TW509878B - Correction method for table characters and the user interface thereof - Google Patents

Abstract

The present invention discloses a correction method for table characters and the user interface thereof, which includes the following steps: first, automatically correcting the possible errors from the recognition techniques according to the character definition and scope; then, automatically correcting the field character value according to the recognition result of the picked items to reduce the amount of characters to be corrected; next, determining the attributes of connected words based on the positions of the character images and the positions of the learning samples to improve the efficiency of the correction operation; finally, displaying the characters to be corrected from different tables and different fields on the user interface to provide the character images, the character values, the periphery image for the character, the connected character processing function, and other related information, so as to improve the operational performance of the character correction.

Description

Field of the Invention: A method for correcting literacy elements * 2 obtained by the recognition device is extracted from the form document identification :: = character especially-uniform. π 得 正 子 子 的 treatment I Background of the invention: Type: 4 more political! In order to investigate or count each of the t: and industry and commerce groups, Bei 1 will print and format the form, and the data will be checked and filled out by =. Take the monthly target-time human resource visits of the Chief Accounting Office of the Executive Yuan as an example, each visit will generate an average of about 25, _ Zhang interview forms, usually each table;: contains about 包含 stops need to be filled in . In order to count this huge form data, traditionally, the content of the form is saved as image data through the image scanning device. Then use the form column imitation positioning technology to find the position of the field on the form, and extract: the written image data of the upper block. Re-use the identification system, such as: The identification system (OCR) identifies the image data-and stores and processes the data to mention that the following three paper standards apply the Chinese National Standard (CNS) M specification (21〇 > < 297 mm) A7 '---------- —_B7 V. Description of the invention (^ ^ ~ ---- Various information required by the community. For the form positioning technology, still There will be some fields that cannot be successfully positioned. These untargeted blocks will be left for subsequent manual correction based on the ^ form as an image. As for the successfully positioned blocks, the recognition system cannot reach 1 〇 ％% success rate of character recognition, so recognition failure characters such as rejection or low recognition accuracy are generated. These recognition failure characters will be displayed on the screen in the form of characters to be corrected. The user performs the correction manually. Refer to Figure 1 for the data processing flowchart of a traditional character correction system. As shown in the figure, the processing flow starts at box 2 and then enters from box 4 Identification system It is worth mentioning that the recognition result of the character correction system entered here is only the part of the recognition failure. As for the data of successful recognition, it will not be included in the characters to be corrected here. Then, the employees of the Central Bureau of Standards of the Ministry of Economic Affairs printed the characters to be corrected and entered the box 6 to perform the character correction process. By displaying the image data of the characters that failed to be identified individually on the screen, it was manually performed. Verification is performed verbatim to complete the character correction procedure. After all the character correction operations are completed, the column-based correction process is performed as shown in box 8. After that, the steps are performed in sequence as in box 1. 〇The paper size shown in the table is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) A7 ~ ^ -----------_ V. Description of the invention (3) ^ ~~钟 "-'--- Single-based correction procedure. Finally, end the character correction, department, and data processing flow at box 12. Because of the correction procedure of the stop and the form correction procedure, it is often necessary to display Partial or full Part of the early image often requires a lot of data processing time. In addition, when moving the form image to correct the f material, φ requires a lot of waiting time and reduces the operating speed. In this conventional technique, Performing a character correction procedure before the form correction procedure will help reduce the burden of subsequent field and form corrections: it can correct "data from several different forms and different form blocks at the same time to save on the character image When it is clear, + you need to view other parts of the image to complete the character correction, which speeds up the overall correction speed. Because the character correction process plays a significant role in the processing speed of the entire form, how can you improve the character? The processing power of the meta-correction program will profoundly affect the speed of reading the entire form data. Therefore, the existing problems of character correction technology will be discussed in more detail below. In the case of the character recognition error of the Central Bureau of Standards of the Ministry of Economic Affairs, first of all, in addition to the failure of the recognition technology to achieve perfection, the following main reasons can be summarized in actual operation: (1) When inputting form images, the quality is poor and noise is high. 5 A7 B7 V. Description of the Invention (Small) (1) The image of the characters on the seat is covered by the seal and there is much noise. (2) The image area of the character is smeared, daylight or creased. (4) The form making error is too large, causing the stop positioning failure. (5) The handwritten font is too scribbled, too large or too small. G.) Writing instruments are too light or too thick, too large or too small. (7) The handwritten fonts are connected, and the two characters are mistaken for a fake character. For further analysis and comparison of these misjudgment modes, the behavior of the identification system's identification errors can be divided into the following four categories: As shown in Figure 2A, in the range captured by the identification system, only noise and no character images are seen. (1) As shown in Figure 2B, although there is a character image failure in the range captured by the recognition system. (3) Deviations or errors occur as shown in Figure 2C (4) As shown in Figure 2D, please ^ Read the notes before reading _ Fill in. J page but too much noise caused identification misordering Central Ministry of Economic Affairs The position of the characters captured by the Standards Bureau ’s Consumer Cooperative ’s print recognition system resulted in incomplete character images. The neutron images in the range captured by the recognition system are connected and the recognition fails. These characters to be corrected will rapidly expand as the number of forms to be processed increases, resulting in a load of character correction operations. In this case ’it ’s urgent to have a powerful set of form characters

The system is dealing with the above problems. In this field, some related research and development have been done in the past. For example, the Republic of China Patent No. 0873 No. 19 (Xu Yingshi, etc.) proposed an image display method for computer-aided data registration, correcting characters in order, correcting stoppages, and finally correcting the entire form. The method of manually correcting the rejected characters in the character field proposed by this reference example is a correction operation method based on a single form. The character correction day is shown in Figure 3. The file labeling information of the original form of the character f to be corrected is displayed at the bottom of the screen at 14. The reference numeral 16 shows the image of the character to be corrected, and You and Yu provide a function for the user to perform correction directly. With this method, click, the handwritten text image cannot be correctly judged because the image is not clear, the image = position deviation, or affected by noise :: The correction operator must ignore the character and use the handwritten text to When the images are connected, the character cannot be corrected at one time. Green Works 2 US patent 5251273 (BettS, etc.) proposed form reading = industry department, system architecture flow, and data structure design, and correct them in order. Scoop: identify the errors after the form. In the device proposed by this reference example

St: identification data processor, and a kind of machine-generated data used to record identification results and correction history, and pass it to the parent processor. When the error correction process is completed, the work (210X297 male |) V. Invention description (& the station image will be displayed on the screen for human correction. Another example is US Patent No. 5305396 (Betts reading operating system architecture flow and data structure design) According to = the error after scanning the identification form. You can select the character recognition process and the identification data correction process for different customer forms. This reference example proposes to input a form template before identification, which contains the settings set according to customer needs. System operating parameters. In summary, the above known techniques can only display the image of the word S to be corrected, the description of the stop and the recognition result on the screen when performing the screen correction procedure. This method is useful for noise images. , Concatenated characters, and character images extracted from the wrong position cannot be corrected effectively. In addition, for the characters to be corrected, the definition of the character attributes and the limits of the character's block itself are not taken into account. Therefore, the number of characters to be corrected cannot be reduced in advance. Therefore, the present invention proposes this method based on such a background. High-efficiency form character correction method and user interface. The purpose and summary of the invention printed by the Employees' Cooperative of the Central Government Bureau of the Ministry of Economic Affairs: The purpose of the present invention is to provide a form character correction method and user interface. Effectively reduce the paper size processed by the identification system. Applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) 509878. 5. Description of the invention (1) The number of characters to be corrected can be corrected automatically in advance. Errors can be corrected directly. Another object of the present invention is to provide a user interface for the correction of form characters? The user interface of the unified application can perform meta corrections across form stops. According to the foregoing background of the invention Related to the problems mentioned in the prior art, the present invention proposes a new form character correction method and a more powerful user interface. The brief content is described below. After the form is processed by the recognition system Characters to be corrected that fail to be identified will be generated, and these characters to be corrected will enter the screen character correction procedure. First, the character correction program will automatically correct according to the attributes, definitions, and reasonable range values of the block corresponding to the character to be corrected. The correction process is as follows: (1) Check the to be corrected Whether the character falls within the meaning and range of its corresponding placeholder. (2) If yes, the value of the character to be corrected is not modified. If not, the character value of the character to be corrected is modified, and Select the best candidate that meets the definition and scope of the block. Gage scale is 509878 A7 B7 5. Description of the invention (exempt) Among them, the so-called candidate is the possible recognition result generated by the recognition system corresponding to the character to be corrected. After the character to be corrected has been checked for definition and range, the character correction method of the present invention will then check whether a check mark is available on the form, and if so, correct the value of the character to be corrected according to the check mark. The steps are as follows: (1) Check whether the recognition result of the checked item is the same as the recognition result of the block character. (B) If yes, the character value is not modified. If not, modify the character value to the result of checking the selected item. It is worth mentioning that the above-mentioned definition of the check block and the check. The procedures of checking the symbols are directly processed in the form character correction system and will not be displayed on the screen. After the characters to be corrected have undergone the first two steps of automatic correction ’, they are then displayed on the screen for manual correction by humans. And: Ming specifically proposes a processing flow for the ligatures encountered during correction: (1) Check the image position of the character to determine whether it falls within a ^ paper standard applicable ^^ quasi (CNS) A4 specification ^ 10 509878 A7-* —______ _B7 V. Description of the invention (^ ~ — 一 ~~-一. The character position on (.)) If yes, it is judged as the connected character attribute. If not, it is judged as Unconnected character attributes. The results of the determined connected characters will be displayed in the user interface provided by the present invention to provide the operator with the ability to identify and directly modify the connected characters. As for the proposed user interface Then at least the following user interface can be provided: '() Child image display area to display images of multiple characters to be corrected > material and the recognition result of the character to be corrected, and allow the operator to correct the ^ Recognize the results. (2) Character surrounding image display area to display the surrounding image of the character to be corrected. (II) Hyphenation function display area 'to provide and display the information of the characters to be corrected. (IV) Other related Character image processing information. ▲ Printed by the Consumer Cooperatives of the Central Government Bureau of the Ministry of Economic Affairs In summary, through the advance character check action, the number of characters to be corrected can be effectively reduced. The use provided by the present invention The user interface has the processing function of connected characters and more powerful text correction capabilities. Therefore, the form character correction method of the present invention 11 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm y 509878 A7)

9878 A7 B7

V. Description of the Invention (I FIG. 6A is a flowchart of using the attribute range of a character to modify the character recognition result in the method of the present invention. FIG. 6B is a flowchart of modifying the character recognition by using the character attribute range in the method of the present invention. A schematic diagram of the results. FIG. 6C is a schematic diagram of using the attribute range of the character to modify the character recognition result in the method of the present invention. FIG. 7A is a flowchart of correcting the character recognition result by using the check mark in the method of the present invention. FIG. 7B is a schematic diagram of using the check mark to modify the character recognition result in the method of the present invention. FIG. 8 is a schematic diagram of a user interface in the method of the present invention. FIG. 9 is a specific implementation of a user community of the present invention. Example: Figure 10A is a single-character image and text diagram of the user interface in the present invention. Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs. Figure 10B is the surrounding image of the user interface field and Schematic diagram of stoppage a month. Figure 10C is a schematic diagram of the perimeter image and description of the stoppage of the user interface of the present invention. Figure 11 shows the connection of the user interface of the present invention. Yuanying's paper size applies Chinese National Standard (CNS) A4 specifications (210 > < 297 mm) ^) 9878 Printed by A7 B7, Consumer Cooperatives of the Central Bureau of Standards of the Ministry of Economic Affairs 5. Description of the invention (Image and text diagram. Section 1 Fig. 2A is a schematic diagram of the criteria for judging left and right connected characters in the present invention. Fig. 12B is a schematic diagram of criteria for judging up and down connected characters in the present invention. Fig. 12C is a flowchart for judging connected characters in the present invention. The figure is a display diagram of the connected character image in the present invention. FIG. 14 is a schematic diagram of the surrounding image block and related field description of the connected character block in the user interface of the present invention. Detailed description of the invention: The present invention proposes a new set of Character correction system to reduce the number of characters that fail recognition in the recognition system in advance. Then, through the method for identifying connected characters and the enhanced user interface disclosed in the present invention, the character correction of the screen character correction process is improved. Ability, and reduce the burden on subsequent identification of the entire block or the entire form. The following will explain the spirit of the present invention with a specific embodiment. First, refer to Figure 4, which shows a schematic diagram of the system function block for form and material extraction. As shown in the figure, as soon as you enter < 亍 身 ^ No, form 14 This paper size is applicable to China National Standard (CNS) A4 specifications ( 210X297 mm) (Please read the precautions of C before filling out this page)

5 (09878

Printed by the Consumer Cooperatives of the Central Bureau of Standards of the Ministry of Economic Affairs

20 Scanning by the scanner 22 will be converted into images and sent to the scanning system for further processing. ：: 后 ’24: will cooperate with all kinds of data and programs stored in the data contained in Zhao 26 (for example, magnetic disk, disc first) to extract the table data and store it. Scanning system 2 4 will store the amount of documents in each batch according to the scanned insert 2〇 ^ Then, ㈣2

The image data in each field will be extracted from the form image 36 according to the pattern template learned in advance, and the identification process will be performed. After the identification of Yuanyuan ’, a certain identification result 38 will be produced. Generally, I ΐ = Γ force: 28 will generate two types of recognition failure recognition results, such as recognition success and recognition failure, according to the reliable process of the recognition effect. 38 will be passed to the word reading positive system 3G, and the information of = recognition failure will be displayed on Calibration and registration of materials on the screen. Last character correction system 30 :::::

=:;: Single data storage "database ... For reference, please refer to Figure 5. After the recognition system w reads and recognizes, the recognition is not preceded by: The table character is in the middle: Character data screen character correction procedure If you make 70 in the middle, you will enter the character correction process shown in the picture. Second: Correction work. For example, the lesser Koala μ Spear King sequence starts at box 44 'Receiver' obtained at box 46 by the recognition system. Sent

This paper is suitable for Zhongsunjia #

---- „----— · (Please read the notes on the back before filling in this I) mn If HI m. • In 1--1 509878 A7 B7 V. Explanation of the invention (丨 If) to be corrected Character data. Then, at box 48, correct the properties or character ranges corresponding to the fields corresponding to each character in advance to reduce the number of characters to be corrected. Because various forms are not only characters, , Usually also contains a check mark corresponding to the character value. If the character has a corresponding check mark, the character to be corrected will be at box 50, and the character value will be corrected according to the check mark. It is worth mentioning Yes, the steps of correcting characters by using the block attribute range in block 48 and correcting characters by using a check mark in block 50 are internal calculation processes of the system and will not be displayed on the screen. Then, the word to be corrected Yuan will judge the situation of connected characters at block 52, and display the characters to be corrected on the screen for manual direct correction. Block 54 will check whether all the characters to be corrected have been corrected, and if not, repeat Return to box 46; if yes, end the entire screen character correction process , As shown in block 56. Next, referring to FIG. 6A, a flowchart of using the character range and attribute correction characters in the present invention is shown. The flow is as follows: (1) Startup flow, as shown in block 57. (2) ) Check whether the result of identifying the failed character falls within the definition and range value of the block character, such as block 59. (3) If step (2) is YES, the character is executed directly without correction. Please read the cloud. 'Note for refill% Printed on this page by the Consumers' Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs

16 509878 A7 B7 V. Description of the invention (block 67 judges whether all characters have been checked: if yes, the process is aborted, as shown in block 69; if not, the next character to be corrected is performed again as shown in block 59. (4) If step (2) is no, continue to block 63 to check whether there are candidate words that fit the definition and range of the character. (5) If step (4) is no, then directly execute the block without correcting the character. 67 to determine whether all characters have been checked: if yes, then the process is aborted, as in block 69; if not, the next character to be corrected is re-performed in the process shown in block 59. (6) if step (4) is yes, Then proceed to block 65 to correct the character with the candidate word. (7) Determine whether all characters have been checked, such as block 67, if it is', then abort the process, such as block 69; if not, the next to-be-corrected character is restarted. Perform the process shown in block 57 until all characters have been checked. Printed by the Consumer Cooperatives of the Central Bureau of Standards of the Ministry of Economic Affairs, refer to Figures 6B and 6C, which show the attributes and corresponding attributes of the characters in the present invention. Schematic diagram of range correction. Before explaining, first define a symbol:

Fm-n: It is the field number m in the form. It is printed by 17 Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs. 5. Description of the invention (6th order η character. As shown in Figure βB) , Read a while, and the identification of the shirt according to 1 60 can not be identified through the recognition class provides the possibility of which-the first candidate word 62 and silk S 8 points from the value of 1. The correction system 、、、 9 According to the character's attributes, Yu 笳 & π starts to grasp the ω and consumes the surroundings, and at the same time cooperates with the identification system, the candidate characters that are provided first are modified characters mm ^ ^ ^ : Recognition: When the recognition is performed collectively, if the recognition accuracy of the character is not: then the possible recognition results of the recognition character are ranked according to the possibility, and several candidate characters for the recognition character are provided. As shown in FIG. 6B The f message shown in the middle frame is clear, and the attribute of the F㈣ character 1 ^ is in the range of 0 ~ 2. The -th candidate word 62 generated by the recognition system is! After inspection, Falls between the range of the character, so the character is not modified. As shown in Figure 6C of the other embodiment, the character shadow 66 has a set of candidate words 68. The value of the first candidate candidate 70 in the == 7, and the value of the second candidate candidate 72 is 丨, and so on, and so on. The box 64 shows the Fm-n The attribute of a character is a number, and its range is between ° 2. After inspection, it is found that the value of the first candidate candidate 70 is 7 is not between 0 and 2, so it is impossible to be a correct number. Then go to To find the second candidate candidate 72, whose value is 1 and falls just between 0 ~ 2

• m ....... I 1 B = 111 · '丨 丨 ---_ installation 丨 丨 / Please read the precautions on the back 1¾ before filling in the clothes f) Order ΐ This paper size applies to Chinese national standards (CNS ) A4 specification (210 × 297 mm) Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 509878 A7 ___B7 5. Between the description of the invention (A), the value of the modified character Fm_n is 1. Through the inspection in this step, the number of characters to be corrected can be reduced initially. Next, referring to FIG. 7A, a flowchart of using the check mark to modify characters in the present invention is shown. The flow is as follows: (1) Start-up flow, such as block 73. (2) Check whether the selected item is the same as the corresponding character value, such as box 7 5. (3) If step (2) is YES, directly execute the block without correcting the characters 7 9 to determine whether all characters have been checked: if yes, abort the process, such as block 81; if not, then the next character to be corrected is Repeat the process shown in block 75. (4) If step (2) is no, proceed to F77 to correct the character with the value of the check mark. (5) Determine whether all characters have been checked, such as block 79, if yes, abort the process, such as block 81; if not, then perform the process shown in block 75 again on the next character to be corrected until all characters have been checked until. Refer to FIG. 7B, which shows the use of a check mark in the present invention to modify «19 This paper size applies the Chinese National Standard (CNS) A4 specification (210 × 297 mm) — ·! 1Γ (Please read the precautions for f · face before filling out this page). Schematic diagram of correcting the character value. Provide a stop to allow the operator to check the form on the 5th. When filling out the form, in addition to the check mark of the field writing content: In addition to writing, there will also be corresponding anger on the left side with option 1 gold option 2M Figure 7B as an example. Shelf 7 is for the operator to write and record the blank area on the right side of the table. If it is: The written character value is 1, the corresponding option 2 is substituted. For the operator, the character value is 2, which means that the book characters and ticks will be recorded at the same time. Normally, the two parts will be entered into the identification system at the same time. Identify. Because the recognition of the check mark symbol is not ^ ', so when writing characters and check marks 亓 =: The book will be corrected with the value of the check mark = = As shown in the figure, the recognition system will write', and The recognition result according to the check mark is the recognition result of the selected character, and the value of recognition = is directly corrected to 丨. Due to the recognition rate of the check mark by the recognition system, Toyoma & for form fields with check marks, the step will be further-reducing the number of characters to be corrected. After that, the characters to be corrected that cannot be corrected through the character attribute range or check mark are displayed on the screen and the screen character correction process is performed manually. It is worth mentioning that 'Figures 6B, 6c, and 7B process the characters to be corrected are all character corrections 509878 A7

I Please read the memorandum items on the back and front before filling in this page. 509878

Printed by the Consumer Cooperatives of the Central Bureau of Standards of the Ministry of Economic Affairs

This paper standard applies the concatenated characters determined by the Chinese national invention for further correction. When the selected character image 92 is displayed as two ligature characters in the ligature function display area, the two characters can be directly modified once, and the conventional technique can only be modified into a single character. Disadvantages of characters. As shown in FIG. 9 ′ is a specific example of the use of the user interface 82, and by using the input device, the operator can directly modify the characters to be corrected in the user interface 82. Refer to Figures U) A to 10c, which show the functions that can be achieved through the image display area around the characters in the aforementioned user interface. As shown in the subordinate figure, the recognized characters 0: of the character image 96 (as shown by reference numeral 98). However, the value of the character image may be true, or it may simply be noise in the form. Looking at the surrounding area of the image around the character in Figure 10B, the surrounding area of the sub-70 image 96 can be seen in macro view. If you refer to Figure ’through the surrounding image of the character ^ Area 108, you can find that the image 96 is only noise. Biography = single firefly! The calibration operator can only read from the recognition system. If the extracted image is only noise from the recognition system and washed in the U image, the operation value is displayed through the surrounding image of the character: Choosing to face 1, 1, / ^ The author will be able to easily report every 22 (210X297 mm)

509878 A7 B7 V. Description of the invention (> 丨 The staff consumer cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs printed the correct value of the child's shirt and made corrections. Refer to Figure 11 for an image of connected characters. When the operator logs in the form data, due to the writing habits, many connected characters often occur. The traditional recognition system recognizes the connected characters as a single character, and the traditional screen character correction system also Only the single-character correction function can be provided. Therefore, the correction of concatenated characters will have to be left in the subsequent block or character correction process, resulting in a burden of data processing in the subsequent paragraphs. As shown in the figure, The image of the character Fm-n 丨 16 shows that the 3 sub-elements are actually two connected characters, but according to the judgment result of the recognition system, the character value is 2 (for example, as indicated by the numeral 丨 丨 8). According to the traditional method, only the unit value can be corrected for the character value, and the character correction cannot be completed with one correction. In order to deal with the situation of connected characters, the present invention proposes The relative character judgment criterion is to quickly determine the connected characters in the screen character correction procedure, and complete the correction of the connected characters with a single correction. Refer to FIG. 1 2A 'to explain the present invention. The criterion for judging horizontally connected characters in English. Through the study of form blocks and characters, we can get the position that each character's learning sample range should occupy. With the learning sample coordinates of sub-element Fm-n, t Six and upper corners (please read the precautions on the back before filling out this page). Pei order 4 23 This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm)

The adjacent word Fin of the child Fm-n, and the coordinates of the lower right corner are (χ

The left of the coordinates of the learning sample of Fm- (n + 1) is actually processed.

The negative value is (χη3, γ, for convenient learning of coordinate plotting, filling the inner 4 without displaying its calculation process on the screen. The S sub-element correction system will capture the character Fm- The position actually occupied by the image of n is compared with the position of the learning coordinates. For the horizontally connected characters shown in Figure 12A, the upper left corner of the actual image of the character Fm-n is (and the lower right corner is (U2 ). If the position occupied by the image of the character Fm-n falls into two learning sample coordinates Fm-n and] pm- (n + 1) 2 at the same time, and the character adjacent to the character Fm-n learns When the overlapping area of the sample coordinate area exceeds the width of the 1/3 learning sample column, the character is determined to be a connected character. The detailed system operation rules are defined as follows: · Definition of OVERLAP (A, B) Calculate the overlapping area of the two areas A and B. Area A ·· The actual image coordinates of the characters Fm-n (χηm!) ~ This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 509878 ≫>

The area occupied by area B (χ ^ 2). The area occupied by the learning sample coordinates (χη3, Υη3) ~ (χη4, Υη4) of a certain field character, and the character is not the original character Fm_n, but the character adjacent to Fm-n The position of the sample coordinates.

The criteria for determining IF connected characters are as follows; f ίί ί f has a field character image FH, then the criterion for determining whether FH is a connected character is as follows: j 疋 φ 马 (FH) AND 〇VERLAP (Α, Β) Greater than or equal to (B / 3); THEN determines that the FH character image is a connected character

Printed by the ENDIF Consumer Standards Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs Refer to Figure 12B, which shows the situation of vertically connected characters encountered in the character correction process. This type of vertical concatenated characters is usually caused by the writing habits of form recorders. Cross-block concatenated characters are mistakenly recognized by the recognition system as a single character. Similar to horizontal ligatures, the coordinates of the upper-left corner of the learning sample of the character Fni — n are (χη1, γη 其), while the coordinates of the lower-right corner are 25. This paper size applies the Chinese National Standard (CNS) Α4 specification (210X297) ) 509878 A7 B7 Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 5. The invention description (min) is (Xn2, Yn2). The coordinates of the upper left corner of the learning sample of an adjacent field Fm'-rT in the vertical direction of the character Fm-n are (χ ^, Υη3), and the coordinates of the lower right corner are (χη4, γη4). As for the box representing the learning sample range of the characters Fm-n, Fm, -n ', it is drawn for convenience of explanation. The range of the learning sample during the actual operation will only be used as a parameter for the internal operation of the character correction system. Still referring to FIG. 12B, the actual image of the character Fm-n is connected up and down across the stop due to the writing habits, so that the recognition system recognizes its coordinates as (XhYJ ~ (χ2, Υ2). Judging the vertical connection When characters are used, the character correction system first reads the actual image coordinates of the connected characters Fm-n, and compares the positions occupied by the images of the characters Fm-n with the positions of the learning coordinates Fm, · n ,. If the position occupied by the actual image of the character Fm-n falls between the two learning sample coordinates Fm-n and Fm, -n, and is perpendicular to the character Fm-n, the character Fm'-n When the overlapping area of the learning sample coordinate area exceeds the height of the 1/2 learning sample block, the character is determined to be a connected character. The calculation rule is as follows: 疋 义 〇VERLAP (A, B) ·· Used to calculate the overlapping area of two areas a and B. Area A: the actual image coordinates of the characters Fm-n (χ " Υι) ild. (Please read the precautions on the back before filling this page)

1T 26 Invention Description Ui)

Area occupied by area B (X2, Y2). The area occupied by the ink feather 4 of a certain seat word 70, and the sample coordinates (χη3, γη3) ~ (χη4, Υη4) of the sample S, and this character is not the original character Fm-n , But with

Fm-n is the position of the sample coordinates of the adjacent characters in the vertical direction. The criteria for judging connected characters are shown below; fff has a block character image 仏〗, then the criteria for judging whether Fi-j is a connected sub is as follows: · IF (Fm-n is not equal to Fi-j) AND OVERLAP (AB ) Greater than or equal to (B / 2) '

THEN determines that the image of the Fi-j character is a conjunctive character ENDIF Refer to Figure 12C. The method for judging concatenated characters can be expressed as the following process: Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs (1) Startup process, such as a box 1 2 1. (2) Check whether the result of the recognition of the failed characters falls in two or more fields in the learning sample. 27 This paper size is applicable to the Chinese National Standard < CNS) A4 specification (210X297 mm) 509878 5. Description of the invention ( ), Such as box 12 3. (3) If step (2) is no, Βί Φ does not correct the characters (please read the precautions on the back before filling this page). Go directly to block 129 to judge | ^ 疋 No All characters have been checked: If yes, then Stop flow borrowing L%, as in block 1 3 1; if not, then the next to be corrected 窆-~ + 7L re-perform the flow shown in block 123. (D) If the answer to step (b) is yes, then proceed to block 125 to check whether the character complies with the hyphenation check rule. (5) If step (4) is no, then directly execute block 129 without correcting the characters to determine the county I. 4 to check whether all characters have been checked: if it is 'then abort the Liujizhenji private' as in block 1 3 1 ; If not, re-perform the flow shown in block 123 in the next to-be-corrected 窆-工 予 π. (6) If step (4) is YES, then 丨% 〃 疋 then go to block 127 to show that the character is a ligature. Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs (7) Judge whether all characters have been checked, such as box 12 9; if 疋, the process is suspended, such as box i 3; if not, the following one is to be corrected The character repeats the flow shown in box 1 23 until all characters have been checked. 28 This paper size applies to China National Standard (CNS) A4 (210X 297 mm) Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs A7 ------------- Β7 ___ V. Description of the invention ( Μ) ^ ~ ~ —— ^ Refer to FIG. 13 and FIG. 14, which show the situation when the character correction system of the present invention processes connected characters. First refer to Figure ^ 3. In the character image display area of the user interface, the Fm-n characters are unrecognizable, but the actual situation may be that the font image is broken there. 疋 Noise causes the recognition system to refuse to recognize the character. yuan. The image displayed by the character Fm- (n + l) is recognized by the recognition system as *. The fool 119 is a special function provided by the user interface of the character correction system of the present invention, and is used to provide whether the two adjacent sub-element images displayed on the user interface are adjacent characters in the real form. number. In other words, when the characters Fm_n and Fm- (n + 1) are two adjacent characters in the real form, the symbol 119 appears in the image display area of the user interface. In the conventional art, the characters Fm-n will not have enough information to be corrected, and the characters Fm < n + 1) cannot directly correct the connected characters. Referring to FIG. 14, if it is judged that the Fm- (n + 1) th character is a concatenated character through the above-mentioned criteria, the character surrounding the character image display area 120 at the lower left of the user interface can be observed. 1 24) and its surrounding image. And the ligature function function display area 1 22 at the bottom right of the user interface will automatically determine whether the character Fm- (n + l) is a concatenated character; if so, the interlocutor in it 29. The paper size applies to Chinese national standards (CNS) A4 specification (2 丨 〇297mm) (Please read the precautions on the back before filling out this page)-Packing _ Order • Note in 4 blocks. When the operator selects the character to be corrected with the cursor, the character correction system provided by the present invention will automatically use the judgment criteria to check whether the character is a ligature and display the judgment result on the ligature function display. Area 122. Therefore, the user can directly modify the value of the character Fm-n to 2 and the character F. (n + 1) according to the connected character information mentioned in the image display area 120 and the hyphen function display area 122 around the character. ) Has a value of 0. In this way, the connected sub-elements can be corrected in the screen character correction process, thereby reducing the burden of subsequent form correction operations. Finally, the character correction system to which the present invention is applied is stored through a data carrier (such as a 'disk, tape, optical disc, memory, etc.), and a central processing device is used as a computing core. The corrected characters are also stored in the Data carrier for subsequent form identification. Cooperating with peripheral devices, such as terminals, scanners, printers, and communication networks, will provide powerful form data processing capabilities. The character correction system for processing digital characters in forms is described above. However, this embodiment Application of the inventive method: correction to any form of character. The comparison of the present invention is as follows: The examples are described above, and are only used to help understand the actual yoke of the present invention, not to limit the spirit of the present invention, and to be familiar with the technology in this field. 30 509878 A7 B7 V. Description of the invention (called) Artist After comprehending the spirit of the present invention, within the scope of the present invention, it can be modified and replaced with some changes (please read the precautions on the back before filling this page), the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs Printed 3 This paper size applies to China National Standard (CNS) A4 specification (210X297 mm)

Claims (1)

Time. Θ gas 、 Applicable patents in the form of form word + steps: early… method, this method at least includes the following characters input at least one to be corrected by the identification system :: some to be corrected characters and corresponding The definition of the stop-the characteristics of the court's tribute, 柃 氺 私 士 # and me /, 4 of these to-be-corrected words-Yu Keshu's definition and scope are outside ... Waiziwu, and generated by the identification system Fall in the candidate that falls within the definition order, and correct the characters to be corrected outside the corresponding τ ^ 4 circle to reduce the number of right children; and the golden screen to display the one pair and the one This to-be-corrected word may be corrected in cooperation with the manual method. = Please correct the characters in the form of the "Scope" member of the patent: Niu Xian ... Check and correct the characters of the data to be corrected η: Department—The result of the internal operation of the microprocessor will not appear on the screen. 3. If the method of patent application scope contains a step, the corresponding symbol should be checked for correction. The form character corrector of the item uses the relative to the character to be corrected. 32 509878 8 8 8 8 ABCD Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs. 6. The scope of patent application. Meta correction method, wherein the step of reusing the check mark for correction includes at least the following steps: checking out the character to be corrected with the corresponding check mark; judging whether the value of the character to be corrected is consistent with the value of the check mark ; Correct the to-be-corrected characters that are inconsistent with the value of the check-mark with the value of the check-mark; and repeatedly check and correct all to-be-corrected characters with corresponding check-marks by the above two steps. 5. If the form character correction method in item 4 of the scope of patent application, wherein the step of correcting by using a check mark, is the result of the internal calculation of a microprocessor, it will not be displayed on the screen. 6. If the form character correction method of item 1 of the patent application scope includes a step of determining a connected character to be corrected to correct the connected character to be corrected. 7. If the form character correction method of item 6 of the scope of patent application is applied (please read the precautions on the back before filling this page) ^ —- I Threading This paper size is applicable to China National Standard (CNS) A4 specifications (210X 297 mm) 509878 It is stated in the patent fan park method that a series of steps are judged to be connected: The sub-element step includes at least checking that the character to be corrected falls into; the a of the learning block on the sentence M checks whether there are more than two should ... The character to be corrected for the ^ position is: when you are waiting for a certain distance; when it is the rated distance, it is judged: the character that is on hold. Ziwu is the connection to be corrected. 8. If the sound suppression method of item 7 of the scope of the patent application, the method of basin φ μ, +,, and table correction is the learning field refers to the system obtained through learning. Mother—The template position of the character. + Bai Suozi to 9. If the scope of the patent application is 帛 7, the form letter of page, where the above-mentioned rated distance ′ includes the 1/3 width of the learning block adjacent to the binary wide to be calibrated horizontally. . Printed by the Employees' Cooperatives of the Central Bureau of Standards of the Ministry of Economic Affairs10. If the scope of patent application is 7 items, the form character correction method 'where the above-mentioned rated distance' includes the learning field adjacent to the character to be corrected in the vertical direction 1/2 of the height. / σ 34 This paper size is applicable to China National Standard (CMS) 8.4 (2i × 297mm) Printed by the Consumers' Cooperative of the Ministry of Health and Economics of the People's Republic of China. 如 · If you apply the binding strike method, the form 7 of the scope 1 in the scope of Feng- 妗 / 、 中 中 should be corrected A _ Wu * Right Zhengfang use Above the user interface ... “Ziwu” will be displayed on the right of the correction character with a symbol. The sub-character can be used by the operator to simultaneously correct all the characters to be corrected of the connected to-be-gamma. In ^ 丨 2 · —A form of character correction is used to face at least A k · User interface 'Please read the ik number first in the user field; At least includes: a turn-in device for inputting character correction operations The command / signal y is not set to display the message of the character to be corrected generated by an identification system. The display device has: a character image display area that displays individual image data of the characters to be corrected And the recognition result, and allow the correction result to be directly corrected; and a peripheral image display area, which displays a surrounding image of the corrector element. ^ 13. The user interface according to item 12 of the scope of patent application, wherein the retrieval device is used to select any one of the characters to be corrected to become the character being corrected. 35 This paper size applies to the Chinese National Standard (CNS) A4 specification (210X297 mm) I Order i 509878 Printed by A8 B8 C8 D8 of the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economics 々. Patent application scope _ 14. If the user interface of the patent application scope item 12, the display device includes a hyphen function display The area 'provides and displays information on whether the character being corrected is a concatenated character. 15. If the user interface of item 12 or 14 of the scope of patent application, the display device therein includes a character description display area to display the position of the form where the character being corrected is located. 16. For the user interface of the scope of patent application No. 15, the character description display area displays the position of the field where the character being corrected is located. 17. For the user interface of the scope of application for patent No. 12, the character image display area includes a plurality of blocks to display the image and recognition result of the character to be corrected. 18. For the user interface of the scope of application for patent No. 12, wherein the above character image display area displays a symbol that can be used to represent adjacent characters. 36 This paper size applies to China National Standard (CNS) A4 specification (210X297 mm) -1 II S-. S- · -.— II —1-1 (Read the precautions on the back and read this page first) Thread -----