CN108830245A - Body method is write in personalized handwritten form replacement - Google Patents
Body method is write in personalized handwritten form replacement Download PDFInfo
- Publication number
- CN108830245A CN108830245A CN201810657283.5A CN201810657283A CN108830245A CN 108830245 A CN108830245 A CN 108830245A CN 201810657283 A CN201810657283 A CN 201810657283A CN 108830245 A CN108830245 A CN 108830245A
- Authority
- CN
- China
- Prior art keywords
- chinese character
- group
- style
- stroke
- user
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/30—Writer recognition; Reading and verifying signatures
- G06V40/33—Writer recognition; Reading and verifying signatures based only on signature image, e.g. static signature recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/103—Formatting, i.e. changing of presentation of documents
- G06F40/109—Font handling; Temporal or kinetic typography
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Character Discrimination (AREA)
Abstract
A kind of personalization handwritten form replacement writing body method, belongs to artificial intelligence and soft project cross discipline, is mainly used for solving the problems, such as that the script of user is substituted for the handwritten form of target handwriting style according to user's handwritten form style of writing style, has main steps that：S1. one group of Chinese character of user is inputted, this group of Chinese character is referred to one of following three kinds of situations according to carefully and neatly done degree by S2.：Situation, time not abundant situation of having plenty of time and signature situation；S3. learn the style of writing style of this group of Chinese character out, gradient, hand-written dynamics and stroke feature including this group of Chinese character etc. using the method for deep learning, and form a four-tuple for characterizing this group of Chinese character style of writing feature according to the situation of classification and these style and features values；S4. the script Chinese character to be replaced according to this four-tuple analog subscriber, and be replaced.
Description
Technical field
The present invention relates to a kind of personalized handwritten form replacements to write body method, belongs to artificial intelligence with soft project and intersects
Section.
Technical background
User in writing Chinese characters can due to write environment influence and write out it is different style of writing styles Chinese character, for example,
When user is in the case where having plenty of time, a stroke that whens writing Chinese characters can be conscientious writes out carefully and neatly done Chinese character, including row
Between smooth, dynamics is moderate and Chinese character personal characteristics uniformly etc., however, user is eager to complete in the case where the time is not abundant
The writing of Chinese character, style of writing style can with have plenty of time in the case where the style of writing style write be very different, such as occur very
Connecting pen more, writing is more based on substantially identifying.Even, the style of writing style as user when doing signature and writing is more with preceding two
Person is different, and naked eyes can not identify completely sometimes.The present invention proposes that body method is write in a kind of personalized handwritten form replacement, is mainly used for
Solve the problems, such as the handwritten form that the script of user is substituted for target handwriting style according to user's handwritten form style of writing style,
It has main steps that：S1. input user one group of Chinese character, S2. by this group of Chinese character according to carefully and neatly spend be referred to following three kinds of situations it
One：Situation, time not abundant situation of having plenty of time and signature situation；S3. this group of Chinese character out is learnt using the method for deep learning
Style of writing style, gradient, hand-written dynamics and stroke feature including this group of Chinese character etc., and according to the situation of classification and these styles
Characteristic value forms a feature four-tuple for characterizing this group of Chinese character style of writing style；S4. it waits replacing according to this four-tuple analog subscriber
The script Chinese character changed, and be replaced.
Summary of the invention
A kind of personalization handwritten form replacement writing body method is mainly used for solving to use according to user's handwritten form style of writing style
The script at family is substituted for the problem of handwritten form of target handwriting style, has main steps that：S1. one group of Chinese of user is inputted
This group of Chinese character is referred to one of following three kinds of situations according to carefully and neatly done degree by word, S2.：It has plenty of time situation, time not abundant situation
With signature situation；S3. the style of writing style of this group of Chinese character out is learnt using the method for deep learning, gradient including this group of Chinese character,
Hand-written dynamics and stroke feature etc., and formed according to the situation of classification and these style and features values and characterize this group of Chinese character style of writing style
A feature four-tuple；S4. the script Chinese character to be replaced according to this four-tuple analog subscriber, and be replaced.It is a kind of
Personalized handwritten form replacement writes body method and has the following steps：
Step S1. inputs the handwritten Chinese character C of one group of userhandwriting={ch1, ch2, …, chn}；
Step S2. calculates the carefully and neatly done degree of this group of text, and distinguishing the handwritten form of user according to carefully and neatly done degree is in the feelings having plenty of time
It is completed under shape, in the case of the time is not abundant and in the case of signature,
Step S3. learns the style of writing style of this group of Chinese character out using the method for deep learning, the row including this group of text entirety
Gradient, the feature of hand-written dynamics and stroke, and the amount of taper, hand-written dynamics value and stroke feature value of this group of Chinese character are calculated separately,
The style of writing style four-tuple of user is formed with these three values after the step S2 selectes hand-written situation, i.e., carefully and neatly done degree (GZ), tiltedly
Angle value (Vinclination), hand-written dynamics value (Vstrength-means), stroke feature value (Fstroke)}；
Step S4. simulates the script Chinese character of user to be replaced according to the style of writing style four-tuple of this group of Chinese character of user
Cscript={cs1, cs2, …, csn, and replaced.
Wherein, step S2 includes：
The present invention sets three kinds of situations for distinguishing user's handwritten form：Situation, time not abundant situation of having plenty of time and signature feelings
Shape.The judgement of three kinds of situations determines by carefully and neatly spending, i.e.,：
Carefully and neatly done degree GZ is more than that threshold value GZ0 indicates that handwritten form is completed in the case of having plenty of time；
Carefully and neatly done degree GZ indicates that handwritten form is completed in the case of the time is not abundant between GZ1 and GZ0；
Carefully and neatly done degree GZ indicates that handwritten form is completed in the case of having plenty of time lower than threshold value GZ1；
The calculating carefully and neatly spent is as follows：
（1）
Wherein, LBnumberIndicate the stroke number of company's pen of one group of Chinese character of user, ZBnumberIndicate the total of one group of Chinese character of user
Stroke number.U is effect length coefficient, is belonged to [0,1], indicate one group of Chinese character length to the weighing factor carefully and neatly spent, if this
The length of group Chinese character is very short, then it is assumed that is likely to be signature, coefficient value is larger, if this group of word length is more than threshold value, then it is assumed that no
It is signature, coefficient value is smaller.
Step S3 includes：
The calculating of one group of Chinese character amount of taper：Learn the gradient of the row of this group of text entirety out using the method for deep learning, it is hand-written
The feature of dynamics and stroke.For the gradient feature of one group of Chinese character, present invention VinclinationIt indicates amount of taper, calculates as follows：
（2）
Wherein, heightch1Indicate first Chinese character ch of this group of Chinese character1Relative altitude, heightchnIndicate this group of Chinese character most
The relative altitude of the latter Chinese character, heightchn/2Show the relative altitude of the Chinese character among this group of Chinese character；
Step S3 further includes：
The calculating of one group of Chinese character hand-written dynamics value：The hand-written dynamics value V of i-th of wordstrengthIndicate, one group of Chinese character it is hand-written
Dynamics mean value computation is as follows：
（3）
Wherein, chiIndicate one group of Chinese character ChandwritingI-th of Chinese character.
Step S3 further includes：
The calculating of one group of Chinese-character stroke characteristic value：Stroke feature FstrokeIt indicates, calculates as follows：
（4）
Wherein, FkValue be 1,2,3,4,5,6,7,0, respectively indicate " just tiltedly ", " negative oblique ", " straight ", " length ", " short ", " big ",
The feature of the strokes such as " small ", " normal ".FiIt is obtained by the statistics of n Chinese character, if the F of stroke kiValue be 1 number of words more than 2
Number of words, then the feature F of stroke kiValue be 1, k=it is horizontal, vertical, skim, right-falling stroke, hook；
Step S3 further includes：
The style of writing style that the present invention defines user is a four-tuple：{GZi, Vinclination, Vstrength-means,
Fstroke }, indicate one group of text C of userhandwritingi={ch1, ch2, …,chnCarefully and neatly done angle value, amount of taper, average
The characteristic value of hand-written dynamics value and stroke.
Step S4 includes：
According to four-tuple { GZi, Vinclination, Vstrength-means, FstrokeCharacterization the style of writing style, simulation use
The script C that family requiresscript={cs1, cs2, …,csm, and be replaced.
Detailed description of the invention
Fig. 1 is a kind of general diagram of one embodiment of personalized handwritten form replacement writing body method；
Fig. 2 is a kind of specific implementation flow chart of personalized handwritten form replacement writing body method.
Specific embodiment
A kind of personalization handwritten form replacement writing body method, inputs one group of Chinese character of user, is classified as the three of setting
Kind situation：Situation, time not abundant situation of having plenty of time and signature situation are completed.User out is learnt using the method for deep learning
The style of writing style, such as gradient, hand-written dynamics and the stroke feature of this group of Chinese character of one group of Chinese character etc..After selecting the situation completed
The four-tuple that the hand-written style of writing style of characterization user is formed with these features, according to style of writing style four-tuple analog subscriber it is given to
Script text is replaced, and is replaced, specific embodiment is as follows：
Step 1）Such as the step 001 in Fig. 2, one group of Chinese character C of user is inputtedhandwriting ={ch1, ch2, …, chn}；
Step 2）Such as the step 002 in Fig. 2, the present invention has plenty of time to three kinds of situations of completion environment set of this group of Chinese character
Situation, time not abundant situation and signature situation, and which kind of situation is belonged to sort out the completion of this group of Chinese character using carefully and neatly done degree, it counts
Calculate this group of word ChandwritingCarefully and neatly done degree GZ it is as follows：
（1）
Wherein, LBnumberIndicate the stroke number of company's pen of one group of Chinese character of user, ZBnumberIndicate the total of one group of Chinese character of user
Stroke number.U is effect length coefficient, is belonged to [0,1], indicate one group of Chinese character length to the weighing factor carefully and neatly spent, if this
The length of group Chinese character is very short, then it is assumed that is likely to be signature, coefficient value is larger, if this group of word length is more than threshold value, then it is assumed that no
It is signature, coefficient value is smaller；
Step 3）Such as the step 003 in Fig. 2, judge that the carefully and neatly done angle value of this group of Chinese character belongs to the range of what situation, i.e.,：
As shown in the step 004-1 in Fig. 2, if carefully and neatly degree GZ is more than threshold value GZ0, being classified as handwritten form was filled in the time
It is completed in the case of abundant；
As shown in the step 004-2 in Fig. 2, the carefully and neatly done GZ that spends then is classified as indicating that handwritten form is in the time in GZ1 and GZ0
It is completed in the case of not abundant；
As shown in the step 004-3 in Fig. 2, the carefully and neatly done GZ that spends is lower than threshold value GZ1, then being classified as handwritten form is to have plenty of time
In the case of complete；
Step 4）Such as the step 005 in Fig. 2, the carefully and neatly done degree of present case is recorded, and calculates amount of taper Vinclination, using depth
The method of degree study learns the gradient of the row of this group of text entirety out, the feature of hand-written dynamics and stroke.For one group of Chinese character
Gradient feature, present invention VinclinationIt indicates amount of taper, calculates as follows：
（2）
Wherein, heightch1Indicate first Chinese character ch of this group of Chinese character1Relative altitude, heightchnIndicate this group of Chinese character most
The relative altitude of the latter Chinese character, heightchn/2Show the Chinese character relative altitude among this group of Chinese character；
Step 5）Such as the step 006 in Fig. 2, hand-written dynamics value V is calculatedstrength-means, the hand-written dynamics value use of i-th of word
VstrengthIt indicates, the hand-written dynamics mean value computation of one group of Chinese character is as follows：
（3）
Wherein, chiIndicate one group of Chinese character ChandwritingI-th of Chinese character；
Step 6）Such as the step 007 in Fig. 2, stroke characteristic value F is calculatedstroke(sk), stroke feature FstrokeIt indicates, calculates
It is as follows：
（4）
Wherein, FkValue be 1,2,3,4,5,6,7,0, respectively indicate " just tiltedly ", " negative oblique ", " straight ", " length ", " short ", " big ",
The feature of the strokes such as " small ", " normal ".FiIt is obtained by the statistics of n Chinese character, if the F of stroke kiValue be 1 number of words more than 2
Number of words, then the feature F of stroke kiValue be 1, k=it is horizontal, vertical, skim, right-falling stroke, hook；
Step 7）Such as the step 008 in Fig. 2, the carefully and neatly done angle value of situation is selected to form the style of writing style four-tuple of this group of text：
{GZi, Vinclination, Vstrength-means, Fstroke, the style of writing style that the present invention defines user is a four-tuple：{GZi,
Vinclination, Vstrength-means, Fstroke, indicate one group of text C of userhandwritingi={ch1, ch2, …,chn?
The carefully and neatly characteristic value of angle value, amount of taper, average hand-written dynamics value and stroke.
Step 8）Such as the step 009 in Fig. 2, the script C that user is required according to style of writing style four-tuplescript=
{cs1, cs2, …,csmBe replaced.
Claims (4)
1. body method is write in a kind of personalization handwritten form replacement, one group of Chinese character of user is inputted, three kinds of setting are classified as
Situation：Situation, time not abundant situation of having plenty of time and signature situation are completed, and learn user's out using the method for deep learning
Style of writing style of one group of Chinese character, such as gradient, hand-written dynamics and the stroke feature of this group of Chinese character etc., select complete situation after with
These features form the four-tuple of the hand-written style of writing style of characterization user, according to style of writing style four-tuple analog subscriber it is given wait replace
Script text is changed, and is replaced, which is characterized in that including：
Step S1. inputs the handwritten Chinese character C of one group of userhandwriting={ch1, ch2, …, chn}；
Step S2. calculates the carefully and neatly done degree of this group of text, and distinguishing the handwritten form of user according to carefully and neatly done degree is in the feelings having plenty of time
It is completed under shape, in the case of the time is not abundant and in the case of signature；
Step S3. learns the style of writing style of this group of Chinese character out using the method for deep learning, the row including this group of text entirety
Gradient, the feature of hand-written dynamics and stroke, and the amount of taper, hand-written dynamics value and stroke feature value of this group of Chinese character are calculated separately,
The style of writing style four-tuple of user is formed with these three values after the step S2 selectes hand-written situation, i.e., carefully and neatly done degree (GZ), tiltedly
Angle value (Vinclination), hand-written dynamics value (Vstrength-means), stroke feature value (Fstroke)}；
Step S4. simulates the script Chinese character of user to be replaced according to the style of writing style four-tuple of this group of Chinese character of user
Cscript={cs1, cs2, …, csn, and replaced.
2. according to method described in claim 1, it is characterised in that step S2 includes：
The present invention sets three kinds of situations for distinguishing user's handwritten form：Situation, time not abundant situation of having plenty of time and signature feelings
Shape；
The judgement of three kinds of situations determines by carefully and neatly spending, i.e.,：
Carefully and neatly done degree GZ is more than that threshold value GZ0 indicates that handwritten form is completed in the case of having plenty of time；
Carefully and neatly done degree GZ indicates that handwritten form is completed in the case of the time is not abundant between GZ1 and GZ0；
Carefully and neatly done degree GZ indicates that handwritten form is completed in the case of having plenty of time lower than threshold value GZ1；
The calculating carefully and neatly spent is as follows：
（1）
Wherein, LBnumberIndicate the stroke number of company's pen of one group of Chinese character of user, ZBnumberIndicate the total of one group of Chinese character of user
Stroke number, u are effect length coefficients, are belonged to [0,1], indicate one group of Chinese character length to the weighing factor carefully and neatly spent, if this
The length of group Chinese character is very short, then it is assumed that is likely to be signature, coefficient value is larger, if this group of word length is more than threshold value, then it is assumed that no
It is signature, coefficient value is smaller.
3. according to method described in claim 1, it is characterised in that step S3 includes：
The calculating of one group of Chinese character amount of taper：Learn the gradient of the row of this group of text entirety out using the method for deep learning, it is hand-written
The feature of dynamics and stroke；For the gradient feature of one group of Chinese character, present invention VinclinationIt indicates amount of taper, calculates as follows：
（2）
Wherein, heightch1Indicate first Chinese character ch of this group of Chinese character1Relative altitude, heightchnIndicate this group of Chinese character most
The relative altitude of the latter Chinese character, heightchn/2Show the Chinese character relative altitude among this group of Chinese character；
The calculating of one group of Chinese character hand-written dynamics value：The hand-written dynamics value V of i-th of wordstrengthIt indicates, the hand-written power of one group of Chinese character
It is as follows to spend mean value computation：
（3）
Wherein, chiIndicate one group of Chinese character ChandwritingI-th of Chinese character；
The calculating of one group of Chinese-character stroke characteristic value：Stroke feature FstrokeIt indicates, calculates as follows：
（4）
Wherein, FkValue be 1,2,3,4,5,6,7,0, respectively indicate " just tiltedly ", " negative oblique ", " straight ", " length ", " short ", " big ",
The feature of the strokes such as " small ", " normal ", FiIt is obtained by the statistics of n Chinese character, if the F of stroke kiValue be 1 number of words more than 2
Number of words, then the feature F of stroke kiValue be 1, k=it is horizontal, vertical, skim, right-falling stroke, hook；
The style of writing style that the present invention defines user is a four-tuple：{GZi, Vinclination, Vstrength-means, Fstroke,
Indicate one group of text C of userhandwritingi={ch1, ch2, …,chnCarefully and neatly done angle value, amount of taper, average hand-written dynamics value
With the characteristic value of stroke.
4. the step S4 includes：
According to four-tuple { GZi, Vinclination, Vstrength-means, FstrokeCharacterization the style of writing style, analog subscriber
It is required that script Cscript={cs1, cs2, …,csm, and be replaced.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810657283.5A CN108830245A (en) | 2018-06-24 | 2018-06-24 | Body method is write in personalized handwritten form replacement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810657283.5A CN108830245A (en) | 2018-06-24 | 2018-06-24 | Body method is write in personalized handwritten form replacement |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108830245A true CN108830245A (en) | 2018-11-16 |
Family
ID=64138299
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810657283.5A Pending CN108830245A (en) | 2018-06-24 | 2018-06-24 | Body method is write in personalized handwritten form replacement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108830245A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1373432A (en) * | 2001-02-28 | 2002-10-09 | 曾立彬 | Method and system for recognizing personal characteristics of scrip |
| CN101308578A (en) * | 2008-06-20 | 2008-11-19 | 华南理工大学 | Beautifying method for hand-written Chinese characters |
| CN101326518A (en) * | 2005-12-13 | 2008-12-17 | 微软公司 | Script recognition for ink notes |
| CN107644006A (en) * | 2017-09-29 | 2018-01-30 | 北京大学 | A kind of Chinese script character library automatic generation method based on deep neural network |
-
2018
- 2018-06-24 CN CN201810657283.5A patent/CN108830245A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1373432A (en) * | 2001-02-28 | 2002-10-09 | 曾立彬 | Method and system for recognizing personal characteristics of scrip |
| CN101326518A (en) * | 2005-12-13 | 2008-12-17 | 微软公司 | Script recognition for ink notes |
| CN101308578A (en) * | 2008-06-20 | 2008-11-19 | 华南理工大学 | Beautifying method for hand-written Chinese characters |
| CN107644006A (en) * | 2017-09-29 | 2018-01-30 | 北京大学 | A kind of Chinese script character library automatic generation method based on deep neural network |
Non-Patent Citations (1)
| Title |
|---|
| SHINJI TSURUOKA等: "Personal Dictionaries for Handwritten Character Recognition Using Characters Written by a Similar Writer", 《IEEE》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108664924B (en) | Multi-label object identification method based on convolutional neural network | |
| CN104463209B (en) | Method for recognizing digital code on PCB based on BP neural network | |
| CN109977234A (en) | A kind of knowledge mapping complementing method based on subject key words filtering | |
| CN109492765A (en) | A kind of image Increment Learning Algorithm based on migration models | |
| CN107316066A (en) | Image classification method and system based on multi-path convolutional neural networks | |
| WO2021051598A1 (en) | Text sentiment analysis model training method, apparatus and device, and readable storage medium | |
| CN105117739A (en) | Clothes classifying method based on convolutional neural network | |
| CN109993057A (en) | Method for recognizing semantics, device, equipment and computer readable storage medium | |
| CN104834941A (en) | Offline handwriting recognition method of sparse autoencoder based on computer input | |
| US20230234235A1 (en) | Method for generating a handwriting vector | |
| CN103020167B (en) | A kind of computer Chinese file classification method | |
| CN109740451A (en) | Road scene image semantic segmentation method based on importance weighting | |
| CN104463243B (en) | Sex-screening method based on average face feature | |
| CN112115967B (en) | Image increment learning method based on data protection | |
| CN108960301A (en) | A kind of ancient Yi nationality's text recognition methods based on convolutional neural networks | |
| CN104834918A (en) | Human behavior recognition method based on Gaussian process classifier | |
| CN106844342B (en) | Term vector generation method and device based on incremental learning | |
| CN111079784A (en) | Flue-cured tobacco baking stage identification method in baking process based on convolutional neural network | |
| CN110163145A (en) | A kind of video teaching emotion feedback system based on convolutional neural networks | |
| CN110569780A (en) | high-precision face recognition method based on deep transfer learning | |
| CN106778834A (en) | A kind of AP based on distance measure study clusters image labeling method | |
| CN108830245A (en) | Body method is write in personalized handwritten form replacement | |
| CN110188662A (en) | A kind of AI intelligent identification Method of water meter number | |
| CN105335372B (en) | The equipment of the direction of document processing device, document processing, method and determining file and picture | |
| CN105184320A (en) | Non-negative sparse coding image classification method based on structural similarity |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181116 |
|
| RJ01 | Rejection of invention patent application after publication |