JPH03252785A - Optical character reader - Google Patents

Abstract

PURPOSE:To facilitate document description by attaching data for recognition to recognize a left or right side bracket similar pattern in a dictionary for KATAKANA(square form of Japanese syllabary) character recognition. CONSTITUTION:A customer describes a left side bracket similar pattern or a right side bracket similar pattern in which a conventional left or right side bracket similar pattern is changed partially and distinction from the KATAKANA character (no), etc., is clarified on a document, etc. In an OCR to read and recognize the document, etc., the data for recognition to recognize the left side bracket similar pattern or the right side bracket similar pattern is attached on the dictionary for KATAKANA character recognition. Therefore, generally speaking, only the KATAKANA character is described in the column of the receiver name of a transfer document, however, an item for which the description of a left side bracket or right side bracket is permitted to be omitted can be recognized. Thereby, it is remarkably convenient for the description for a customer, and descriptive time can be shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an optical character reading device, and more particularly to an optical character reading device that recognizes katakana characters and parentheses.

[Prior Art] For example, in money transfer operations at financial institutions, an optical character reader (hereinafter referred to as OCR) may be used to automatically read money order forms and perform subsequent business processing.

FIG. 2 is a diagram showing an example of filling in the above-mentioned exchange form, FIG. 3 is a diagram showing an example of a screen display in which the result of reading and recognizing the exchange form by OCR is displayed on the screen of the display unit, and FIG. FIG. 2 is a diagram illustrating abbreviations that can be used in the gold and silver system as abbreviations for exchange forms.

A conventional document reading and recognition method using OCR will be described with reference to FIGS. 2 to 4.

The customer enters the necessary information (recipient's name, transfer bank name, amount,
Fill in the client's name, etc.) and submit it to the bank employee at the bank counter (hereinafter referred to as the teller).

After the teller fills in the necessary information for the bank on the form, the teller or another operator causes the OCR to read the form. The data read by OCR from the form is displayed for each entry as shown in the display example in Figure 3.
Characters as a result of OCR recognition and image data as an image before recognition are displayed on the display unit in vertical comparison. An operator checks the results displayed on this representation, corrects any errors, and performs operations to transmit the corrected data to a higher-level station such as a host computer.

Incidentally, an example of an item to be entered in a form to be read by OCR is the recipient's name on a money transfer form, but this item is to be entered in katakana characters. For example, if the beneficiary name is "Tozai Co., Ltd.," it is necessary to enter "Tozai Riki Bushiki Gaisha," but as shown in the diagram of current abbreviations in Figure 4, abbreviations are also allowed in foreign exchange transactions. ``Tozai force)'' or ``
It has the same meaning even if it is written as "Tozai (power)".

In the conventional technology, a dictionary containing only katakana characters is used as a built-in dictionary of OCR to read and recognize documents that include items such as recipient names and client names, and parentheses cannot be recognized. Or “(” or “)
In order to recognize ``, a dictionary containing katakana characters, English symbols, and letters was used.

[Problems to be Solved by the Invention] In the conventional optical character reading device (OCR) as described above, when reading items such as the above-mentioned recipient name and client name using a dictionary containing only katakana characters, , OCR cannot recognize "(" or ")" for abbreviations, so
Customers were not able to write in abbreviations, and it took time and inconvenience to fill in forms.

Furthermore, when using a dictionary that includes katakana characters, English symbols, and letters, it is convenient for the customer, but there is a problem in that the capacity of the dictionary becomes large and the cost of the device increases.

Furthermore, in this case, there is another problem in recognition that, for example, there is a possibility that a recognition error may occur between the kana character "ノ" and the English symbol ")".

The present invention was made in order to solve the above-mentioned problems, and provides customers with an optical character reading device that allows them to easily fill out forms, does not significantly increase the cost of the device, and improves the recognition rate. The purpose is to

[Means for Solving the Problems] An optical character reading device according to the present invention is an optical character reading device that recognizes at least katakana characters, and is capable of changing the conventional left-hand parenthesis or right-hand parenthesis written on a form, etc. In order to read and recognize left parenthesis-like patterns or right-hand parenthesis similar patterns that have been deformed to clearly distinguish them from katakana characters such as “ノ”, the left side parenthesis-like pattern or right-hand parenthesis similar pattern is written in advance in a katakana character recognition dictionary for recognizing katakana characters. It is provided with additional recognition data for recognizing at least either a parenthesis similar pattern or a right parenthesis similar pattern.

[Operations] In the present invention, in an optical character reading device that recognizes at least katakana characters, the conventional left-hand parenthesis or right-hand parenthesis is partially modified, and the left-hand side parenthesis is clearly distinguished from the ``)'' etc. of katakana characters. The left parenthesis similar pattern or the right parenthesis similar pattern is written in a form, etc., and in order to read and recognize the form, etc., the left parenthesis similar pattern or right parenthesis similar pattern is written in a katakana character recognition dictionary for recognizing katakana characters. To correctly recognize a left parenthesis and a right parenthesis without erroneously recognizing them as katakana characters by adding recognition data for recognizing at least one of similar patterns.

[Embodiment] FIG. 1 is a flowchart showing an embodiment of OCR recognition processing according to the present invention. Further, FIG. 5 is a diagram showing an example of the dictionary configuration of the OCR according to the present invention, and FIG.
It is a figure which shows the example of parenthesis entry format of CR.

FIG. 7 is an equipment configuration diagram of a money exchange data communication device using OCR according to the present invention. In the same figure, 1 is the exchange form,
2 is a tabletop optical data reader (hereinafter referred to as a tabletop OCR) for reading the character and image data of the exchange form 1; 3 is a simple keyboard; 4 is a remote OCR control device (hereinafter referred to as 0CRCTL); 6 is a public communication network; and 7 is a network control unit (hereinafter N) for transmitting and receiving modem signals via the public communication network 6.
8 is a remote OCR recognition device f (hereinafter referred to as M
9 is the MCU Workstation (hereinafter referred to as M
CU-WS), 10 is a CR7 display (hereinafter referred to as CRT).
), 11 is a keyboard, 12 is a printer, 13 is a dedicated communication line, and 14 is a host device.

41 is a microprocessor (hereinafter referred to as CPU);
It has a built-in memory of about megabytes (hereinafter referred to as MB). 42 is a communication control unit that performs data communication between OCRCTL4 and external equipment. 43 is a floppy disk (
For example, it has a capacity of about two 5-inch FDs (hereinafter referred to as FDs). 44 is a disk (DISK), which has a capacity of, for example, about 20 MB. A companding processor 45 performs data compression or data expansion. 46 is GPIB
Desktop 0CR2 with interface (hereinafter referred to as GPIB)
This is an interface for data communication with. 41-4 above
All 6 devices are built into 0CRCTL4.

The 81 is a CPU with about 5MB of built-in memory.

A communication control unit 82 performs data communication with the host device 14 via the dedicated communication line 13. 83 is an FD and has a capacity of about two 5-inch FDs. A disk 84 has a capacity of, for example, about 130 MB. Reference numeral 85 denotes a reception processor, which in this example is equipped with data reception lines of up to 6 channels (hereinafter referred to as CH), and performs data reception processing using a DMA interface method with a total throughput of about 24 KB (kilobytes)/second. Therefore, the communication line from each branch via the public communication network 6 and NCU 7 is CH
You can input from #1 to CH#6. Reference numeral 86 denotes a kanji recognition processor, which is equipped with a shared memory interface and performs recognition processing of kanji, which is multi-gradation, for example, 18-gradation image data. Reference numeral 87 denotes a local interface (hereinafter referred to as LIF), which is an interface for performing data communication between the MCU and the WS 9. All of the above-mentioned devices 81 to 87 are built into the MCU 3.

91 has a built-in CPU and about 5MB of memory.

Reference numeral 92 is an LIF interface for data communication with the MCU 3. 93 is an FD and has a capacity of, for example, two 5-inch FDs. 94 is a disk with a capacity of, for example, about 20 MB. 95 is a companding processor. All of the above devices 91 to 95 are built into the MCU-WS9.

The tabletop OCR2 can read money order forms written in handwriting or printed letters with a size of A6 to B4 at a reading speed of about 4 pages/minute. The read data is processed as character data and/or image data. The simple keyboard 3 has a communication interface with the OCRCTL 4, and also has a liquid crystal display or CRT capable of displaying about 40 characters and about 2 lines. Further, unless an identification (hereinafter referred to as ID) card is inserted into the simple keyboard 3, the keyboard cannot be operated. Therefore, only a specific operator who is allowed to possess this ID card can operate the simple keyboard 3. CPU U41.81 and 91 are each 5MB
It has a built-in IC memory of about 100 yen, but uses disks 44, 84 and 94 as large-capacity auxiliary storage devices, and FDs (floppy disks) 43, 83 and 93 as medium-capacity auxiliary storage devices, respectively. The FD is used to store an auxiliary control program such as loading an initial program at the time of initial operation.

First, the operation of each device as hardware in the money order data communication device using an optical character reader (OCR) shown in FIG. 7 will be explained.

The sales branch reads the money order form 1 on the tabletop 0CR2.

The actual operating procedure is as follows: First, an authorized operator inserts an ID card into the simple keyboard 3, and sets a number of exchange forms 1 to be automatically sent (one or more is fine) on the tabletop 0CR2. After that, by pressing the reading/sending start button, reading of the form is started. The information to be read from the exchange form 1 is shown, for example, in the example of form entry in FIG.

The tabletop OCR2 reads Kanji, characters, and image information as shown in FIG. 2. Kanji are read as multi-gradation data, and among the characters, kana letters or numbers are recognized after being read. Also, important information such as the amount, transfer account, and recipient's name, which are essential items in a money order form, are processed in parallel as image information for verification, even if the tabletop 0CR2 can be recognized as a character. Further, when the tabletop OCR2 cannot recognize the read character, it processes this read as image information. The tabletop 0CR2 reads the data as described above and performs some recognition processing, and converts it to 0CRCT.
0CR via interface CPI346 in L4
Supply to CTL4.

The OCRCTL4 writes the data read by the tabletop OCR2 to the disk 44 under the control of the built-in CPU41. When the data to be written on the disk 44 is image data, the CPU 41 compresses the data using the companding processor 45 to reduce the number of data and write the data on the disk 44.
Write in 4. Then, a transmission message (not shown) is created by the CPU 41 in the 0CRCTL 4, and sent to the remote office of the district center via the NCU 3, public communication network 6, and MCU
It is transmitted to the CR processing unit (MCU) 8.

At the district center, a reception processor 85 within the MCUS receives data transmitted from the sales branch and input via the MCU 7. In this device example, the receive processor 8
5 can receive up to 6 channels of input signals with a total throughput of 24 K B /sec.

The MCU 3 performs kanji character recognition on the received kanji image data and converts it into kana data.

In this process, the kanji recognition processor 86 in the MCUS recognizes the kanji image data supplied from the reception processor 85 and converts it into kana data.

Further, the CPU 81 in the MCU 3 writes the data received by the reception processor 85 onto the disk 84 . In this case, the kanji image data is written to the disk 84 as the converted kana data. The exchange form data written to the disk 84 is automatically displayed on the CRTIO via the ncu-ws 9. That is, the CPU 81 in the MCU 3
The exchange form data is distributed to the MCU-WS9 via the F87 LIF92, and the CPU91 in the MCU-WS Q
For example, as shown in the screen display example in Figure 3, the distributed data is sent to CRTIO by comparing the characters recognized by OCR for each item with the image data as images before recognition. indicate.

At the time of this display, image data that cannot be recognized by the tabletop OCR2 or the Kanji recognition processor 86 is also displayed on the CRTlO. Therefore, the operator must correct data that was incorrectly recognized by this device, read image data that could not be recognized by this device, and add the interpreted data using the keyboard. This can be done using

When displaying image data on the CRT 10, the CPU 91 uses the companding processor 95 to restore (expand) the compressed image data and display it.

When the form data is completed by correcting or adding data in this manner, the operator operates the send button on the keyboard 11 to transmit the completed exchange form data to the host device 14 via the dedicated communication line 13. .

That is, the CPU 81 in the MCU 3 uses the communication control unit 82,
Data is sent and received to and from the host device 14 via the dedicated communication line 13. This transmitted and received data is transferred to the MC.
The CPU 91 of the U-WS 9 outputs the data to the CRT 10 and printer 12 for recording.

This concludes the explanation of the operation of the money exchange data communication device as hardware using OCR according to the present invention.

Next, the OCR dictionary structure according to the present invention will be explained with reference to FIG. The dictionary shown in the figure includes a kana dictionary part which has recognition data and kana codes for recognizing all or part of kana characters, and in addition, a left parenthesis similar pattern according to the present invention. and the right parenthesis similar pattern, or the recognition data and corresponding code for either one of the parenthesis similar patterns.

This parenthesis similar pattern according to the present invention is shown, for example, in the entry format example of FIG. The left- or right-side parenthesis-like pattern in the figure is not the traditional left-hand parenthesis “(° or right-hand parenthesis)” itself, but a short horizontal line protruding from the arcuate center of the parenthesis to the left or right, respectively, and the kana character “ In order to clearly distinguish it from ノ°, we use a parenthesis-like (bracket-like) pattern that is similar to parentheses but partially modified. In the present invention, when a customer enters an abbreviation symbol (see Figure 4) in an item such as the recipient's name on a form, the customer is asked to fill in the parenthetical similar pattern, and the built-in OCR reads and recognizes this form. This parenthesis-like pattern is recognized using a dictionary, and the code for the correct left parenthesis or right parenthesis is extracted.

Furthermore, since the left parenthesis similar pattern and the right parenthesis similar pattern according to the present invention are left-right symmetrical patterns, the recognition data of either one of the parenthesis similar patterns does not need to be provided in the recognition dictionary. By providing this, it is possible to identify both left-right symmetrical patterns and to recognize both bracket-like patterns.

Note that the parentheses-like pattern in the present invention is not limited to the entry format example shown in FIG. Other bracket-like patterns may also be used. For example, a similar action can be performed with a crescent moon or a stringed bow pattern.

Next, an embodiment of the OCR recognition process according to the present invention will be described with reference to FIG. 1 and FIGS. 5 and 6.

As shown in the example of filling out a form in FIG. 2, the customer writes kana characters in the recipient's name and client's name fields on the form and submits it to the teller. An operator such as a teller sets this form on the tabletop 0CR2 and causes it to be read.

In step S10 of FIG. 1, the tabletop 0CR2 reads the data entered in the corresponding item above, and in step S
ll, the read image data is 0CRCTL
Temporarily store it in the memory in 4.

Next, in step S12, OCRCTL4 reads data corresponding to one character from the once stored image data.

In step S13, recognition processing is performed based on the recognition data of the <kana dictionary section> of the OCR dictionary shown in FIG. 5, and a corresponding kana code is searched. Then, in step S14, it is determined whether there is a corresponding kana code. If there is, the process moves to step S15, and if there is not, the process moves to step 818.

If there is a corresponding kana code, the corresponding kana code (for example, the corresponding kana code in JIS 7 units) is extracted in step S15, and this code is stored in the memory in step S20.

If there is no corresponding kana code, in step S1B, a recognition process is performed based on data for recognizing a left parenthesis similar pattern in the OCR dictionary shown in FIG. 5, and it is determined whether it can be recognized as a left parenthesis similar pattern. If the determination result can be recognized, the process moves to step S17, and if it cannot be recognized, the process moves to step S318.

If it is recognized in step 31B, step S
In step S20, the "(" code (for example, the "(" code in JIS 7 units) is extracted, and this code is sent to step S20.
, and store it in memory.

If the pattern cannot be recognized in step S1B, a recognition process is performed in step 818 based on data for recognizing a right-side parenthesis similar pattern in the OCR dictionary shown in FIG. 5, and it is determined whether it can be recognized as a right-side parenthesis similar pattern. If the determination result can be recognized, the process moves to step S19, and if it cannot be recognized, the process moves to step S21.

If it is recognized in step 31g, step S
In step S19, the ")" code (for example, the ")" code in JIS 7 unit) is extracted, and this code is sent to step S20.
Store it in memory at .

If none of the above is recognized, step 521
In this case, the corresponding character is assumed to be an unrecognized character, and this fact is stored in the memory.

In step S22, it is determined whether the recognition process for all the characters entered in the corresponding item has been completed, and if all the characters have not been recognized, the process returns to step S12 and the above process is repeated. If all characters have been completed, the process moves to step S23.

In step 823, the character code (including image data of unrecognized characters) is retrieved from the memory as the recognized character of the corresponding item, and is assembled into a transmission message together with other necessary information, as explained in FIG. It is sent to a higher-level station such as MC08 at the district center. Sonogo M CU
8 C MCU-WS9 (7) Display on the CRTIO, for example, the screen display example shown in FIG.

In the case of the money exchange processing, the abbreviations shown in FIG. 5 can be sent as they are to a higher-level station such as the host device 14. However, in the case of a system where abbreviations are not allowed in a higher-level station such as the host device 14, for example,
A correspondence dictionary (not shown) between the abbreviations and official names shown in FIG. 5 is built into the CR2, and in step S22 of FIG. After comparing and searching the code with the corresponding dictionary and converting the abbreviation part into an official name, the message may be assembled in the next step 823 and sent to the higher-level station.

[Effects of the Invention] As described above, according to the present invention, at least in OCR that recognizes katakana characters, the conventional left or right parentheses are partially changed to clearly distinguish them from katakana characters such as ")". The left parenthesis similar pattern or the right parenthesis similar pattern that was set in ``2'' is written by the customer on a form, etc., and the OCR that reads and recognizes the form etc. is equipped with a katakana character recognition dictionary to recognize katakana characters in advance. , Recognition data for recognizing the similar pattern in the left parenthesis or the similar pattern in the right parenthesis is added, so that, like the recipient column 8 of the transfer form in the gold and silver system for exchange transactions, is filled in only in kana characters, but in order to omit entry, it is possible to recognize items that allow entry in left parentheses or right parentheses, which is convenient for the customer and has the effect of shortening the entry time.

Also, for OCR, it is only necessary to add the recognition data and the corresponding code for the left or right similar pattern to the conventional katakana character recognition dictionary, so when recognizing using the conventional dictionary for recognizing English symbols, In comparison, the memory capacity of the dictionary is small, and the recognition rate can be improved by clearly distinguishing the characters from katakana characters without increasing the cost too much.

[Brief explanation of drawings]

FIG. 1 is a flowchart showing an example of OCR recognition processing according to the present invention, FIG. 2 is a diagram showing an example of form entry, and FIG.
The figure shows an example of screen display, and Figure 4 is a figure explaining the currently used abbreviations. FIG. 5 is a diagram showing an example of an OCR dictionary structure according to the present invention, FIG. 6 is a diagram showing an example of an OCR bracket entry format according to the present invention, and FIG. 7 is a money exchange data communication device using OCR according to the present invention. FIG. In the figure, 1 is a money order form, 2 is a desktop OCR, 3 is a simple keyboard, 4 is 0CRCTL, and 5.7 is an NCU. 6 is a public communication network, 8 is MCCU, 9 is MC1l
-VS, 10 CRT, 11 keyboard, 12 printer, 13 dedicated communication line, 14 host device, 41.
81.91 is the CPU. 42.82 is the communication control unit, 48.83.93 is the FD, 4
4.84.94 is a disk, 45.49 is a companding processor, 46 is CPIB, 85 is a receiving processor, 86 is a Kanji recognition processor, and 87.92 is LIF. Diagram Current Abbreviation 1 Congratulations Diagram 4r! ! J wa Ko (key in the neck 1115 showing p 2nd cause 6t 11
Shows γb bow 5th prisoner 5th prisoner

Claims (1)

[Claims] In an optical character reading device that recognizes at least katakana characters, at least one of a left parenthesis similar pattern or a right parenthesis similar pattern is included in a katakana character recognition dictionary for recognizing katakana characters. An optical character reading device characterized by being provided with additional recognition data for recognition.