WO2001013325A1

WO2001013325A1 - Document input method, recording medium recording document input program and document input device

Info

Publication number: WO2001013325A1
Application number: PCT/JP2000/002484
Authority: WO
Inventors: Masaki Nakagawa
Original assignee: Japan Science And Technology Corporation
Priority date: 1999-08-11
Filing date: 2000-04-17
Publication date: 2001-02-22
Also published as: JP2001052111A; JP3954247B2

Abstract

Erroneously recognized characters are corrected without necessarily entering characters in correcting characters. An input document (10) is image-input (S101), characters are picked up (S103), characters are recognized (S105) and recognition results are output (S107). When outputting on paper, the operator instructs correction descriptions to characters/documents short of sufficient recognition (S111) and entered correction symbols are picked up (S113). Word processing (S117), linking (S119) or separating (S117) is executed according to the picked up correction characters, a document file is corrected according to each processing (S123) and the correction result is output (S125). If a satisfactory result is not obtained (S129), the system returns to Step S111 to repeat the sequences of processing which are terminated after a sufficient correction has been made.

Description

TECHNICAL FIELD The present invention relates to a document input method, a recording medium recording a document input program, and a document input device. The present invention particularly relates to a document input method usable in the field of information devices used in offices and the like, and capable of correcting recognized characters, a recording medium storing a document input program, and a document input device. . BACKGROUND ART Conventionally, there is a document input device that inputs a document using an optical character reader (Optical Character Reader, OCR) or the like, and corrects a read result without using a keyboard and a display (for example, a special document input device). See 63-22020383). FIG. 13 shows a configuration diagram of a conventional document input device. FIG. 14 is an explanatory diagram of a conventional document input process. This document input device includes a photoelectric conversion unit 101, a character recognition unit 102, a document file storage unit 103, a printing unit 104, a correction symbol / character recognition unit 106, a correspondence unit 107, a correction unit. Equipped with 1 08. In this document input device, first, an input document 110 to be read is converted into electronic data by a photoelectric conversion unit 101, read by a character recognition unit 102, and the result is stored in a document file storage unit 103. The document file stored in the document file storage means 103 is printed by the printing means 104, and an output document 105 (recognition output 111) is output. Here, in order to correct an error on the recognition output 1 1 1, a symbol and a character are entered and a correction input 1 1 2 is created. The corrected input 1 1 and 2 were converted into electronic data again by the photoelectric conversion means 101, and entered in the corrected symbol / character recognition means 106. Read symbols and letters. Further, the correspondence between the positions where these are entered and the positions of the document files is performed by the associating means 107, and the processing according to the entered correction symbols is performed by the correcting means 108, and the correction result 1 is obtained. You can get 1 3 In this way, people who cannot use a keyboard and display can create and modify electronic documents. Disclosure of the invention

However, conventionally, it is assumed that the recognition of characters such as kanji written by hand in an output document for correction is performed correctly. In the past, in other words, what was not accurately recognized in the original recognition process was simply repeated repeatedly by the same recognition process.Accurately recognizing handwritten characters can be quite realized at present. Have difficulty. In view of the above, the present invention stores intermediate processing results such as a character image and a candidate character of a reading result, and always corrects characters by using the intermediate processing result and recognition means and correction means. The purpose is to correct misrecognized characters without writing them. Further, according to the present invention, it is another object of the present invention to apply a modification in a case where a character is entered as in the past. According to a first solution of the present invention, an output step of displaying or outputting a document file created by character recognition, and a correction recognition for recognizing a correction content instructed for a document output by the output step And a matching step of determining the correspondence between the correction content recognized in the correction recognition step and the character to be corrected; and correcting the document file according to the correction content based on the determination result by the mapping means. And a recording medium storing a document input program. According to a second solution of the present invention, a document file storage means for storing a document file created by character recognition, an output means for displaying or outputting the document file stored in the character file storage means, A correction recognizing means for recognizing the correction content instructed for the document output by the output means, and a correlating means for determining a correspondence between the correction content recognized by the correction recognizing means and the character to be corrected. And a correction means for correcting a document file stored in the document file storage means in accordance with the content of the correction based on the result of determination by the associating means. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a configuration diagram of a document input device according to the invention. FIG. 2 is a flowchart of a first embodiment of a document input method according to the present invention. FIG. 3 is an explanatory diagram relating to the writing input process. FIG. 4 is an explanatory diagram of word processing. FIG. 5 is an explanatory diagram of the combining process. FIG. 6 is an explanatory diagram of the separation processing. FIG. 7 is a flowchart of the document input method according to the second embodiment of the present invention. FIG. 8 is a flowchart of a third embodiment of the document input method according to the present invention. Fig. 9 shows the extraction and identification of correction symbols in the case of dot characters or gray color. These are the flowcharts. FIG. 10 is an explanatory diagram relating to the document processing of No. 9. FIG. 11 is an explanatory diagram of detecting a lint-out character position. FIG. 12 is an explanatory diagram of recognition of a positive sign. FIG. 13 is a configuration diagram of a conventional document input device. FIG. 14 is an explanatory diagram of the conventional document input process. BEST MODE FOR CARRYING OUT THE INVENTION FIG. 1 shows a configuration diagram of a document input device according to the present invention. This document input device includes photoelectric conversion means 1, character recognition means 2, document file storage means 3, output means 4, input means 5, correction recognition means 6, association means 7, correction means 8, intermediate processing file storage means. 9 is provided. The photoelectric conversion means 1 is composed of, for example, an OCR, reads a document to be read, and converts it into electronic data. The character recognition means 2 recognizes a character or a document from the electronic data read by the photoelectric conversion means 1. The document file storage means 3 stores the document file created by the character recognition means 2. The output means 4 displays or outputs the document file stored in the character file storage means 3. The output unit 4 is a device for printing out the output document by the printing unit, a device for displaying the output document on a display, or the like. As the output means 4, a device for interfacing with an external device, a transmission device, a drive device for various recording media, or the like can be used. The input means 5 is a suitable input device such as a pointing device such as a mouse and a touch pen, a keyboard and the like. The operator refers to the recognition result output by the output unit 4 and instructs the content of the correction. For example, in the case of an output printout by the output unit 4 such as a printer, the correction content may be indicated by inputting a correction symbol in the print result. This place In this case, the entered correction symbol is cut out by the correction recognition means 6. In the case of a display screen by the output means 4 such as a display, the instruction of the content of the correction may be made by pointing the correction content and the correction position by the pointing device or the like of the input means 5 while watching the screen. The correction recognizing means 6 recognizes correction contents (correction symbols, characters, etc.) specified for the document or display screen output by the output means 4. When a document as a recognition result is printed out by the output unit 4, the correction content is input from the photoelectric conversion unit 1. On the other hand, when the document as the recognition result is displayed on the display, the input means 5 appropriately inputs the correction content and the position. The correction recognizing means 6 can recognize not only the correction symbols but also the above-described conventional correction characters and correction portions. The correction recognition means 6 includes, for example, word processing for recognizing a plurality of characters as a word, combining processing for recognizing a character pattern for a plurality of characters as one character pattern, and a plurality of character patterns corresponding to one character. It is possible to recognize a correction symbol or a correction character when performing a separation process or the like for recognizing the character pattern as a character pattern. The associating means 7 determines the correspondence between the content of the correction recognized by the correction recognizing means 6 and the character or the position of the character to be corrected. That is, the entered correction symbol and character are associated with the processing of the output character. The correction means 8 refers to the intermediate processing file storage means 9 based on the result of the determination made by the associating means 7 when re-recognized, and refers to the intermediate processing file storage means 9 to store the document file stored in the document file storage means 3 in accordance with the content of the correction. Fix it. For example, the correction unit 8 may be configured to convert the character indicated by the position information of the document file from the candidates stored in the intermediate processing file storage unit 9 at the time of the previous recognition, according to the association method 7, for example. Modify to characters. The correction means 8 can also execute the above-described conventional correction processing. For example, in the case of word processing, the correcting means 8 recognizes a designated plurality of characters as one word, and in the case of combining processing, recognizes a character pattern for a plurality of characters as one character pattern. In the case of processing, character flutter corresponding to one character Can be separated and recognized as a plurality of character patterns, and the corresponding correction processing is executed for each. It should be noted that the present invention can be appropriately applied to other instructions for correction. When the character recognition means 2 performs recognition, the intermediate processing file storage means 9 stores processing data in the middle, that is, a character pattern and its position, a line position, a recognition result including a recognition candidate character, and the like. The character recognition means 2 stores the information in the intermediate processing file storage means 9 during the character recognition processing. When executing the character re-recognition and correction processing, the correction means 8 extracts the position information and the candidate character for the corresponding character or line from the intermediate processing file storage means 9. FIG. 2 shows a flowchart of the first embodiment of the document input method according to the present invention. FIG. 3 is an explanatory diagram related to the document input process. Here, a case where the output result is printed out on paper or the like will be described as an example. First, the input document 10 is image-input from the photoelectric conversion means 1 (S101), and the data is stored in the intermediate processing file storage means 9. The character recognizing means 2 cuts out characters from the input image (S103), and stores character or line position information, recognition candidates and the like in the intermediate processing file storage means 9. The character recognition means 2 performs character recognition (S105), and stores the recognition result in the document file storage means 3. The output means 4 outputs the recognition result (S107). Here, it is assumed that the recognition output 11 is output. In the case of paper output, the operator gives an instruction of the correction content to a character or a document that has not been sufficiently recognized (SI11). The instruction for the correction content here is made, for example, by using the correction symbol of the first symbol 12a, the second symbol 12b, and the third symbol 12c in the correction input 12 in which the symbol characters are entered. Will be In the example of the modified symbols, the first symbol 12a, the second symbol 12b, and the third symbol 12c are symbols for instructing word processing, combining character patterns, and separating character patterns, respectively. The correction recognizing means 6 cuts out the written correction symbol (S113). The extraction of the correction symbol is performed by detecting the line spacing and the character spacing of the print position, for example. Alternatively, a correction symbol inserted between characters can be detected. When the correction symbol is a color such as blue or red, the correction symbol can be detected by adding a light-receiving element or a filter that detects only that color as the photoelectric conversion unit 1. . Conversely, when the character printed by the output means 4 is a color character, the correction symbol can be similarly detected. Further, when character position information is printed on the printout by the output means 4, it is possible to recognize what is written in other positions as correction symbols. Further, when the printout is performed by the output unit 4 in dot characters or gray color, the correction symbol can be identified by the continuity and thickness of the pixels. This point will be described later. Here, if the extracted modified character is recognized as a word process, a combination of characters corresponding to this symbol in a candidate character to be recognized is determined from a combination of candidate characters that is most likely to be a word. The word processing is performed (S117). Figure 4 shows an illustration of word processing. Generally, in character recognition, each character pattern is recognized, and the first recognition candidate character is output as a recognition result. If this is determined as a word, for example, by a combination of two characters, the number of erroneous recognitions is reduced because there are restrictions on the combinations of the first and second characters that can be established as words. For example, suppose that as a result of recognizing the first character, four characters are determined as recognition candidates. That is, the first candidate is “mochi”, the second candidate is “special”, the third candidate is “samurai”, and the fourth candidate is “earning”. Also, assume that three characters are determined as recognition candidates as a result of recognizing the second character. That is, the first candidate is “fine”, the second candidate is “sign”, and the third candidate is “work”. In the recognition results for each character, the first candidates “Toku” and “Fine” were selected, respectively. However, in word processing, the user was instructed to recognize these two characters as words, so we consider combinations of each candidate for each character. For example, the combination of the second candidate “Toku” and the second candidate “Sho” is the sum of the candidate priorities 4 (2 + 2), and the fourth candidate “Earn” and the third candidate “Work” Is determined to be the sum of the candidate priorities 7 (4 + 3). Therefore, as the word processing result, the first candidate “feature” and the second candidate “operation” are determined as recognition candidates in descending order of priority (ie, in order of decreasing sum of priorities). Here, the “feature” power of the first candidate Be recognized. If it is recognized that the character pattern is to be combined, the corresponding character patterns are combined, and a combining process for recognizing one character is executed (SI 19). FIG. 5 is an explanatory diagram of the combining process. In the combining process, a character pattern (stored in the intermediate processing file storage means 9) corresponding to two or more characters specified by the correction symbol is recognized again as one character pattern. Here, an instruction to combine “T” and “Why” in the recognition result was given, and as a result of re-recognition, “Extract” was recognized. If it is recognized that the separation process of the character and the 'turn' is performed, the separation process is performed again from the character extraction for the corresponding character pattern (SI17). Figure 6 shows an illustration of the separation process. In the separation processing, a character pattern (stored in the intermediate processing file storage means 9) corresponding to the character indicated by the correction symbol is cut out as two character butters and recognized again. Here, the instruction to separate the “sum” of the recognition results was given, and as a result of the recognition again, “f” and “e” were recognized. The correction means 8 corrects the document file storage means 3 according to each process (S123), and outputs the correction result 13 (S125). If a satisfactory result is not obtained (S129), the process returns to step S111 to repeat the process. When the correction has been made sufficiently, the processing ends. Next, FIG. 7 shows a flowchart of a document input method according to a second embodiment of the present invention. This shows the operation when the recognition output 11 is output by the output means 4 by the display. The processes in steps S101 to S107 and S117 to S127 are the same as those in FIG. Steps S111 and S113 in FIG. 2 are replaced by steps S115 in this figure. In the case of display output, the contents of correction and the corresponding character position are indicated by using the input means 5 such as a mouse (S115). Therefore, the correction recognizing means 6 and the associating means 7 can directly grasp the content and position of the correction from the instruction by the mouse or the like. Further, the correction character itself can be input from the keyboard of the input means 4 or the like. FIG. 8 shows a flowchart of the third embodiment of the document input method according to the present invention. The processes in steps S101 to S107 and S115 to S127 are the same as those in FIG. Further, the processing in step S115 is the same as that in FIG. Here, a recognition output 11 1 S by the output unit 4 and a step S 109 for determining whether the output is paper output or display output are added. In the case of paper output, processing similar to that of the first embodiment is executed, and in the case of display output, processing similar to that of the second embodiment is executed. In the following, a description will be given of the extraction and identification of a correction symbol when the printout by the output unit 4 is a dot character or a gray color. Fig. 9 shows a flowchart for cutting out and identifying modified symbols in the case of dot characters or gray color. FIG. 10 is an explanatory diagram related to the document processing of FIG. The output means 4 outputs a printout of the document file stored in the document file storage means 3 in a dot character or gray color image (S201). The operator enters a symbol for correction and characters as necessary on this printout, and creates a correction input document 14 (S203). Further, the corrected input document 14 is image-input by the photoelectric conversion means 1, converted into electronic data, and stored in the electronic data storage means 6 (S205). The correction recognizing means 6 detects the position of the character in which the correction symbol is entered in the document output by the output means 4 (S207). Further, the correction recognizing means 6 detects lines and character positions from the digitized document image, noting that the printed characters are printed in dots. Details will be described later. Further, the correction recognizing means 6 recognizes the correction symbols and characters entered in the correction input document 14 (S211). The correct character to be corrected can be recognized by, for example, the character recognition means 2 or the like. The associating means 7 associates the position of the character detected by the correction recognizing means 6 with the correction symbol recognized by the correction recognizing means 6 and specifies the character to be corrected (S 2 13 ) . That is, it is associated with where the entered correction symbol or character corresponds to the output character. In this way, a symbol detection / recognition result 15 is obtained. The correction means 8 corrects the corresponding character by the correction processing corresponding to the correction symbol in accordance with the correspondence of the correspondence means 7 (S215). The correction means 8 corrects the document file stored in the document file storage means 3 according to the correction symbol recognized by the correction recognition means 6. Here, in FIG. 10, the hatched instruction indicates, for example, a symbol that replaces the corresponding character with the character written thereon, so that the correcting means 8 responds to such a character by The process of correcting “filled” into “physical” is performed on the document file storage means 3. Here, the character “rea” is recognized by the character recognition means 2 or the correction recognition means 6 as a correct character to be corrected. In addition, the correction means 8 can perform, for example, word recognition processing, combination processing, and division processing to correct the document file. The output means 4 reads the corrected document file from the document file storage means 3 and outputs a correction result 16 (S2 17). At this time, it is okay to output and confirm on the display as well as the printout. Fig. 11 is an explanatory diagram of printout character position detection. The correction recognizing means 6 detects the position of the character by multiplying the number of pixels in each of the horizontal direction and the vertical direction in the edge image of the input document. Specifically, for example, first, a contour image is created, emphasized, and on this contour image, the number of black pixels is counted in the horizontal direction to detect a row position. The character position is detected by counting the number of black pixels. In this example, the position detection of the characters printed as “this paper” on the first line and “capturing device” on the second line by the correction recognition means 6 will be described. First, a character image 41 in which a correction symbol has been entered is input. The contour of each pixel is detected based on this character image, and a contour image 42 is obtained. Further, with respect to the detected contour pixel 42, pixels are added or added in the horizontal direction to obtain a horizontal peripheral distribution 43. The position of the row can be detected from the horizontal distribution. Next, for each of the obtained rows, similarly, the detected contour pixels 44 are cut out, and the pixels are added or added in the vertical direction to obtain the vertical peripheral distribution 45. From the vertical marginal distribution 45, the character position (horizontal position) of each line can be obtained. As described above, for a certain character, the row and column are specified, and a character position detection image 46 is obtained. Furthermore, each character can be cut out by detecting the outermost pixel of each character. FIG. 12 shows an explanatory diagram of recognition of a correction symbol. The correction recognizing means 6 calculates the connected components of the pixels from the corrected input document 14 in the extraction of the corrected symbols (detection of the entered characters), and determines the connected components of the pixels in advance based on the obtained number of pixels of the connected components. By removing black connected components smaller than or equal to the size, only the modified symbols are left, and each of the remaining modified symbols is cut out. When cutting out a correction symbol, the correction recognition means 6 includes, for example, a contraction process of setting a black pixel adjacent to a white pixel as an edge and removing the edge from the original image. In this example, a description will be given of the correction symbol written in the “package” of the character printed out “of the collecting device”. First, a character image 51 in which a correction symbol has been entered is input. Next, each pixel is converted into a contracted image 52 in which contraction processing (contraction processing) is performed. This shrinking process is performed an appropriate number of times to erase the dot character, thereby obtaining a corrected symbol extracted image 53 from which the corrected symbol has been extracted. As a method of erasing a dot character, for example, a connected component which is a portion where black pixels are connected is obtained, and the number of black pixels of each connected component is obtained. Furthermore, a modified symbol can be extracted by leaving a connected component in which the number of black pixels is equal to or larger than a predetermined threshold. Next, the correction recognizing means 6 cuts out the correction symbol and recognizes what kind of correction symbol the correction symbol is to be issued, thereby obtaining a correction symbol recognition result 54. For input and output between the processes, communication such as the Internet may be used. Industrial Possibility According to the present invention, as described above, the intermediate processing results such as the character image and the candidate characters of the read result are stored, and the intermediate processing results are used and the recognition means and the correction means are used. It is possible to correct an erroneously recognized character without necessarily correcting the character. Further, according to the present invention, it is also possible to apply a modification in a case where a character is entered as in the related art.

Claims

The scope of the claims

1. An output step of displaying or outputting a document file created by character recognition, a correction recognition step of recognizing the correction content instructed for the document output by the output step, and a recognition by the correction recognition step An associating step of determining the correspondence between the corrected content and the character to be corrected, and a correcting step of correcting the document file according to the content of the correction based on the determination result by the associating means. Document input method.

2. A character recognition step of storing position information and a candidate character of a character or a line recognized based on the input document in an intermediate processing file, wherein the correction step includes the intermediate step stored in the character recognition step. 2. The document input method according to claim 1, wherein the character designated by the position information of the document file is corrected to another candidate character with reference to the processing file.

3. If the recognition result is printed out by the output step, the method further comprises a symbol cutout step of writing out a correction symbol for indicating a correction content by filling in the recognition result. Document input method described in 1 or 2.

4. The method according to claim 1, further comprising an input step for designating a correction content and a correction position on the displayed screen when the recognition result is displayed on a display by the output step. Document input method described in any of them.

5. The correction recognition step includes a word process for recognizing a plurality of characters as a word, a combining process for recognizing a character pattern for a plurality of characters as one character pattern, or a character pattern corresponding to one character. 5. The document input method according to claim 1, wherein any one or a plurality of separation processes for recognizing a plurality of character patterns is recognized.

6. In the correcting step, when the specified correction content is word processing, the specified plural characters are combined into one word, and when the specified correction content is the combining processing, the character corresponding to the plurality of characters is processed. 6. The method according to claim 1, wherein the pattern is recognized as one character pattern, and when the specified correction content is a separation process, a character pattern corresponding to one character is recognized as a plurality of separated character patterns. Document input method according to any of the above.

7. An output step of displaying or outputting the document file created by the character recognition, and a correction recognition step of recognizing a correction content specified for the document output by the output step. An associating step of judging a correspondence relationship between the amendment recognized in the amendment recognizing step and a character to be amended, and an amendment step of amending a document file according to the amendment based on the judgment result by the associating means. A recording medium on which a document input program comprising and is recorded.

8. Document file storage means for storing a document file created by character recognition, output means for displaying or outputting the document file stored in the character file storage means, and for a document output by the output means. A correction recognizing means for recognizing the instructed correction content, an associating means for judging a correspondence relationship between the correction content recognized by the correction recognizing means and a character to be corrected; and A document input device comprising: a correction unit that corrects a document file stored in the document file storage unit according to the content of the correction.

9. Further provided is an intermediate processing file storage unit that stores position information and candidate characters for characters or lines recognized based on the input document, and the correction unit refers to the intermediate processing file storage unit, 9. The document input device according to claim 8, wherein the character designated by the position information of the document file is corrected to another candidate character.