JP2002279353A - Character recognition device, method therefor, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character recognition device, improving the working efficiency of postprocessing of a character recognition result. SOLUTION: This character recognition device recognizes characters relative to a character region of an inputted image, calculates sureness of the character recognition result of the respective characters, searches characters in the front/ rear of a character, having low sureness of the recognition result relative to the respective characters, when post-processing the character recognition result, and corrects the word so as to improve the working efficiency of the postprocessing of the user.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device, a method thereof, and a recording medium, and more particularly to post-processing of a recognition result after a character recognition processing result.

[0002]

2. Description of the Related Art Optical character reading (OCR) software for identifying a character image in document image data taken into a computer from a scanner or the like and outputting it as a character code has been widely used in recent years. This is because information that has been converted into electronic data as a character code has the advantage that it can be easily reused and stored and exchanged more efficiently than paper-based information and image data. However, in character recognition, it is extremely difficult to completely eliminate erroneous recognition by character-by-character recognition from a character image. For this reason, as a post-processing, a character string resulting from character recognition is often subjected to linguistic constraints to select a candidate, or automatically corrected by word matching, morphological analysis, or the like. . Although this method can eliminate many errors, depending on how it is applied, it may create new errors due to erroneous correction. In addition, since this correction does not result in 100% correction, a recognition result having a higher degree of certainty may be adopted as compared with a recognition result obtained by another character cutout method (for example, see Japanese Patent Application Laid-Open No. 9-274645). For example, a certainty factor is assigned to each character to notify the user and leave the correction to the user. For example, as a post-processing method of character recognition, the technology disclosed in Japanese Patent Application Laid-Open No. 5-40853 creates candidate words and calculates certainty for each word of the character recognition result, and determines whether or not the word has a large certainty. It corrects the recognition result of a word with low certainty using grammatical relations. As the calculation of the certainty factor, for example, as in the technique of Japanese Patent Application Laid-Open No. 9-134410, the language processing prior to the certainty factor calculation, the word notation length, the part of speech, the appearance frequency, the strength of the connection with the preceding word, etc. It is calculated by combining parameters.

[0003]

However, even if language processing is performed using a word dictionary and grammatical knowledge as post-processing of character recognition, grammatical errors or unrecognizable areas may occur.
It is not expected to improve the results in grammatically specific areas. Eventually, the result of presenting the character recognition result to the user and correcting the result will calm down. As a technique for assisting the correction of the recognition result presented to the user, for example, in Japanese Patent Application Laid-Open No. 5-46803, a recognition result having a low likelihood of the recognition result is shaded, and further distinction can be made according to the character type. By displaying the identifier at the same time,
It is easy for the user to correct the recognition result for characters similar to katakana “ri” and hiragana “ri”. Also, in order to easily correct the misrecognition result, Japanese Patent Laid-Open No.
In Japanese Patent Application Publication No. 176189, a total of three character candidates including a character whose recognition result is determined to be suspicious and one character before and after the character are displayed, and Japanese Patent Publication No. Hei 7-113951 and Japanese Patent No. 29154.
In Japanese Patent Publication No. 17, a recognition result and an original image are simultaneously displayed. However, in the above-mentioned conventional technology, since the display method mainly relies on visual information, it is impossible for a visually impaired person or the like to determine which character is an uncertain recognition result. SUMMARY OF THE INVENTION The present invention provides a character recognition apparatus, a method and a recording method for correcting a recognition result using visual information, auditory information, tactile information, and the like to improve the efficiency of post-processing. The purpose is to provide a medium. In particular, it is an object of the present invention to provide a character recognition device, a method thereof, and a recording medium that improve the work efficiency of post-processing after character recognition even for a visually impaired person.

[0004]

According to a first aspect of the present invention, there is provided a character recognition apparatus for obtaining a recognition result by performing post-processing on a character recognition result of a document image. An image input unit for inputting a document image, a character recognizing unit for recognizing a character in a character area of the image input by the image input unit, and a certainty of a character recognition result for each character obtained from the character recognizing unit. And a result output unit that post-processes the character recognition result obtained from the character recognition unit and outputs the result, wherein the result output unit performs post-processing of the character recognition result. Searching for characters before and after a character with low probability of recognition result for each character calculated by the certainty factor calculation unit, extracting a word including a character with low probability of recognition result, and post-processing this word What I did And it features. Further, in the character recognition device according to claim 2 of the present invention, in the character recognition device according to claim 1, the result output unit does not determine a word whose probability of recognition result is determined to be low and a word whose probability is low. The word is visually distinguished and output, and the user can easily perform post-processing. According to a third aspect of the present invention, there is provided the character recognition device according to the first aspect,
The result output unit is characterized in that a word determined to have low probability of the recognition result and a word not determined to be low are audibly distinguished and output to facilitate post-processing by the user. Also, in the character recognition device according to claim 4 of the present invention, in the character recognition device according to claim 1, the result output unit does not determine a word whose probability of recognition result is determined to be low and a word whose probability is low. The word is tactilely distinguished and output, so that the user can easily perform post-processing.

According to a fifth aspect of the present invention, there is provided the character recognition apparatus according to any one of the first to fourth aspects, wherein the result output unit corrects an uncertain word. In some cases, the uncertain word is replaced by the corrected word, and the process proceeds to post-processing of the next uncertain word. In a character recognition method according to a sixth aspect of the present invention, in a character recognition method for performing character recognition on a character image and performing post-processing on the recognition result, a character before and after a character having a low probability of the recognition result is searched. Then, a word including a character having a low probability of the recognition result is extracted, and the word is corrected so that the post-processing work efficiency is improved.
Further, a recording medium according to claim 7 of the present invention is a computer-readable recording medium recording a program for causing a computer to function as a character recognition device that performs post-processing on a character recognition result of a document image to obtain a recognition result. An image input unit for inputting a document image, a character recognition unit for recognizing characters in a character area of the image input by the image input unit, and a character recognition result for each character obtained from the character recognition unit And a result output unit for post-processing the character recognition result obtained from the character recognition unit and outputting the result, wherein the result output unit is configured to output the result after the character recognition result. At the time of processing, search for characters before and after a character with low probability of recognition result for each character calculated by the certainty factor calculation unit, and extract a word including a character with low probability of recognition result. They were recorded character recognition program to be worked up words.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The construction and operation of an embodiment of the present invention will be described below in detail with reference to the drawings. Embodiment (1) Configuration of Embodiment FIG. 1 is a block diagram illustrating a configuration of a character recognition device according to an embodiment of the present invention. Character recognition device 100 according to an embodiment of the present invention
Are the control unit 10, the image input unit 20, the character recognition unit 30, the certainty calculation unit 40, the result output unit 50, the original image storage unit 25,
It comprises a recognition dictionary 35, a recognition result storage unit 45, and a language dictionary 55. Further, the result output unit 50 includes a low confidence extractor 60. The control unit 10 controls the entire process from reading an image from a scanner or a file, recognizing characters from the image information, and outputting a final recognition result. The image input unit 20 reads an image from a scanner or a file, and stores the image data in the original image storage unit 25. The character recognizing unit 30 determines a character image area from the image information stored in the original image storage unit 25, cuts out a line from the character area, cuts out a character from the cut out line, and sets a rectangular pair surrounding the character part. The angular coordinate values are extracted, the size of the character portion is normalized, noise (dirt, etc.) is removed, the feature amount is calculated, and pattern matching is performed between the feature amount and the recognition dictionary 35 holding the standard pattern. The recognition result storage unit 45 stores one or more recognition candidate characters per character, their order, and the distance value between the character and the standard pattern corresponding to the character. The recognition dictionary 35 holds information such as a character code for each character and a feature amount of a standard pattern of the character.

Next, the character recognizing unit 30 generates a word candidate starting from each character position, searches the word dictionary of the language dictionary 55, and extracts the matched word and its part of speech information.
A candidate word is connected from the beginning of the processing target area to generate a word path, and at the same time, the cost of the word path is calculated using the part-of-speech connection cost table of the language dictionary 55. The word paths are selected in descending order of cost so that the number of generated word paths is equal to or less than a certain number, the recognition candidate character is corrected based on the word path with the smallest cost, and the recognition result storage unit 45 is updated. . The language dictionary 55 is a notation of words,
It consists of a word dictionary that holds information such as reading and part of speech, and a part-of-speech connection cost table that holds weights that indicate whether or not the part of speech of connected words can be connected. For example, in the case of Japanese, in addition to words containing only kanji,
"Kanji + Kana" and "Kanji + English" can also be registered. The certainty calculating unit 40 assigns numerical values (weights) that meet the following conditions to the parameters for calculating the certainty, and calculates the certainty as a linear combination or an average value of these parameters (for example, JP-A-9-134410). (A) The higher the character is in the recognition result for the same character in the image, the higher the certainty is. (B) The higher the similarity in the character recognition of the character,
Confidence is high. (C) The longer the written length of the word determined to belong to the character, the greater the certainty. (D) The greater the likelihood of connection between the character or the category of the character and that of the preceding and following characters, the higher the certainty. (E) The degree of certainty increases as the possibility of connection between the word and the preceding and succeeding words increases. This rule is stored in advance as a statistical property or a heuristic rule of many documents in, for example, a rule table or a part-of-speech connection cost table. The result output unit 50 outputs the recognition result stored in the recognition result storage unit 45 and the corresponding original image data in order to perform post-processing. By referring to the output result, the user can correct the recognition result if it is incorrect by referring to the output result, output the final result to an output device such as a display, a printer or a file, or transmit the final result to another computer via a network. Or

In the post-processing, in the conventional method, one character "single character" is input when the correct character is input in order to correct only the incorrectly recognized character. However, the user cannot easily select the correct character in the case of kanji because there are many single kanji of the same sound. In many cases, it is more efficient to correct not only the characters but also the correct characters before and after it, in units of words. On the other hand, the fact that there is a word including a character with a low degree of certainty in the recognition result calculated by the certainty degree calculating unit 40 often means that matching with the language dictionary 55 could not be achieved. In such a case, the low confidence extracting unit 60 first searches the word path for a character having a low confidence in the recognition result, and the character is registered as two or more words before and after the detected recognized character. Non-adjacent single characters are searched for in a word path, concatenated as one character string, and the character string is extracted as one word (hereinafter, this character string is also called a word in the present invention). For example, referring to FIG. 2, as a result of character recognition of image data of “peripheral image display processing”, a character string of “peripheral image wheat display processing” is obtained, and the word paths are “peripheral”, “image”, “wheat”. , "Show", and "process" are obtained. The confidence for each character is “90”, “89”, “80”, “75”, “7”
It is assumed that the calculated values are “0”, “80”, “50”, “80”, “99”, and “92”. Since "barley" with a low confidence of "50" is detected, when searching forward and backward from "barley", there is a word "image" in the forward direction and a single character "indicator" in the backward direction. Since there is a word “processing”, “barley” is extracted as a word with low probability.

Next, when performing post-processing of the recognized result, the result output unit 50 outputs the low confidence word extracted by the low confidence extraction unit 60 as described above and the high confidence word. Are displayed separately. For example, as a display method, a color, a font size, an underline, and a wink are displayed so that the characters can be visually distinguished. In this post-processing, when one word is corrected, the recognition result storage unit 45 is updated with the corrected word, and then the process moves to the next word having a lower degree of certainty. As a result, the user can efficiently correct erroneous recognition in word units without releasing the hand from the keyboard. By configuring the embodiment in this manner, a word having a low degree of certainty is displayed in units of words, instead of being corrected in units of characters as in the related art, and correction is performed in units of words. This makes it easier to correct the correct word and improves the efficiency of editing work.

(2) Processing Procedure FIG. 3 is a flowchart showing the processing procedure of this embodiment. An image from a scanner or a file is read, and the image data is stored in the original image storage unit 25 (step S10).
0). Thus, the image input unit 20 is configured. The character image area is determined from the image information stored in the original image storage unit 25, and a line is cut out from the character area (step S11).
0). A character is cut out from the cut-out line, diagonal coordinate values of a rectangle surrounding the character portion are extracted, the size of the character portion is normalized, noise (dirt, etc.) is removed, and a feature amount is calculated (step S120). ). Pattern matching is performed from the feature amount and the recognition dictionary 35 that holds the standard pattern, and one or more recognition candidate characters per character, their rank, and the distance value between the corresponding standard pattern and the recognition result storage unit 45 are stored. (Step S13)
0). Based on the recognition candidates registered in the recognition result storage unit 45,
A word candidate starting from each character position is generated, a word dictionary of the language dictionary 55 is searched, a matched word and its part of speech information are extracted, and a candidate word is connected from the beginning of the processing target area to generate a word path. At the same time, the cost of the word path is calculated using the part-of-speech connection cost table of the language dictionary 55. The word paths are selected in descending order of cost so that the number of generated word paths is equal to or less than a certain number, the recognition candidate character is corrected based on the word path with the smallest cost, and the recognition result storage unit 45 is updated. (Step S14
0). The character recognition unit 30 is configured by steps S110 to S140. For parameters for calculating a certainty factor, for example, a weight held in a rule table, a part-of-speech connection cost table, or the like is extracted, and the certainty factor is calculated as a linear combination or an average value (step S15).
0). This constitutes the certainty factor calculation unit 40. In the word path generated in step S140, a character having a low degree of certainty is searched for, and two characters before and after the detected recognized character are searched.
A search is made for adjacent single characters that are not registered as words of one or more characters in the word path, concatenate them as one character string, and extract the character string as one word (step S160). This constitutes the low confidence extractor 60. In order to perform post-processing of the recognition result, the recognition result stored in the recognition result storage unit 45 and the corresponding original image data are associated with each other, and the low confidence word and the high confidence word extracted in step S160 are extracted. By visually changing the color, changing the font size, underlining, or winking, the output is visually distinguished. The user corrects any error in the recognition result by referring to the output result (step S170). In this post-processing, when one word is corrected, the recognition result storage unit 45 is updated with the corrected word, and then the next word having a low degree of certainty is moved to. After the post-processing of all the recognition results is completed, the final result stored in the recognition result storage unit 45 is output to an output device such as a display, a printer or a file, or transmitted to another computer via a network. (Step S180). From step S160 to step S180, the result output unit 5
0.

<Modification 1> On the other hand, in the case of a visually impaired person, information is often obtained by reading out the recognition result. However, according to the method of the above embodiment, it is possible to visually judge by certainty, but it is not available to visually impaired persons. Therefore, the first modification example
Then, at the time of post-processing of the result output unit 50, instead of extracting a word with low confidence and visually displaying this word,
The voice information to be read can be distinguished by changing the sound quality or changing the loudness of the sound according to the certainty factor. When reading this low confidence word, it does not make sense as a word, so
Read multiple candidates for each character. For example,
When outputting sound in stereo, if the confidence is high, the right and left signals are output simultaneously, but if the confidence is low, the output destination of the sound is output, such as outputting only the right or left signal. It may be changed. Further, before and after reading out the recognition result, a signal sound or a message for distinguishing may be inserted. Also, it may not be possible for visually impaired people to see the image, so it may not be possible to completely correct the characters, but it is possible to guess and correct or imagine from the flow before and after Will. Therefore, instead of reading out the next low-confidence word directly after correcting the low-confidence word in post-processing, what the visually impaired does Make sure you know what to fix.

Alternatively, in the case of a hearing-impaired person, instead of jumping to the next word of low confidence, all the words may be read in order. For a healthy person, jumping to the next low confidence word without such reading increases work efficiency, so in the case of a healthy person and a visually impaired person,
You may specify whether to jump and use it properly. With the configuration as in the first modification, even a visually impaired person reads out a word having a low degree of certainty on a word-by-word basis, so that the user can easily correct a correct word, and the efficiency of editing work is improved. Also, when outputting the recognition result in this post-processing, if the visual and auditory outputs are output together, the visually impaired person and his / her helper can work together, so that the editing work is more efficient. Is improved. Also, when visually checked, similar shaped characters may cause omission of correction, so if you output both visual and auditory, you can read out similar shaped characters aloud Thereby, even a healthy person can easily find an error, and the efficiency of editing work is further improved.

<Modification 2> In the case of a visually impaired person,
In many cases, the recognition result is obtained by using Braille. However, according to the method of the above embodiment, it is possible to visually judge by certainty, but it is not available to visually impaired persons. Therefore, in the second modification, at the time of post-processing of the result output unit 50, instead of extracting a word with a low degree of certainty and visually displaying this word, a method that can be tactilely determined by a braille output device is used. For example, the height of the unevenness of Braille is changed, and a mark or the like for distinguishing before and after characters or words with low confidence is inserted. When the low confidence word is corrected in the post-processing, the jump to the low confidence word is performed by using the braille cursor line (the braille pin comes out) on or around the line including the low confidence word.
As shown. With the configuration as in the second modification, even a visually impaired person outputs a word having a low degree of certainty as braille in word units, so that the user can easily correct a correct word and the efficiency of editing work is improved. In addition, when outputting the recognition result in this post-processing, if the tactile and auditory are output together, even if the visually impaired auditoryly misses, it can be checked tactilely,
Editing work is more efficient. In the present embodiment, a word path is created from the candidates recognized by the character recognition unit 30 by using the language dictionary 55, and a character string (word) including a character having a low degree of certainty is extracted from the word path. Although the extraction is performed, the same effect can be obtained even if the recognition result is searched for an unknown word that is not registered as a word in the language dictionary 55 in the post-processing of the recognition result of the result output unit 50. Can be.

<Embodiment by Computer> Further, the present invention is not limited to the above embodiment. For example, the character recognition device 100 shown in FIG.
Computer device 20 having a hardware configuration such as
It can be realized by setting the value to 0. That is, the computer device 200 includes a keyboard, a mouse, a touch panel, a scanner, a Braille input device, and the like, and includes an input device 1 used for inputting information, various output information, information input from the input device 1, and the like. Output device 2 for displaying or outputting to a printer or Braille output device, a CPU (Central Processing Unit) 3 for operating various programs, and a program itself. A memory 4 for holding information temporarily created when executed, an original image storage unit 25, a recognition dictionary 35, a recognition result storage unit 45, a language dictionary 55, a program and a program of the character recognition device of the present invention Storage device 5 for storing temporary information at the time of execution
A medium drive device 6 used to load a recording medium storing programs, data, and the like, read them, and store them in the memory 4 or the storage device 5;
And a network connection device 7 which is an interface for connecting to the network. The network 9 is connected to a computer device 20.
This is a transmission path for connecting the computer 0 and another computer device 200, and is generally realized by a cable, and TCP / IP is used as a communication protocol. However, the transmission path is not limited to cables, but may be any of wireless, wired, or broadcast waves as long as the communication protocol between them is the same. For example, LAN (Local Area Network)
k), WAN (Wide Area Network), Internet, analog telephone network, digital telephone network (ISDN: Integral S)
ervice Digital Network), PHS (Personal Handy Phone System), mobile phone network, satellite communication network, and the like.

In such a configuration of the computer device 200, each function constituting the character recognition device shown in FIG. 1 is programmed and written in advance on a recording medium such as a CD-ROM. By installing the program in a computer 200 equipped with a medium drive device 6 such as a CD-ROM drive at a site, storing these programs in the memory 4 or the storage device 5 of each computer device 200, and executing the program. The functions similar to those of the above-described embodiment can be realized. In addition, as a recording medium, a semiconductor medium (for example, RO
M, IC memory card, etc.), optical media (eg, DVD,
MO, MD, CD-R, etc.), magnetic media (eg, magnetic tape, flexible disk, etc.) and the like. In addition, not only the functions of the above-described embodiment are realized by executing the program loaded into the memory 4 of the computer device 200, but also the operating system or the like can execute a part of the actual processing or It also includes a case where all the operations are performed and the functions of the above-described embodiment are realized by the processing. Also,
When a program for realizing the above-described embodiment is a semiconductor recording medium such as a ROM, the program is directly loaded into the memory 4 instead of the medium driving device 6 and executed.

<Operation in Network Environment of the Present Invention> FIG. 5 shows a configuration of an embodiment in which the present invention is connected to a wired or wireless communication network for operation. For example, a server 210 holding a character recognition program and a terminal 220 used by a plurality of users are connected via the network 9. In this case, the server 210 and the user terminal 22
0 is constituted by the general-purpose computer device 200 shown in FIG. The user logs in to the server 210 from the terminal 220 or inputs image data for character recognition, and requests the character recognition program of the server 210 to execute character recognition. The character recognition program of the server 210 returns a character recognition result for the character area of the transmitted image data to the requesting terminal 220. User terminal 2
20 outputs the recognition result and the original image data while comparing the recognition result and the original image data, and performs post-processing. This has the advantage that the latest character recognition program can always be used. When the server 210 and the terminal 220 are connected via a wired or wireless communication network as shown in FIG. 5, a character recognition program for realizing the functions of the present invention is stored in a storage device such as a magnetic disk of the server 210. It can also be distributed to the terminal 220 in a form such as download. Further, a character recognition program for realizing the functions of the present invention may be provided by distribution through a medium or a broadcast wave.

[0017]

As described above, according to the present invention,
It is possible to confirm the certainty of the recognition result irrespective of the presence / absence of visual or auditory disturbance. In addition, since a word that is not certain is extracted for each word, it is easy to correct the word, and the work efficiency of post-processing is improved.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a functional configuration of an embodiment.

FIG. 2 is a diagram illustrating character recognition results of a conventional example and the present invention.

FIG. 3 is a flowchart illustrating a processing procedure according to the embodiment.

FIG. 4 is a diagram showing a computer device for operating the character recognition device of the present invention.

FIG. 5 is a diagram for explaining an operation example in a network environment according to the present invention.

[Explanation of symbols]

Reference Signs List 1 input device, 2 output device, 3 CPU, 4 memory, 5 storage device, 6 medium drive device, 7 network connection device, 8 bus, 9 network, 10 control unit, 20 image input unit, 25 original image storage unit, 30 Character recognition unit, 35 recognition dictionary, 40 confidence calculation unit, 45
Recognition result storage unit, 50 result output unit, 55 language dictionary,
60 low confidence extractor, 100 character recognition device, 200
Computer device, 210 server, 220 terminal

Claims (7)

[Claims] 1. A character recognition apparatus for performing a post-process on a character recognition result of a document image to obtain a recognition result, comprising: an image input unit for inputting a document image; and a character area of the image input by the image input unit. A character recognition unit for recognizing a character, a confidence calculation unit for calculating the certainty of the character recognition result for each character obtained from the character recognition unit, and a post-processing of the character recognition result obtained from the character recognition unit. A result output unit for outputting a result, wherein the result output unit is a character before and after a character for which the likelihood of the recognition result for each character calculated by the certainty factor calculation unit is low during post-processing of the character recognition result. A character recognition device characterized by searching for a word, extracting a word including a character having a low probability of the recognition result, and post-processing this word. 2. The character recognition device according to claim 1, wherein the result output unit visually distinguishes a word determined to have low probability of the recognition result from a word not determined to be low. A character recognizing device which outputs and facilitates post-processing by a user. 3. The character recognition device according to claim 1, wherein the result output unit aurally distinguishes between a word determined to have low probability of the recognition result and a word not determined to be low. A character recognizing device which outputs and facilitates post-processing by a user. 4. The character recognition device according to claim 1, wherein the result output unit tactilely distinguishes between a word determined to have low probability of the recognition result and a word not determined to be low. A character recognizing device which outputs and facilitates post-processing by a user. 5. The character recognition device according to claim 1, wherein, when the uncertain word is corrected, the unreliable word is replaced by the corrected word. A character recognition device, characterized in that the character is replaced and the process proceeds to the post-processing of the next uncertain word. 6. A character recognition method for performing character recognition on a character image and post-processing the recognition result, searching for characters before and after a character having a low probability of the recognition result, and determining the probability of the recognition result. A character recognition method characterized by extracting a word including a low character and correcting the word so as to improve post-processing work efficiency. 7. A computer-readable recording medium in which a program for causing a computer to function as a character recognition device that performs post-processing on a character recognition result of a document image to obtain a recognition result is provided. An image input unit, a character recognition unit that recognizes a character in a character region of an image input by the image input unit, and a confidence that calculates a certainty of a character recognition result for each character obtained from the character recognition unit. A calculation unit, and a result output unit that post-processes the character recognition result obtained from the character recognition unit and outputs the result, wherein the result output unit calculates the certainty factor during the post-processing of the character recognition result. Search for characters before and after the character with low confidence in the recognition result for each character calculated in the section, extract words containing characters with low confidence in the recognition result, and post-process this word Computer readable recording medium character recognition program recorded you so that.