JPH03222080A - Character pattern recognition circuit - Google Patents

Abstract

PURPOSE:To reduce the number of time of correction operation to the result of an erroneous recognition by providing a picture pattern designating means for erroneous recognition, a correction information input means, a storing means, a comparison deciding means, and a picture processing means. CONSTITUTION:For example, it is assumed that a character 'I' (Japanese sylla bary) is inputted, and two characters 'SHI' (Japanese syllabary) and 1 are erroneously recognized. A correction is ordered to a CPU 1 by a key 5. This pattern is temporarily stored in a RAM 3, the correct input 'I' is received, a picture pattern of an instructed range is transferred to a working area, and an erroneous recognition data case number information becomes +1. Finally, the character pattern of 'SHI' and 1 and the correct 'I' are respectively stored in an external storage device 7 as patterns 1, 2, and the registration of the erroneous recognition/correction pattern is ended. For the character recognition after this, the correct recognition processing is executed for a similar picture is executed based on the registered erroneous recognition pattern and the correc tion pattern.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a character recognition circuit that recognizes characters from image information read from a document.

[Conventional technology]

Currently, a character recognition circuit extracts a character image to be recognized by determining from where a single character's worth of character image starts from within one screen's worth of read image information. Therefore, for example, if the character image ``i'' extends beyond the extraction frame for one character and is extracted as two characters, the character image ``i'' is extracted as two characters, such as ``J'' becomes the letter ``J'' and ``1''. Misrecognition may occur. In this case, the operator inputs the correct character code and the character recognition circuit executes character processing in which the incorrectly recognized character code is replaced with the correct character code.

[Problem to be solved by the invention]

For this reason, the operator has had to repeat the above-described correction operation for the erroneously recognized image pattern every time image information is read from the document.

SUMMARY OF THE INVENTION In view of the above-mentioned problems, an object of the present invention is to provide a character recognition circuit that can reduce the number of correction operations performed by an operator in response to erroneous recognition results.

[Means to solve the problem]

In order to achieve such an object, a first aspect of the present invention provides an instruction means for instructing the erroneously recognized image pattern when an erroneous recognition occurs in character recognition of an image pattern read from a document; an input means for inputting correction information for correcting the erroneously recognized image pattern to a correctly recognizable image pattern; and an input means for inputting correction information for correcting the erroneously recognized image pattern to an image pattern that can be correctly recognized; a storage means for storing information; a determination means for determining by pattern comparison whether or not there is an image pattern matching the image pattern to be recognized in the storage means during the next character recognition; and an image processing means for reading correction information corresponding to the matching image pattern from the storage means and correcting the image pattern to be recognized using the read correction information when the judgment result is affirmative. Features.

A second aspect of the present invention provides a storage means that stores in advance a first image pattern to be converted and a character code string after conversion of the first image pattern, and a first image pattern that matches the second image pattern to be recognized. a determining means for determining whether or not is present in the storage means by pattern comparison; and when the determination result of the determining means is affirmative, character recognition is performed for a character code string corresponding to the matching first image pattern. and reading means for reading out the result from the storage means.

A third aspect of the present invention is to set a recognition result in a character recognition circuit that calculates the distance between a registered image pattern whose character code is known and an image pattern to be recognized for character recognition. determining means for predetermining a threshold value for the distance and determining whether or not the calculation result for the distance is within the range of the threshold value by comparing the predetermined threshold value and the calculation result for the distance; a first storage means for storing the character code of the registered image pattern having the shortest distance as a recognition result when the determination result of the determination means is an affirmative determination;
The present invention is characterized by comprising second storage means for storing the image pattern to be recognized as it is as a recognition result when the determination result of the determination means is a negative determination.

A fourth aspect of the present invention provides a character generating means for converting the character code stored in the first storage means of the third aspect into a character pattern in a predetermined font, and a character pattern generated by the character generating means. and image processing means for combining the image patterns stored in the second storage means into one piece of document information by associating the character positions thereof.

A fifth aspect of the present invention is characterized in that it further includes display means for displaying the image pattern stored in the second storage means of the third aspect.

[Function] In the first embodiment of the present invention, for example, an image pattern that causes an error in character extraction is stored together with correction information. Next, when performing character recognition of a similar image pattern, if the determining means determines that the image pattern to be recognized matches the image pattern in the storage means, the image processing means can modify the image pattern to be recognized based on the modification information. This is done by Therefore, an image pattern for which correction information has been input once will not be erroneously recognized. As a result, the number of correction operations performed by the user is reduced.

In the second embodiment of the present invention, a specific figure is detected by the determination means and the corresponding character code string is read out from the storage means, so that not only can the figure be converted into a character string, but also the recognition process for each character can be performed. and there is no need to perform correction processing,
Recognition processing time is also reduced.

In the third embodiment of the present invention, when the result of distance calculation performed in character recognition exceeds a threshold value, the determination means determines that recognition is impossible, and the image pattern to be recognized is set as the recognition result, so that the recognition circuit side can prevent erroneous recognition. It never occurs.

In the fourth embodiment of the present invention, a character pattern with a fixed font is created by the character generation means based on the character recognition result, so it is possible to clean up handwritten characters, and unrecognized characters can be Since the characters are synthesized as they are, there will be no misrecognition.

In the fifth embodiment of the present invention, the unrecognizable result is displayed as a character image to be recognized on the display means, so that the user can easily know the character to be corrected.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

Lost in summer 1 to 5 show a first embodiment of the present invention.

FIG. 1 shows the circuit configuration of a first embodiment of the present invention.

In FIG. 1, l denotes a central processing unit (CPU), which not only controls the entire device but also extracts character images from image information and performs character recognition. 2 is a ROM in which various programs executed by the CPUI are stored. C.P.
The UI operates as the determination means 1 image processing means in the first to second embodiments of the present invention.

3 is a RAM used to store CPUI input/output;
4 is an image reading device (OCR) that reads an image from a document and outputs image information in the form of a dot pattern. 5
is a keyboard (instruction means, input means of the first form of the present invention) for inputting information such as operation instructions to the CPUI,
6 is a display device that displays input information from the keyboard 5 and processing results of the CPUI.

7 is an external storage device for storing the standard image pattern used for character recognition and the read image pattern (first to second of the present invention);
(a third form of storage means).

8 is a printer that prints information, and 9 is a system bus that commonly connects each of the above circuits.

The operation of such a circuit according to the present invention will be explained with reference to FIGS. 2 to 5.

FIG. 2 shows a data format for storing image information when erroneously recognized. FIG. 3 shows an image to be recognized. FIG. 4 shows a control procedure for delimiter correction executed by the CPUI. FIG. 5 shows a control procedure for character recognition executed by the CPUI.

When the 0CR4 reads the image of the original shown in Figure 3, the CP
LII performs character recognition according to the control procedure shown in FIG.

At step 52-1 in FIG. 5, one screen worth of image information is received from 0CR4. Next, in step 52-2, one character extraction processing is performed. Thereafter, according to the processing procedure of steps 52-3 to 52-7, the standard character image pattern registered in the external storage device 7 is compared with the read character image pattern to convert each character image into a character code.

As a result, input the letter "i" in Figure 3.
” is recognized as two characters.

Therefore, the operator selects the correction mode from the keyboard 5.
Instruct put. Based on this instruction, the CPU executes the control procedure shown in FIG. 4.

At step S1-1 in FIG. 4, the read image is displayed on the display device 6. Next, when the operator inputs the position of the erroneous character "shi" (b in Figure 3) and the next character "1" (C in Figure 3), the CPU , are stored in the work area of RAM3.

The procedure then proceeds to step 51-2, where the operator inputs the correct character range for the character "i" (F in FIG. 3) to be corrected. In step 5l-3, the image pattern in the designated range is transferred to the work area. Subsequently, in step 51-4, the information on the number of misrecognized data shown in FIG. 2 is incremented by 1.

Finally, in step 5l-5, according to the format shown in FIG.
The character pattern and the correct character pattern ("i") instructed in step 51-3 are stored in the external storage device 7 as patterns 1 and 2, respectively. This completes the registration of the erroneous recognition pattern and correction pattern. In the next character recognition, based on the registered erroneous recognition pattern and corrected pattern, correct recognition processing is performed for an image substantially similar to the image shown in FIG. 2 in which the erroneous recognition occurred.

That is, when the 0CR4 reads the image of the document shown in FIG. 3, the CPUI performs character recognition according to the control procedure shown in FIG. 1 from 0CR4 in step 52-1 of FIG.
Receive image information for a screen. Next, in step 52-2, one character extraction processing is performed.

In step 52-3, a co-op that counts registered data is used.
Clear unt to O. Subsequently, in step 52-4,
The count-th corrected registration data is read from the external storage device, and pattern matching is performed on the registered erroneous recognition pattern l and the pattern extracted from the original. Here, if both patterns match, step 52-6
Go to and replace the extracted pattern with the registered modified pattern 2.

Next, the process proceeds to step 52-7, where character recognition processing is performed on the modified pattern. Finally, in step 52-8, it is determined whether the processing has progressed to the end position of the document, and if the determination is affirmative, the control procedure is ended. If the character image to be recognized is not at the end position of the document, the process returns to step 52-2 and repeats the above-described procedure.

As a result of the pattern matching in step S5, if the two patterns do not match, the process goes to step 52-9, and whether or not pattern matching has been performed for all of the misrecognized patterns 1 is determined based on the number of registrations and the execution of pattern matching. Check by comparing the number of times. If pattern matching has been performed for all of the misrecognition patterns l, it is determined that the image pattern extracted from the document is not registered as a misrecognition pattern, and the process goes to step 52-7.
Character recognition is performed using patterns extracted from the original. If the determination result in step 52-9 is negative, the count is increased by 1 in step S2-10, and step 52-4
Return to. Pattern matching with the next misrecognized pattern l is continued.

By executing this procedure, when a character image "i" similar to the character image "i" in Figure 2, which was misrecognized the first time, is extracted from another manuscript, the extracted character image is be recognized correctly.

The second embodiment is an example in which an image of a specific shape in an image extracted from a document is detected and converted into a pre-registered character code string. The operation of the circuit shown in FIG. 1 in this example will be explained below according to the flowchart shown in FIG.

First, in step 53-1, a process selection screen for pattern registration or character recognition is displayed on the display device 6. First, when the operator selects pattern registration from the keyboard input device 5, the procedure moves from step 53-2 to step S3-14.

In step S3-14, a document image is received.

As shown in FIG. 7, characters or graphic patterns to be registered are written on this manuscript. In step S3-15, an image (the length of a character or graphic pattern and a dot pattern) surrounded by a rectangle is detected. Next, the process goes to step S3-16, and the detected character pattern is changed to -
Store it in the work area at the time. Next, on the display device 6,
Display guidance for entering the character code string to be converted.

In step S3-18, input of a character code string from the keyboard human input device 5 is accepted. Finally, the length and pattern of the character pattern stored in the work area and the input character code string and its length are stored in the external storage device 7 according to the data format shown in FIG. Furthermore, the number of registered items is increased by 1.

On the other hand, if pattern recognition is selected in step 53-2, the procedure advances to step 53-3. Step 33-3
Now, read the original image, and then in step 53-4
Perform character extraction processing.

In step 53-5, registered patterns are counted.
Clear ount to 0. Next, step 53-6
Now, from the registration data stored in the external storage device 7, co
The unt-th data is read out, and pattern matching is performed between the character pattern and the pattern extracted in step 53-4.

Here, if both patterns match, the procedure is changed to step 53.
The process advances from step 53-8 to step 53-8, and the character code string stored in the count number in the external storage device 7 is read out. Next, the read character code string is outputted from the printer 8 as a result of character recognition. Next, in step 53-9, the end of character recognition is determined based on the reading position of the image.

If the image reading position is the end position of the document, this procedure ends. If the image reading position is not the end position,
The process returns to step 53-4 and the character recognition process is repeated.

On the other hand, as a result of pattern matching, if it is determined in step 53-7 that the two patterns do not match, the process proceeds to step S3-10, and it is determined whether or not the pattern check for the number of registered items has been completed.

When pattern checking has been completed for all registered patterns, the process goes to step S3-11 and normal character recognition processing is performed.

Thereafter, in step 53-9, it is determined whether the document is finished. If the number of registered items is not reached in step S3-10, it is determined that there are remaining patterns to be checked, the count is incremented by 1 in step S3-12, and step 53-
Return to step 6 and repeat the pattern check for pattern conversion.

By executing this procedure, it becomes possible to convert a specific image pattern into a specific character code string.

This example is an example in which when character recognition is performed on a read image pattern, if the read image pattern is unrecognizable or has a low recognition probability, the read image pattern is output as is.

FIG. 9 shows a schematic configuration of the apparatus of this embodiment. 101 in the diagram
is a CPU (determination means according to the third embodiment of the present invention, image processing means according to the fourth embodiment) that controls the entire apparatus, and operates according to a program stored in a program area 102a in the memory 102. Note that a program shown in the flowchart of FIG. 1O is stored in the memory 102a.

103 is a keyboard input device, 104 is a CRT display device (display means of the fifth embodiment of the present invention), and 105 is a character pattern generator capable of generating a plurality of character patterns in correspondence with character codes. 106 is CRT104
It is a transparent pressure-sensitive touch panel located on the front of the computer, and the positional information touched by the operator's fingertip is taken into the CPU10I.

107 is a printer that outputs image information, etc.;
8 is a telephone line, and 109 is a telephone set. 110 is an NCU device that switches telephone lines, etc.; 111 is a signal modulation device;
This is a modem that performs demodulation. 112 is an external storage device for storing data (first storage means and second storage means of the third embodiment of the present invention); 113 is an F for reproducing encoded information;
It is an M sound source. 114 is an AD that encodes and plays audio
A PCM device 115 is an image reading device that reads images.

FIG. 1O shows the control procedure when creating a text file.

When creating a text file, characters are written on paper in a grid-like fixed format as shown in FIG. 11, and the image reading device 115 reads the character image line by line. The CPU 101 performs character recognition processing on this read image according to the processing procedure shown in FIG. 10 to convert the character image into a character code, thereby creating a document file.

In FIG. 1O, after reading one line of image information, the CPU 10I performs character recognition character by character. In this example, the distance between the standard image pattern for each character registered in advance and the image pattern to be read is calculated, and the character code of the standard image pattern with the shortest distance and shorter than a certain distance and its character position are recognized as the result. The data is stored in the external storage device 112 as (steps 54-1 to S4-3).

Further, if there is no standard image pattern that falls within a certain distance, it is determined that character recognition cannot be performed correctly, and the image and position are stored in the external storage device 112 (step 54).
-2 to S4-4). By repeating the above process for all lines, a text file in character code form is created from the image read from the original.

Furthermore, for characters in image format, it is preferable to insert a control code indicating this at the character position.

The created text file is displayed on the display device 105 or the printer 10.
6, the recognized character code in the document file is converted in advance into a corresponding character pattern, for example, in the Sui font, by a pattern generator 105. Output the converted character pattern to the output device.

The control code indicating that the image pattern in the document file is stored is stored in the external storage device 11 by the CPU10I.
2, the image pattern of the character position is read out and directly transmitted to the output device.

By performing such processing, the user can confirm the character recognition results. Furthermore, a document file of all character codes is created by instructing correct character codes for the characters of the image pattern and performing the correction process of the first embodiment.

The created document file is NCLI in the form of character code.
III to an external device.

Natsudoshi 1 This example performs character recognition on a character image for reading,
This is an example in which the character code is converted into a character code, the converted character code is reconverted into a character pattern of, for example, the Sui script by the pattern excavator 105, and the neatly written character image is sent to an external device.

FIG. 12 shows the execution control procedure of CPU10I for this process. When transmitting after character recognition, the image reading device 115 reads a character image written on a sheet of paper in a grid-like fixed format as shown in FIG. 11 line by line. After reading each line, character recognition is performed one by one, and for those judged to be correctly recognized, the character code and position are stored in memory, and for those that are likely to be misrecognized, the image and position are stored in memory (step 55-1-55-3
, 55-4).

When the recognition of the characters on the l line is completed, the characters that have been converted into character codes are reconverted into image information by the character pattern generator 115. Character images that cannot be recognized are used as is.

In this way, it is combined with the image, converted into MH code, and stored in the external storage device 112 (step 55-5.S5
-61° Repeat this process for all lines to create a file converted into one page of MH code. Thereafter, dialing is performed in the conventional manner, and the MH file is transmitted via the telephone line via the start protocol, and after the end protocol is transmitted, the telephone line is disconnected to end the transmission.

As described above, in this embodiment, a read handwritten character image can be converted into a font character image and transmitted to an external device, so that the output result on the receiving device side is easy to see.

FIGS. 16 and 17 show specific processing steps for editing a document file using unrecognizable character images and recognized character codes. In FIG. 16, as a result of recognition processing performed on the character image as shown in FIG. 13 in steps 56-1 and 56-2, if the characters can be recognized, the character code of the recognition result is 'A', 'I', 'U', and 'O' are set in the external storage device 112 in the order of recognition as shown in FIG.

However, if it is determined in step 56-3 that the character cannot be recognized, the 14th
An unconfirmed code corresponding to the figure in the diagram is set. Subsequently, in step 56-5, the position of the unconfirmed character (in this case, row 9 column) and the storage address of the unconfirmed image data are set in the external storage device 112 as shown in FIG.

Furthermore, unconfirmed image data is stored at the set storage address (step S6-6).

Such a processing procedure is repeatedly executed, and when it is completed for one document, the process moves on to editing the next document file.

FIG. 17 shows a processing procedure for correcting characters during editing.

In step 57-1, the cursor is moved to the character desired to be corrected. If the character is an unconfirmed character, the storage address of the corresponding image data is searched and the image data of that character is displayed (step 57).
-2 to S7-4).

The operator re-enters the characters from the keyboard device 103 while looking at the image (step S7-5).
). If editing is to be continued, the process returns to step 57-1 and the cursor is moved again to continue editing.

[Effects of the Invention] As explained above, in the first embodiment of the present invention, similar erroneous recognition can be prevented by registering erroneous recognition patterns and referring to them at the time of recognition, which improves operability and practicality. It has the effect of improving sex.

In the second embodiment of the present invention, an image pattern and a character string are registered, and the registered pattern is converted into a registered character string during character recognition, thereby maintaining the ease of handwriting input and reducing the tediousness. can only be reduced.

According to the third aspect of the present invention, when handwritten characters are recognized and a text file is created, characters that are likely to be misrecognized are stored as images, making it possible to create a more accurate text file.

According to the fourth aspect of the present invention, since handwritten characters are converted into a font and transmitted, the quality is high, and since characters that are not recognized are output as they are in the image, it is possible to transmit information with accurate content. .

In the fifth embodiment of the present invention, by storing images of unrecognized characters and displaying the images when editing, there is no need to make corrections while comparing with the original.
Misrecognized characters can be corrected easily and quickly, which has the effect of improving operability.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing the circuit configuration of the embodiment of the present invention, Fig. 2 is an explanatory diagram showing the contents of the registered data format of the embodiment of the invention, and Fig. 3 is an example of an erroneous recognition pattern of the embodiment of the invention. FIG. 4 is a flowchart showing a control procedure for delimiter correction in an embodiment of the present invention; FIG. 5 is a flowchart showing a control procedure for character recognition in an embodiment of the present invention; FIG. 6 is a flowchart showing a control procedure for character recognition in an embodiment of the present invention. An explanatory diagram showing an example of the registration data format of the second embodiment, FIG. 7 is an explanatory diagram showing an example of the registration pattern of the second embodiment of the present invention, and FIG. 8 is an explanatory diagram showing an example of the registration pattern of the second embodiment of the present invention. FIG. 9 is a block diagram showing the circuit configuration of the third embodiment of the present invention. FIG. 10 is a flowchart for creating a document file according to the third embodiment of the present invention. An explanatory diagram showing an example of the standard format of the third embodiment, FIG. 12 is a flowchart showing a control procedure for information transmission in the third embodiment of the present invention, and FIG. An explanatory diagram showing an example; FIGS. 14 and 15 are explanatory diagrams showing a buffer structure for storing recognized characters and image data according to a fourth embodiment of the present invention; FIGS. 3 is a flowchart illustrating an example processing procedure. 1.101. ...cpu. 2・・・・・・・・・R where the program is stored
OM, 3......RAM, 4.1
15... Image reading device, 5.103...
・Keyboard, 6.104...Display device, 7.112...External storage device, 8.107...
...Printer, 9...System bus, 102...
......Memory, 102a...Program area, 105
・・・・・・Character pattern generator, 106...
...Touch panel, 108...
Telephone line, 109...Telephone, 110...NCU. 111...Modem, 113...FM sound source, 114...ADPCM. Figure 2 Figure Figure Figure Figure Figure 10 Figure 11 Figure 13-1 Figure 4-2 Figure 15 Figure 16 Figure 17

Claims (1)

[Scope of Claims] 1) When an erroneous recognition occurs in character recognition of an image pattern read from a document, an instruction means for instructing the erroneously recognized image pattern, and capable of correctly recognizing the erroneously recognized image pattern. an input means for inputting correction information for correcting the image pattern to a correct image pattern; a storage means for storing the misrecognized image pattern instructed by the instruction means and the correction information input from the input means; determining means for determining, by pattern comparison, whether or not there is an image pattern in the storage means that matches the image pattern to be recognized during character recognition; A character pattern recognition circuit characterized in that it comprises an image processing means for reading correction information corresponding to a matching image pattern from the storage means and correcting the image pattern to be recognized using the read correction information. 2) A storage means that stores in advance a first image pattern to be converted and a character code string after conversion of the first image pattern, and a first image pattern that matches the second image pattern to be recognized is stored in the storage means. determining means for determining whether or not the matching first image pattern exists by pattern comparison; and when the determination result of the determining means is affirmative, the storage means stores a character code string corresponding to the matching first image pattern as a result of character recognition. A character pattern recognition circuit comprising: reading means for reading from a character pattern; and a character pattern recognition circuit. 3) For character recognition, in a character recognition circuit that calculates the distance between a registered image pattern whose character code is known and an image pattern to be recognized, a threshold value of the distance for setting the recognition result is set in advance. determining means for determining whether or not the calculated distance is within the range of the threshold by comparing the determined threshold with the calculated result of the distance; In the case of a positive determination, a first storage means stores the character code of the registered image pattern having the shortest distance as a recognition result, and in the case of a negative determination by the determination means, the image pattern to be recognized is directly stored as a recognition result. A character pattern recognition circuit comprising a second storage means for storing information. 4) character generation means for converting the character code stored in the first storage means into a character pattern with a predetermined font; 4. The character pattern recognition circuit according to claim 3, further comprising image processing means for combining the image patterns into one piece of document information by associating the character positions of the image patterns. 5) Claim 3, further comprising display means for displaying the image pattern stored in the second storage means.
Character pattern recognition circuit described in .