WO2022169123A1 - Procédé de reconnaissance de données manuscrites en tant que caractères, et dispositif associé - Google Patents

Procédé de reconnaissance de données manuscrites en tant que caractères, et dispositif associé Download PDF

Info

Publication number
WO2022169123A1
WO2022169123A1 PCT/KR2022/000402 KR2022000402W WO2022169123A1 WO 2022169123 A1 WO2022169123 A1 WO 2022169123A1 KR 2022000402 W KR2022000402 W KR 2022000402W WO 2022169123 A1 WO2022169123 A1 WO 2022169123A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
handwriting data
handwriting
received
rotation angle
Prior art date
Application number
PCT/KR2022/000402
Other languages
English (en)
Korean (ko)
Inventor
정강훈
최화영
이상규
Original Assignee
주식회사 네오랩컨버전스
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 주식회사 네오랩컨버전스 filed Critical 주식회사 네오랩컨버전스
Publication of WO2022169123A1 publication Critical patent/WO2022169123A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/32—Digital ink
    • G06V30/333—Preprocessing; Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/60—Rotation of whole images or parts thereof
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/14—Image acquisition
    • G06V30/146—Aligning or centring of the image pick-up or image-field
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/32—Digital ink

Definitions

  • the present invention relates to a method and apparatus for recognizing characters in handwriting data, and more particularly, to rotating handwriting data that is not horizontally aligned so that character recognition is impossible to enable character recognition, and to move the recognized character writing data to a position before recognition.
  • the present invention relates to a method and apparatus for recognizing characters of handwriting data to be rearranged.
  • the electronic pen transmits writing trace related information to a predetermined device (eg, a smartphone or computer), and the predetermined device responds to the received writing trace related information.
  • a predetermined device eg, a smartphone or computer
  • the predetermined device responds to the received writing trace related information.
  • a technique for electronically reconstructing and displaying a writing trace written on paper is used.
  • the writing trace is restored in the application and displayed.
  • a handwriting trace is expressed as a vector graphic and stored as a file, which is displayed on a screen through a predetermined application.
  • the handwriting trace displayed in the above device may be converted into text through handwriting recognition through a character recognition engine and provided to the user.
  • Handwriting data is unstructured data having very large fluctuations depending on individual writing habits, such as writing speed, size of characters/vowels, crooked writing, and writing direction. Such unstructured data characteristics are directly related to a decrease in the recognition rate at the time of handwriting recognition according to individual writing habits.
  • 1 and 2 are diagrams illustrating examples of electronically implemented conventional handwriting trajectories.
  • the writing trajectory is arranged to be horizontal in the horizontal direction.
  • the recognition rate is excellent, and handwriting can be recognized as a character almost as it is.
  • the writing is not arranged so as to be horizontal, but is inclined in an oblique direction. Recognition is impossible with the currently applied character recognition engine, and also intermediate curves or figures cannot be recognized, so there is a problem in that a broken character is provided to the user due to an error in the character recognition result.
  • the character recognition rate can be increased through the rotation of the handwriting data for handwriting in any direction.
  • the character recognition data by rotating the character recognition data again and rearranging the handwriting data generated as graphic data according to the coordinates, the overall shape and position of the handwriting data that has undergone the character recognition procedure is expressed in a form similar to the received original handwriting data, It is possible to resolve the objection to the handwriting recognition result.
  • the unrecognized handwriting data is re-expressed as a vector graphic, the original handwriting content can be checked even if the character recognition fails.
  • 1 and 2 are diagrams illustrating examples of electronically implemented conventional handwriting trajectories.
  • FIG. 3 is a flowchart illustrating a method for recognizing characters in handwriting data according to an embodiment of the present invention.
  • FIG. 5 is a diagram illustrating an example of a process of dividing a word using the coordinates of the stroke data of FIG. 4 .
  • FIG. 6 is a diagram illustrating an equation according to an embodiment of the present invention.
  • FIG. 7 is a diagram illustrating an example of a word-separated result according to an embodiment of the present invention.
  • FIG. 8 is a diagram illustrating an example of an optimization function according to an embodiment of the present invention.
  • 9 to 21 are diagrams illustrating equations according to another embodiment of the present invention.
  • 22 is a diagram illustrating an example of an optimization function of stroke data in word units according to an embodiment of the present invention.
  • FIG. 23 is a diagram illustrating an example in which the handwriting data of FIG. 2 is recognized and rearranged.
  • 24 is a block diagram of an apparatus for recognizing handwriting data according to an embodiment of the present invention.
  • a method for recognizing a character of handwriting data includes: receiving handwriting data; extracting a rotation angle of the received handwriting data; rotating the received handwriting data based on the extracted rotation angle; and character recognition of the rotated handwriting data.
  • Separating the received handwriting data into word units may include dividing the received handwriting data into word units according to data related to a stroke time of the received handwriting data and a weight based on coordinates of the received handwriting data. may include
  • the extracting of the rotation angle of the received writing data may include extracting the rotation angle of the received writing data according to an optimization function calculated using a least-squares method.
  • the character recognition of the rotated handwriting data may include normalizing a size of the rotated handwriting data; and character recognition of the handwriting data whose size is normalized.
  • the character recognition of the rotated handwriting data may include: arranging the rotated handwriting data in a virtual page space according to a time sequence in which the handwriting data was written; and character recognition of the arranged handwriting data.
  • the handwriting data character recognition method may further include outputting the character-recognized handwriting data.
  • the outputting of the character-recognized writing data may include re-rotating and outputting the character-recognized writing data based on the extracted rotation angle.
  • the handwriting data character recognition method includes the steps of: extracting unrecognized handwriting data from the received handwriting data; generating the received handwriting data corresponding to the unrecognized handwriting data as graphic data; and superimposing the generated graphic data on the output handwriting data and outputting the overlapping step.
  • the handwriting data character recognition method includes: extracting handwriting data recognized as a figure from the received handwriting data; generating handwriting data recognized as the figure as graphic data; and superimposing the generated graphic data on the output handwriting data and outputting the overlapping step.
  • the step of re-rotating and outputting the character-recognized handwriting data based on the extracted rotation angle may include uniformizing the size, horizontal spacing, and inter-character spacing of the character-recognized writing data; and re-rotating and outputting the character-recognized handwriting data having the uniform size, horizontal spacing, and spacing based on the extracted rotation angle.
  • a receiving unit for receiving handwriting data a preprocessing unit extracting a rotation angle of the received writing data and rotating the received writing data based on the extracted rotation angle; and a character recognition unit for character recognition of the rotated handwriting data.
  • each component shown in the embodiment of the present invention is shown independently to represent different characteristic functions, and it does not mean that each component is composed of separate hardware or a single software component. That is, each component is listed as each component for convenience of description, and at least two components of each component are combined to form one component, or one component can be divided into a plurality of components to perform a function, and each Integrated embodiments and separate embodiments of components are also included in the scope of the present invention without departing from the essence of the present invention.
  • the components are not essential components for performing essential functions in the present invention, but may be optional components for merely improving performance.
  • the present invention can be implemented by including only essential components to implement the essence of the present invention except for components used for improving performance, and only having a structure including essential components excluding optional components used for improving performance Also included in the scope of the present invention.
  • FIG. 3 is a flowchart illustrating a method for recognizing characters in handwriting data according to an embodiment of the present invention.
  • step 310 an apparatus (hereinafter, referred to as a 'character recognition apparatus') that performs a method for recognizing text on handwriting data receives handwriting data.
  • a 'character recognition apparatus' an apparatus that performs a method for recognizing text on handwriting data
  • the handwriting data refers to handwriting trace data stored in a file in a smartphone, tablet, or computer, or handwriting trace data currently displayed on a screen.
  • the handwriting data consists of stroke data consisting of a continuous set of coordinates.
  • the stroke data represents one stroke or spelling for each language.
  • Stroke data may be divided into data as detailed elements such as figures and pictures. Stroke data itself has no special meaning in cases other than figures, and is stored in contact with an input tool (electronic pen, stylus, smart pen, etc.) on paper or other media such as a smartphone or tablet, A single character or symbol data is stored through a contact-non-contact process.
  • step 320 the character recognition apparatus determines whether the handwriting data is a figure. If it is a figure, the process proceeds to step 360; otherwise, the process proceeds to step 330.
  • step 330 the character recognition apparatus extracts a rotation angle of the handwriting data.
  • the character recognition apparatus separates handwriting data in a page according to a predetermined criterion, and then extracts a rotation angle for each separated unit of handwriting data.
  • the character recognition apparatus separates handwriting data in a page in units of words.
  • a superset to which a meaning is given to a collection of stroke data is called a word or a word.
  • a word is meaningful data and serves as a minimum unit for handwriting recognition.
  • handwriting data is divided into word units.
  • the handwriting data storage time that is, data related to the stroke time of the handwriting data and the received handwriting Set weights based on the coordinates of the data.
  • a method of setting a weight according to an embodiment of the present invention is as follows.
  • Timestamp average of all handwriting data that is, T avg , which is the average of the required writing time for each stroke
  • the average of the time difference between handwriting data that is, the average of the time obtained by subtracting the start time of the next stroke from the end time of the previous stroke
  • T th T avg + T' avg .
  • a weight is added. In one embodiment of the present invention, a weight +2.5 is added.
  • T'/T th >1.7, a weight is added.
  • 1.7 is the experimental value obtained through the experiment.
  • a weight of +2.5 is added.
  • FIG. 4 is a diagram of received stroke data
  • FIG. 5 is a diagram illustrating a process of dividing a word using the coordinates of the stroke.
  • (x i , y i ) is the midpoint of the current stroke
  • (x i-1 , y i-1 ) is the midpoint of the previous stroke.
  • the stroke data is divided into word units.
  • the criterion of weight 5 is a numerical value obtained experimentally.
  • the importance of writing data storage time is set higher than that of the coordinates.
  • FIG. 7 is a diagram illustrating an example of a word-separated result according to an embodiment of the present invention.
  • the character recognition apparatus After separating into word units, the character recognition apparatus obtains the directionality of the stroke data of the word unit, and then extracts the rotation angle of the stroke data of the word unit based on the directionality.
  • an optimization function representing a correlation is obtained using a least-squares method, and a rotation angle of stroke data in word units is obtained according to the optimization function.
  • the correlation of data is analyzed by finding the regularity of the dependent variable y that changes according to the independent variable x.
  • the least squares method is used for correlation analysis of the atypical stroke data.
  • the directionality, slope, and scale factor of the word can be calculated, and the words are normalized based on this value.
  • the least squares method is used to predict one reference variable as one or more predictive variables by a linear assumption so that the sum of the squares of the distance between the actual reference variable and the reference variable predicted by the linear assumption is minimized.
  • the least-squares method is generally used to quantitatively determine the causal relationship.
  • Each measurement (x 1 , y 1 ), ( x2, y2 ), ... , the distance from (x n , y n ) to the optimal function has a minimum value.
  • the equation of the least squares method is the same as the equation of FIG. 9 .
  • the sum of squared deviations ⁇ 2 is referred to as a residual, and the error is the same as the equation of FIG. 10 .
  • a and b have a value that minimizes the error ⁇ 2 when the partial derivative of each becomes 0.
  • the suitability of the optimization function can be determined through the correlation coefficient.
  • the value of the correlation coefficient is a value indicating the data representation suitability of the optimization function, and the closer to 1, the more optimized the function is.
  • the correlation coefficient is 1, all data exactly match the optimization function; when it is close to 1, it does not match but is close to a straight line; when it is 0, it means that all data coordinates are evenly distributed and do not approach a straight line. .
  • a correlation coefficient r 2 can be obtained as shown in the equation of FIG. 21 .
  • the correlation coefficient r 2 is closer to 0, it indicates the degree of inappropriateness of the optimization function, and as the correlation coefficient r 2 is closer to 1, it indicates that the optimization function is suitable. In this way, it is possible to determine the suitability of the optimization function using the correlation coefficient.
  • FIG. 22 is a diagram illustrating an example of an optimization function of stroke data in word units according to an embodiment of the present invention.
  • the red solid line is the optimization function.
  • the rotation angle of the stroke data can be known due to the coefficient a of the variable x due to the characteristic of the linear function. For example, when a is -1, it can be seen that the angle is tilted by 45° in the clockwise direction.
  • the text recognition apparatus horizontally rotates the received handwriting data based on the extracted rotation angle.
  • the entire coordinates of the stroke data may be rotated based on the rotation angle.
  • the vertical and horizontal ratio of the word unit handwriting arranged to be rotated and horizontal cannot exceed 1.0. If the ratio is greater than 1.0, since the direction of the word is vertical, it is possible to additionally rotate the handwriting in word units by 90° or 270°.
  • the character recognition apparatus recognizes the rotated handwriting data as a character.
  • step 360 If the character recognition apparatus does not recognize the rotated handwriting data, the process moves to step 360 .
  • step 360 the character recognition device generates the handwriting data as graphic data.
  • the character recognition apparatus generates handwriting data recognized as a figure or not recognized as a character as vector graphic data.
  • the character recognition apparatus generates text by applying a predetermined font to the character-recognized handwriting data.
  • the size of the text is output to be the same as the normalized size when character recognition is performed for each word unit, and the spacing and horizontal spacing of the text within the word unit are the same and outputted. Thereafter, the character recognition apparatus rearranges the text according to the rotation angle for each word unit and the received coordinates and outputs the rearranged text on a new page.
  • the character recognition apparatus superimposes the handwriting data generated as graphic data on the text according to the coordinates and outputs it.
  • FIG. 23 is a diagram illustrating an example in which the handwriting data of FIG. 2 is recognized and rearranged.
  • the character recognition apparatus may rearrange and output the text box form for each word unit.
  • the handwriting recognition result may be provided to an external document through copying and pasting, or the like, or the original form and order of saving in a rich format may be maintained.
  • 24 is a block diagram of an apparatus for recognizing handwriting data according to an embodiment of the present invention.
  • the receiver 1010 receives handwriting data.
  • the handwriting data refers to handwriting trace data stored in a file in a smartphone, tablet, or computer, or handwriting trace data currently displayed on a screen.
  • the handwriting data consists of stroke data consisting of a continuous set of coordinates.
  • the stroke data represents one stroke or spelling for each language.
  • Stroke data may be divided into data as detailed elements such as figures and pictures. Stroke data itself has no special meaning in cases other than figures, and is stored in contact with an input tool (electronic pen, stylus, smart pen, etc.) on paper or other media such as a smartphone or tablet, A single character or symbol data is stored through a contact-non-contact process.
  • the received handwriting data is data continuously including coordinates and time information, and has various forms in which figures and symbols are mixed.
  • the preprocessor 1020 determines whether the handwriting data is a figure. In the case of a figure, vector graphic data is generated according to the corresponding stroke coordinates.
  • Whether to determine whether a figure is a figure may be determined by determining handwriting data corresponding to a predefined figure as a figure, or writing data determined not to be a character as a figure. Whether to determine the figure may be determined in advance through learning.
  • the preprocessor 1020 extracts the rotation angle of the handwriting data.
  • the preprocessor 1020 separates the writing data in the page according to a predetermined criterion, and then extracts a rotation angle for each separated unit of the writing data.
  • the preprocessor 1020 separates the handwriting data in the page in word units.
  • a superset to which a meaning is given to a collection of stroke data is called a word or a word.
  • a word is meaningful data and becomes the minimum unit for handwriting recognition.
  • handwriting data is divided into word units.
  • the preprocessor 1020 determines the writing data storage time, that is, the stroke time of the writing data and A weight is set based on the associated data and the coordinates of the received handwriting data.
  • a method of setting a weight according to an embodiment of the present invention is as follows.
  • the preprocessor 1020 obtains T th , which is a time threshold.
  • T th T avg + T' avg .
  • the preprocessor 1020 obtains the average coordinate for each stroke and uses it as the measured value (x i , y i ), obtains the maximum and minimum of the coordinate value, and calculates the distance from the most distant value to the radius (R) of the stroke i ) is determined.
  • the preprocessor 1020 uses a known algorithm for an algorithm for obtaining the maximum and minimum of the coordinate values.
  • the preprocessor 1020 adds weights when the time difference T' between the current stroke and the subsequent stroke is greater than the time threshold value T th . In an embodiment of the present invention, the preprocessor 1020 adds a weight of +2.5.
  • the preprocessor 1020 adds weights when T'/T th >1.7.
  • 1.7 is the experimental value obtained through the experiment.
  • the pre-processing unit 1020 adds a weight of +2.5.
  • the preprocessor 1020 adds a weight of word separation based on the coordinates as follows. The conditions of the coordinate-based processing will be described below with reference to FIG. 5 mentioned above.
  • the preprocessor 1020 obtains a distance between the midpoint of the current stroke as a reference and the midpoint of the previous stroke.
  • the distance l is the same as the equation of FIG. 6 .
  • (x i , y i ) is the midpoint of the current stroke
  • (x i-1 , y i-1 ) is the midpoint of the previous stroke.
  • the preprocessor 1020 performs the following through the radius R i of the current stroke and the radii R i-1 , R i-2 , R i-3 , and R i-4 of the previous stroke.
  • the preprocessor 1020 compares the R sum > R i condition with respect to the sum of the previous stroke radii R sum , and updates the R sum value when the condition is true for all stroke data (the first time is R sum ) assign the real maximum to R sum to calculate). Setting the previous stroke to four was obtained experimentally. In another embodiment of the invention, the number of previous strokes may be changed.
  • the preprocessor 1020 increases the weight by +1 when l > (R i + R i-1) with respect to the case of R sum > R i . If l > (R i + R i-1) does not hold, the preprocessor 1020 repeats the previous radius value and (l > (R i + R i-1), 2 ⁇ i ⁇ 4) and They are compared together, and if at least one condition is satisfied, the weight is increased by +1. If all the coordinate comparison conditions are not met, the preprocessor 1020 reduces the weight by -1 so that the word is not separated.
  • the preprocessor 1020 separates stroke data in word units when the weight is 5 or more according to conditions 4 to 6 above.
  • the criterion of weight 5 is a numerical value obtained experimentally.
  • the importance of the writing data storage time is set higher than that of the coordinates.
  • the preprocessor 1020 separates the word units, acquires the directionality of the word unit stroke data, and extracts a rotation angle of the word unit stroke data based on the directionality.
  • the preprocessor 1020 obtains an optimization function representing a correlation by using the least squares method, and obtains a rotation angle of the stroke data in word units according to the optimization function.
  • the correlation of data is analyzed by finding the regularity of the dependent variable y that changes according to the independent variable x.
  • the least squares method is used for correlation analysis of the atypical stroke data.
  • the directionality, slope, and scale factor of the word can be calculated, and the words are normalized based on this value.
  • the least squares method is used to predict one reference variable as one or more predictive variables by a linear assumption so that the sum of the squares of the distance between the actual reference variable and the reference variable predicted by the linear assumption is minimized.
  • the least-squares method is generally used to quantitatively determine the causal relationship.
  • the rotation angle of the stroke data can be known due to the coefficient a of the variable x due to the characteristic of the linear function. For example, when a is -1, it can be seen that the angle is tilted by 45° in the clockwise direction.
  • the suitability of the optimization function can be determined through the correlation coefficient.
  • the value of the correlation coefficient is a value indicating the data expression suitability of the optimization function, and the closer to 1, the more optimized the function is.
  • the correlation coefficient is 1, all data exactly match the optimization function; when it is close to 1, it does not match but is close to a straight line; when it is 0, it means that all data coordinates are evenly distributed and do not approach a straight line. .
  • the preprocessor 1020 horizontally rotates the received handwriting data based on the extracted rotation angle.
  • the preprocessor 1020 removes the angular element of the rotation angle extracted for each word unit. For example, when the rotation angle extracted above is 45°, the preprocessor 1020 rotates the word part by -45° to remove the angle element. In this case, the received handwriting data are horizontally and parallelly aligned.
  • the preprocessor 1020 may rotate the entire coordinates of the stroke data based on the rotation angle.
  • the preprocessor 1020 may normalize the size of the handwriting in units of words.
  • the size of each handwriting is normalized based on the vertical and horizontal ratios of the word, so that the handwritings in the word unit have the same size scale factor.
  • the horizontal spacing and the letter spacing of the handwriting data in the word unit may be changed equally, and the spacing between the word units may also be changed in the same way.
  • the vertical and horizontal ratios of the word unit handwriting arranged to be rotated and horizontal cannot exceed 1.0. If the ratio is greater than 1.0, since the direction of the word is vertical, the preprocessor 1020 may additionally rotate the handwriting in word units by 90° or 270°.
  • the character recognition unit 1030 recognizes the rotated handwriting data as a character.
  • the character recognition unit 1030 If the character recognition unit 1030 does not recognize the rotated writing data as characters, the writing data in word units is generated as graphic data. In an embodiment of the present invention, the character recognition unit 1030 generates handwriting data recognized as a figure or not recognized as a character as vector graphic data.
  • the output unit 1040 rearranges the character-recognized handwriting data on a screen or page and outputs it.
  • the output unit 1040 generates text by applying a predetermined font to the character-recognized handwriting data.
  • the output unit 1040 outputs the size of the text to be the same as the normalized size when recognizing characters for each word unit, and outputs the same spacing and horizontal spacing of the text within the word unit. Thereafter, the output unit 1040 rearranges the text according to the rotation angle for each word unit and the received coordinates and outputs the rearranged text on a new page.
  • the output unit 1040 superimposes the handwriting data generated as graphic data on the text according to the coordinates and outputs it.
  • the output unit 1040 may rearrange and output the text box form for each word unit.
  • the handwriting recognition result may be provided to an external document through copying and pasting, or the like, or the original form and order of saving in a rich format may be maintained.
  • the handwriting data character recognition method as described above can also be implemented as computer-readable codes on a computer-readable recording medium.
  • the computer-readable recording medium includes any type of recording medium in which data readable by a computer system is stored. Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage device.
  • the computer-readable recording medium is distributed in a computer system connected through a network, so that the computer-readable code can be stored and executed in a distributed manner.
  • functional programs, codes, and code segments for implementing the disk management method can be easily inferred by programmers in the art to which the present invention pertains.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Character Discrimination (AREA)

Abstract

L'invention concerne un procédé de reconnaissance de données manuscrites en tant que caractères, et un dispositif associé, le procédé consistant à recevoir des données manuscrites, à extraire l'angle de rotation des données manuscrites reçues, et à appliquer une rotation aux données manuscrites reçues sur la base de l'angle de rotation extrait pour reconnaître les données manuscrites ayant subi la rotation en tant que caractères.
PCT/KR2022/000402 2021-02-04 2022-01-10 Procédé de reconnaissance de données manuscrites en tant que caractères, et dispositif associé WO2022169123A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR1020210015814A KR20220112368A (ko) 2021-02-04 2021-02-04 필기 데이터 문자 인식 방법 및 그 장치
KR10-2021-0015814 2021-02-04

Publications (1)

Publication Number Publication Date
WO2022169123A1 true WO2022169123A1 (fr) 2022-08-11

Family

ID=82741304

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2022/000402 WO2022169123A1 (fr) 2021-02-04 2022-01-10 Procédé de reconnaissance de données manuscrites en tant que caractères, et dispositif associé

Country Status (2)

Country Link
KR (1) KR20220112368A (fr)
WO (1) WO2022169123A1 (fr)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20030066774A (ko) * 2000-12-27 2003-08-09 아스라브 쏘시에떼 아노님 입력 존에서 수작업으로 그려넣어진 문자들을 인식하기위한 방법 및 장치
JP2013242826A (ja) * 2012-05-23 2013-12-05 Panasonic Corp 画像処理装置およびこれを備えた原稿読取システム
KR20150044697A (ko) * 2013-10-17 2015-04-27 삼성전자주식회사 문서 보정 방법 및 그 전자 장치
KR20160053544A (ko) * 2014-11-05 2016-05-13 주식회사 디오텍 후보 문자의 추출 방법
KR101985612B1 (ko) * 2018-01-16 2019-06-03 김학선 종이문서의 디지털화 방법

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20030066774A (ko) * 2000-12-27 2003-08-09 아스라브 쏘시에떼 아노님 입력 존에서 수작업으로 그려넣어진 문자들을 인식하기위한 방법 및 장치
JP2013242826A (ja) * 2012-05-23 2013-12-05 Panasonic Corp 画像処理装置およびこれを備えた原稿読取システム
KR20150044697A (ko) * 2013-10-17 2015-04-27 삼성전자주식회사 문서 보정 방법 및 그 전자 장치
KR20160053544A (ko) * 2014-11-05 2016-05-13 주식회사 디오텍 후보 문자의 추출 방법
KR101985612B1 (ko) * 2018-01-16 2019-06-03 김학선 종이문서의 디지털화 방법

Also Published As

Publication number Publication date
KR20220112368A (ko) 2022-08-11

Similar Documents

Publication Publication Date Title
WO2021132927A1 (fr) Dispositif informatique et procédé de classification de catégorie de données
WO2021025290A1 (fr) Procédé et dispositif électronique permettant de convertir une entrée manuscrite en texte
WO2020180013A1 (fr) Appareil d'automatisation de tâche de téléphone intelligent assistée par langage et vision et procédé associé
WO2018062580A1 (fr) Procédé de traduction de caractères et appareil associé
WO2020082562A1 (fr) Procédé, appareil, dispositif et support de mémoire d'identification de symbole
WO2020251233A1 (fr) Procédé, appareil et programme d'obtention de caractéristiques abstraites de données d'image
WO2017039287A1 (fr) Système et procédé d'authentification de signature manuelle sur la base de segments
WO2020045714A1 (fr) Procédé et système de reconnaissance de contenu
WO2016080713A1 (fr) Dispositif d'affichage d'image à commande vocale et procédé de commande vocale de dispositif d'affichage d'image
WO2015050321A1 (fr) Appareil pour générer un corpus d'alignement basé sur un alignement d'auto-apprentissage, procédé associé, appareil pour analyser un morphème d'expression destructrice par utilisation d'un corpus d'alignement et procédé d'analyse de morphème associé
WO2020060019A1 (fr) Dispositif, procédé et système de détection de caractère
WO2021141419A1 (fr) Procédé et appareil pour générer un contenu personnalisé en fonction de l'intention de l'utilisateur
WO2016208941A1 (fr) Procédé de prétraitement de texte et système de prétraitement permettant de mettre en œuvre ledit procédé
WO2020082766A1 (fr) Procédé et appareil d'association pour un procédé d'entrée, dispositif et support d'informations lisible
EP3685279A1 (fr) Procédé de recherche de contenu et dispositif électronique associé
WO2021251539A1 (fr) Procédé permettant de mettre en œuvre un message interactif en utilisant un réseau neuronal artificiel et dispositif associé
WO2020159140A1 (fr) Dispositif électronique et son procédé de commande
WO2022169123A1 (fr) Procédé de reconnaissance de données manuscrites en tant que caractères, et dispositif associé
WO2021091124A1 (fr) Dispositif électronique et procédé de fonctionnement permettant de rechercher un fichier similaire à un fichier de référence sur la base d'informations de distribution concernant des caractéristiques de chaque fichier de la pluralité de fichiers
WO2018084503A1 (fr) Procédé de vérification à l'aide d'un clavier et modèle de comportement d'entrée de souris d'utilisateur, et support d'enregistrement enregistré avec un programme destiné à mettre en œuvre le procédé
WO2013187587A1 (fr) Procédé d'échantillonnage de données et dispositif d'échantillonnage de données
WO2016117854A1 (fr) Appareil d'édition de texte et procédé d'édition de texte sur la base d'un signal de parole
WO2024019226A1 (fr) Procédé de détection d'urls malveillantes
WO2016088954A1 (fr) Procédé de classement de spams, support d'enregistrement destiné à le mettre en œuvre et dispositif de classement de spams
WO2022139327A1 (fr) Procédé et appareil de détection d'énoncés non pris en charge dans la compréhension du langage naturel

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22749877

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22749877

Country of ref document: EP

Kind code of ref document: A1