that is considered to be close to the character as a candidate character. Therefore, for example, when many characters having a character shape similar to a certain character are registered in the character set table 221, the correct character may be excluded from the candidate characters. That is, the candidate character sent from the character recognition device 200 to the word recognition processing device 100 is limited to how many characters each character has (how many words the character recognition device 200 wants from the word recognition processing device 100 side). Therefore, if the correct character has a higher rank, it will not be sent as a candidate character.

As a result, the word recognition processing unit 111 of the word recognition processing apparatus 100 recognizes a word in the dictionary file 122 from the received candidate characters, and therefore, when the correct character is not received as the candidate character, , There is a risk of becoming unrecognizable.

Further, even if the correct answer character is sent to the word recognition processing apparatus 100 as a candidate character, the word recognition processing unit 111 becomes closer as the rank gets closer to the lowest rank.
The processing efficiency of word recognition in will be reduced.

From this point of view, in the first embodiment, it is possible to perform accurate and efficient word recognition processing even when there are restrictions on the candidate characters sent from the character recognition device 200 side.
The data of the frequency-of-use table is used when selecting the candidate character of the character recognition device 200, and the operation will be described as the operation of the first embodiment.

<< Operation of Embodiment 1 >> FIGS. 3 and 4 show a word recognition processing apparatus 100 and a character recognition apparatus 2, respectively.
10 is an operation flowchart of 00. First, when a form is inserted into the character recognition device 200, the reading unit 211 starts reading (step S41) and makes a form format data request as described above (step S42).

When the word recognition processing device 100 receives from the character recognition device 200, the word recognition processing device 100 determines the received content (step S31). If this is a form format data request, as described in the basic operation, the form format data 121 corresponding to the form identification data is taken out and transferred (step S3).
2) At the same time, it is determined whether to use the usage frequency table 123 from the extracted form format data 121 (step S33). Note that the usage frequency table 1
Whether or not to use 23 is recorded in the form format data 121 at the time of designing the form format.

When the usage frequency table 123 is "used" in step S33, the table data transfer unit 113 transfers the usage frequency table 123 to the character recognition device 200 (step S34), and then proceeds to step S31. Return. If the usage frequency table 123 is "not used" in step S33, the process directly returns to step S31.

The character recognition device 200 transfers the form format data 121 from the word recognition processing device 100, and when the usage frequency table 123 is transferred from the word recognition processing device 100 side in step S34, this data is transferred. Upon reception (step S43), the character recognition unit 212 starts character analysis processing (step S4).
4).

That is, the character recognition unit 212 performs a predetermined character analysis process based on the format data of the form from the image data of the character read by the reading unit 211, as described in << Basic operation >>, From the characters registered in the set table 221, the characters of the dictionary file used in that field are selected as candidates for the candidate character. Then, from these target characters, the characters that are considered to be similar to the characters recognized in the analysis process are selected as candidate characters by ranking them in the order of high possibility based on the characteristics of the glyphs (step S45). .

Next, the rearrangement unit 213 determines in step S4.
From the form format data received in step 3, it is determined whether or not to use the usage frequency table 123 (step S
46). If it is determined in step S46 that the usage frequency table 123 is used, the order of the candidate characters is rearranged in consideration of the contents of the usage frequency table 123 (step S47).

That is, for all the candidate characters selected for the recognized character, the value of the frequency counter in the corresponding dictionary is checked, and the ranking is re-evaluated based on the value of this frequency counter. For example, when the rank value of the candidate character is low, but the counter value of the usage frequency table 123 is high, this counter value is reflected to increase the rank of the candidate character. It should be noted that the degree of reflection of this counter value in the ranking is previously determined empirically.

When the rearrangement of the candidate characters is completed in step S47, or when it is determined in step S46 that the use frequency table 123 is not used, the candidate characters of one sheet or the read characters are read in the order of the candidate characters. The character recognition result such as the image data of the written character is transferred to the word recognition processing device 100 (step S4).
8).

The word recognition processing apparatus 100 has step S3.
In the reception of 1, when the received content is the character recognition result including the candidate character from the character recognition device 200, the word recognition processing unit 111 performs the word matching process (step S3).
5). That is, for each field, the corresponding dictionary file 122 is referred to, and character recognition as a word is performed. When this is finished, the word recognition processing unit 111 displays the processing result of the word recognition and the image data of the written character received from the character recognition device 200, and the operator confirms and
A correction process is prompted (step S36).

When the operator confirms and inputs the recognition contents, the table data management unit 112 uses the form format data 121 to retrieve the usage frequency table 12 of the form.
The use / non-use of No. 3 is checked (step S37), and if it is "use", the contents of the use frequency table 123 are updated (step S38). That is, if the character is not registered, the character code is registered, and if it is registered, the frequency counter value in the corresponding dictionary is incremented. In step S37,
If the usage frequency table 123 is "not used", the process directly returns to step S31.

In this way, each time the form is read by the character recognition device 200, the contents of the use frequency table 123 up to the previous time are transferred to the character recognition device 200. Therefore, even a candidate whose rank as a candidate character is low at first becomes higher as a correct character while reading the form, and the recognition rate can be improved.

In the first embodiment, the character recognition device 20
Although all the data of the usage frequency table 123 is transferred every time the form format data request of 0 is received, it may be configured to transfer only the data of the updated character code. By doing so, the transfer time of the usage frequency table 123 can be shortened.

As described above, in the first embodiment, the data of the usage frequency table 123 is transferred from the word recognition processing device 100 to the character recognition device 200, and the contents of the usage frequency table 123 are transferred on the character recognition device 200 side. Since the order of the candidate characters is rearranged in consideration, the order of the candidate characters output from the character recognition device 200 is in the order of high use frequency. The rate can be improved and the time for word matching processing can be shortened. In addition, the first embodiment is capable of performing accurate word recognition processing even when there are restrictions on the candidate characters sent from the character recognition device 200 side.

In the first embodiment, the character recognition device 200 is arranged to rearrange the candidate characters, but the word recognition processing device 100 may be arranged to rearrange the candidate characters. This will be described below as a second embodiment.

<Second Embodiment> FIG. 5 is a block diagram of a character recognition system according to a second embodiment. That is, the word recognition processing device 10
The control unit 110 of 0 is provided with a word recognition processing unit 111, a table data management unit 112, and a rearrangement unit 114. In this case, the table data transfer unit 113 of the first embodiment is unnecessary.

This rearrangement unit 114 has the same function as the rearrangement unit 213 of the character recognition device 200 in the first embodiment, and the contents of the usage frequency table 123 for the candidate characters received from the character recognition device 200. The order is rearranged in consideration of. Since the other configuration of the word recognition processing device 100 is the same as that of the above-described first embodiment, the description thereof is omitted here.

On the other hand, the character recognition processing section 210 of the character recognition apparatus 200 side is provided with a reading section 211 and a character recognition section 212, and the storage section 220 is provided with a character set table 221. That is, the rearrangement unit 213 and the usage frequency table storage unit 222 provided in the first embodiment are not necessary. Therefore, the character recognition device 200 only needs to have a basic function as a character recognition device.

Next, the operation of the second embodiment thus constructed will be described. << Operation of Second Embodiment >> FIGS. 6 and 7 are operation flowcharts of the word recognition processing apparatus 100 and the character recognition apparatus 200 in the second embodiment, respectively. First, when a form is inserted into the character recognition device 200, the reading unit 211 starts reading (step S71), and a form format data request is made as in the first embodiment (step S72).

The word recognition processing device 100 judges the contents received from the character recognition device 200 (step S61). If this is a request for the form format data, the form format data 121 corresponding to the form identification data is obtained.
Is taken out and transferred (step S62), and the process returns to step S61. That is, unlike the first embodiment, data transfer of the usage frequency table 123 is not performed.

When the character recognition device 200 receives the form format data 121 from the word recognition processing device 100 (step S73), the character recognition unit 212 starts the character analysis process (step S74), and the character set table 221 selects candidates. Select characters (step S7)
5). The above operation is similar to the operation of the first embodiment. Then, when the candidate characters are selected, the character recognition processing unit 210 transfers the character recognition result of one form including these candidate characters to the word recognition processing device 100 (step S).
77).

The word recognition processing apparatus 100 has step S6.
In the reception of 1, when the received content is the character recognition result from the character recognition device 200, the word recognition processing unit 111
First, based on the form format data 121, it is determined whether or not the form uses the use frequency table 123 (step S63). If the usage frequency table 123 is "used" in step S63, the sorting unit 114 sorts the ranks of the candidate characters in consideration of the contents of the usage frequency table 123 (step S64).

When the rearrangement in step S64 is completed, and when it is "not used" in step S63, the word recognition processing unit 111 performs a word collation process (step S65). That is, for each field, the corresponding dictionary file 122 is referred to, and character recognition as a word is performed. When this is completed, the word recognition processing unit 111
Is the recognition result of the word recognition process and the character recognition device 200.
The image data of the written characters received from the computer is displayed to prompt the operator for confirmation / correction processing (step S6).
6).

When the operator confirms the recognition contents, the table data management unit 112 determines that the form format data 1
From 21, the use frequency table 123 of the form is checked whether it is used or not (step S67), and “used”
If it is, the contents of the usage frequency table 123 are updated (step S69). That is, if the character is not registered, the character code is registered, and if it is registered, the frequency counter value in the corresponding dictionary is incremented. If the usage frequency table 123 is "not used" in step S67, the process directly returns to step S61.

As described above, also in the second embodiment, the use frequency table 123 up to the last time is used for the rank of the candidate character.
The content of is reflected. Therefore, even a candidate character that has a lower rank as a candidate character at the beginning becomes higher as a correct character while the form is being read, and the recognition rate can be improved.

Further, in the second embodiment, the character recognition device 200 only performs basic operations,
The character recognition device 200 is not required to have a high calculation processing capability, and is advantageous for a system in which the word recognition processing device 100 has a relatively higher calculation processing capability than the character recognition device 200. However, a sufficient number of candidate characters must be transferred from the character recognition device 200 side.

In each of the above embodiments, the form format data 121 is sent from the word recognition processing device 100 side every time the character recognition device 200 reads a form, but the present invention is not limited to this. If the form format data 121 is the same as the previous one, it may be configured not to be sent. Also, this form format data 1
21 is not sent from the word recognition processing device 100 side,
The character recognition device 200 may have it. Further, although the character recognition device 200 is an optical character reading device, the character recognition device 200 is not particularly limited to an optical character reading system, and may be, for example, magnetic reading.

[0055]

As described above, according to the character recognition system of the present invention, the order of candidate characters selected by the character recognition unit is rearranged in consideration of the frequency of use based on the word recognition results up to the last time, Since the word recognition processing is performed in this rearranged order, the efficiency of the character recognition processing can be improved and the operator's burden at the time of correction can be reduced.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a first embodiment of a character recognition system of the present invention.

FIG. 2 is an explanatory diagram of a usage frequency table in the character recognition system of the present invention.

FIG. 3 is a flowchart showing the operation of the word recognition processing device in the first embodiment of the character recognition system of the present invention.

FIG. 4 is a flowchart showing the operation of the character recognition device in the first embodiment of the character recognition system of the present invention.

FIG. 5 is a configuration diagram of a second embodiment of the character recognition system of the present invention.

FIG. 6 is a flowchart showing the operation of the word recognition processing device in the second embodiment of the character recognition system of the present invention.

FIG. 7 is a flowchart showing the operation of the character recognition device in the second embodiment of the character recognition system of the present invention.

[Explanation of symbols]

 100 word recognition processing device 111 word recognition processing unit 112 table data management unit 113 table data transfer unit 114 sorting unit 123 usage frequency table 200 character recognition device 212 character recognition unit 213 sorting unit 222 usage frequency table storage unit

Claims (3)

[Claims] 1. A character recognition system that reads characters on a form and performs word recognition processing on the form, performs character recognition for each character, and ranks a plurality of candidate characters for each character. A character recognition unit to be selected, a usage frequency table indicating the usage frequency of any character up to the previous time, and a sorting unit that sorts the ranks of the candidate characters selected by the character recognition unit based on the usage frequency of the usage frequency table. And a word recognition processing unit that performs a recognition process of a word composed of each character based on the rank of the candidate characters rearranged by the rearrangement unit. 2. A character recognition device that reads characters on a form and performs character recognition processing on each character, and word recognition that performs word recognition processing based on the character recognition result of each character received from this character recognition device. In a character recognition system comprising a processing device, the word recognition processing device, a usage frequency table indicating the usage frequency of any character up to the last time, and a table data transfer unit for transferring the usage frequency table to the character recognition device. The character recognition device performs character recognition processing for each character, and a plurality of candidate characters are given a candidate order for setting the order in which the word recognition processing device performs the word recognition processing, and each character is recognized. A character recognition unit that is selected for, a use frequency table storage unit that receives the use frequency table transferred from the table data transfer unit, and retains the contents. When the character recognition unit selects a plurality of candidate characters, based on the frequency of use of the frequency-of-use table stored in the table storage unit, the candidate ranks of the plurality of candidate characters are set to the rank of the candidate character with high frequency of use. The character recognition system further comprises: a rearrangement unit that rearranges the candidate characters so that the candidate characters are rearranged in the rearranged order and sends the candidate characters to the word recognition processing device. 3. A character recognition device for reading characters on a form, performing character recognition processing on each character, and selecting a plurality of candidate characters for each character in a candidate order, and selecting by this character recognition device. In a character recognition system including a word recognition processing device that performs a word recognition process based on the candidate characters, the word recognition processing device, a usage frequency table showing the usage frequency of any character up to the last time, the character recognition A rearrangement unit for rearranging the rank of each candidate character with respect to each of the plurality of candidate characters for each character received from the device so that the rank of the candidate character having a high frequency of use becomes high using the frequency of use table. A character recognition system comprising: a word recognition processing unit that performs a word recognition process based on the order of the candidate characters rearranged by the rearrangement unit.