CN1128074A - Method of stroke segmentation for handwritten input - Google Patents
Method of stroke segmentation for handwritten input Download PDFInfo
- Publication number
- CN1128074A CN1128074A CN95190347A CN95190347A CN1128074A CN 1128074 A CN1128074 A CN 1128074A CN 95190347 A CN95190347 A CN 95190347A CN 95190347 A CN95190347 A CN 95190347A CN 1128074 A CN1128074 A CN 1128074A
- Authority
- CN
- China
- Prior art keywords
- stroke
- flexibility
- point
- derivative value
- value
- 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
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
- G06V30/347—Sampling; Contour coding; Stroke extraction
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Character Discrimination (AREA)
- Image Analysis (AREA)
- Character Input (AREA)
Abstract
The method of the present invention includes a step of calculating the derivative (140), or instantaneous rate of change, of the curvature at points in the handwritten input (110). The method then selects as stroke boundary points certain points (or pixels) in the input which lie at a midpoint between a point of high curvature derivative and a succeeding point of low curvature derivative (150). Such boundary points are not influenced by absolute curvature values, but rather only by relative changes in the curvature. The stroke segmentation boundary points are provided to a stroke-based recognizer for interpretation of the handwritten input (170).
Description
The present invention relates to handwriting recognition, be specifically related to the method for stroke segmentation for handwritten input.
Machine recognition for people's handwritten form is difficult to, and (developing rapidly of calculation element of pen-based), this identification had become a major issue already and had been suggested with the base of beginning to write recently.To this, now there have been many distinct solutions, and a kind of useful scheme wherein is that handwritten form is divided into the basic traverse of a sequence or " stroke ", and with the input of these strokes (with them according to certain mode parametrization) as Character recognizer.
A key request that with the stroke is based recognizer is that the multiple situation (for example different number of times and the different letters that the writer write " A ") of same character class all should be divided in the similar group of strokes at every turn.This helps to guarantee that identification is not too because of difficulty, because the description of the multiple situation of this character " it seems " it is similar cause for Character recognizer itself.In the ideal case, all of a given character are write the stroke that situation always contains similar number, and these strokes all are positioned on the identical relative position, and the feature description of these strokes is all very similar for various situations.Though this ideal situation can not realize in practice, to a certain extent can be near it, thus can improve the accuracy of identification.
It is that the stroke border is set in vertically nib speed on (or " y ") direction is on some points of zero that a kind of recognition technology is arranged in prior art, just, writes to begin to move up or begin on these aspects and moves down.Formed like this group of strokes can be described as " stroke that makes progress " and " downward stroke ".This method is at Mermelstein ﹠amp; Eden, Information And Control Vol.7 has discussed in the article that PP.255-270,1964 delivers " Experiments on Computer Recognition of ConnectedHandwritten Words ".A problem of this method is in vertical direction stroke to be changed extrasensitivity, and in the horizontal direction stroke is changed inresponsive.Yet many characters are made up of some horizontal lines, for example, three horizontal strokes of the horizontal and letter " E " of letter one in " t ", even the writing of carelessness, they under normal circumstances also are that mostly be level rather than vertical.Y velocity profile stroke segmentation device is stroke because of the shake of nib in vertical direction in writing makes cross-section being split into of level sometimes, and often it is broken into two sections, three sections even multistage more.This causes the identification accuracy of difference, because the multiple font situation of same character often is segmented into the group of strokes for seeming to have nothing in common with each other.The effort of doing for the inaccuracy of this method of correction is included in the vertical direction variation that produces new stroke requirement minimum before, and this has obtained limited success, and many same basic problems still exist.
In the existing technology of another kind, this problem is, the stroke border solves by setting on some points like this, promptly occurs local maximum deflection on such point, and surpasses certain and be listed as crooked threshold value by force corresponding to this clerical type.May take place with the moving direction of nib because of strong row are crooked irrelevant, so this method is to the handwriting input orientation-insensitive of the each several part of statement or character for example.Yet, it self problem is also arranged with the technology that is bent into the basis.For example, suppose that someone has a very bending of gradual change when writing letter " L ", rather than the bending of strong row, so that it begins to seem that it more resembles letter " C ".In this case, if the required flexibility threshold value in stroke border can not satisfy, then this method can not segmentation.Reducing threshold value simply can not address this problem, because this can merely cause too much number of strokes.Has too much additional strokes and to have very few stroke all bad, because this multiple handwriting that means same character again is segmented into different stroke types often.
In view of the above, the stroke segmentation technology that needs a kind of problem that more accurately, exists in no above-mentioned method such as y tachometric method and the existing flexibility method now.
For this reason, the invention provides a kind ofly becomes the method for a plurality of strokes with the handwriting input segmentation, for the multiple handwriting of each concrete character class, can both segmentation go out consistent number of strokes.
The invention provides and a kind of handwritten word is imported the method that segmentation becomes a plurality of strokes, have similar shape and position for the multiple handwriting segmented stroke of each concrete character class of input.
Generally speaking, method of the present invention comprises the derivative of the flexibility on the each point in the calculating handwriting input or the step of instantaneous rate of change.Then, this method selects some point (or pixel) in handwriting input as the stroke frontier point, and they are between a high flexibility derivative point and a follow-up low flexibility derivative point.Such frontier point is not subjected to the influence of absolute flexibility value, but the only degree influence of variation relatively by bending.
Fig. 1 illustrates the process flow diagram in order to the operation of identification stroke boundary according to a preferred embodiment of the present invention;
Fig. 2 illustrates the example that becomes a plurality of strokes with the handwriting input segmentation of the y speed method generation of prior art;
Fig. 3 illustrates the example that the handwriting input segmentation that produces with flexibility method in the prior art becomes a plurality of strokes;
Fig. 4 illustrates the example that the handwriting input segmentation that produces with the preferred embodiments of the present invention becomes a plurality of strokes;
Fig. 5 illustrates the point of foundation from the letter " L " of digitalizer reception;
Fig. 6 illustrates according to the preferred embodiments of the present invention back of taking a sample again and sets up the point of letter " L " according to constant distance;
Fig. 7 illustrates the exploded view of the flexibility calculating of the preferred embodiments of the present invention;
Fig. 8 illustrates the curve map of the flexibility value that calculates for Fig. 7 each point;
Fig. 9 illustrates the curve map of the flexibility derivative value that calculates for Fig. 7 each point.
The hand-written character input is normally collected with the form of discrete continuous segment from the user.A discrete continuous segment is made up of one or more strokes, and wherein, stroke is a pen and the left trace of input media period of contact of a digitizing figure input card or paper and so on.
In the present invention, one or more discrete continuous segments are a plurality of unit of the handwriting input that is identified.Handwriting input is the input that electricity is caught, and it includes but not limited to following input: handwriting input; The electronics input; The input that pressure is caught, for example Ya Yin input; The input that electronic method for example receives with facsimile recorder, pager or other device.
Stroke can be expressed as into by input media with the point of a sequence of sampling at interval clocklike roughly.Each available at least x of point and y coordinate are described.Stroke can utilize a digitizing figure input card to catch with electrical method, or obtains from image scanning or fax in the testing process of lines in image; The electricity method of catching is known in the present technique field like this.In a method for optimizing, handwriting input is received by a device such as PDA(Personal Digital Assistant) device or other device.Having other device that receives hand-write input function includes but not limited to lower device: computing machine, modulator-demodular unit, pager, telephone set, digital television, interactive television, the device that has digitizing figure input card, facsimile unit, scanister and have other device of catching the handwriting input ability.Usually, when stroke was caught with electrical method, each o'clock was represented with a pixel, so that a stroke can be represented by a series of pixels on this device.
According to the present invention, handwriting input can be other form of character or symbol in character style, ideographic character or the hand-written communication of letter.
Referring to accompanying drawing, Fig. 2 and Fig. 3 illustrate: when stroke segmentation was based recognizer by one with the stroke, the stroke segmentation of alphanumeric handwriting input had the inaccuracy of high probability in to the understanding of input.Fig. 4 illustrates the stroke segmentation for the identical alphanumeric input of Fig. 2 and Fig. 3, and stroke segmentation is to carry out such stroke segmentation among the figure in accordance with the teachings of the present invention by an accuracy that has high probability with the stroke during for based recognizer.
Referring to Fig. 1, the process flow diagram of a kind of method for optimizing of instructing according to the present invention shown in the figure.The handwriting input that comes from digitalizer or other device with the form of x and y coordinate (together with relevant nib move or nib under shifting state) be received in step 110.These points are represented by pixel usually.Generally speaking, method of the present invention is taken a sample in step 120 pair handwriting input again, to obtain along the point of handwriting input length equidistantly to leave.Fig. 5 illustrates letter " L " 500 as before taking a sample again, the example of series of points or pixel.Fig. 6 illustrates series of points or the pixel of same letter " L " 600 after taking a sample again.Sampling is to use idea spacing d610 to realize that the d value is constant in whole handwriting input again.The d value is preferably selected to such an extent that make the middle input of handwriting input highly be about 15 to 30 times of d values.For example, d value shown in Figure 6 is selected to such an extent that can make the intermediate altitude of letter greatly in the scope of 15 to 30 times of d values.
The preferred embodiment of Fig. 1 calculates each flexibility at sampling spot place again in step 130.Fig. 7 is illustrated in a R (710) and locates the data that flexibility calculates and describe.The distance that is defined as in the flexibility of sampling spot R (710) again, from more preceding (R-1,750) that this R is ordered obtain through the linear projection of R own 1 P (720) to R order after a bit (put R+1,730) spacing.This distance spacing 740 as shown in Figure 7.Flexibility at handwriting input end points place is defined as the flexibility at the nearest neighbor point place that equals corresponding.The flexibility at the internal point place of handwriting input also can be calculated, from 2 of ordering away from R rather than a spot projection (with preceding 2 that of using that R orders rather than R-1 point), to obtain a stronger valuation.Fig. 8 illustrates the curve map of the flexibility value that each point shown in Figure 7 obtains.
For example, two " straight line " part that the letter " L " with bending gradually of last example needs only " L " is more straight than the bending between them significantly, just can segmentation become a vertical stroke and a horizontal strokes.So flexibility increases (being that the flexibility derivative uprises) in the trend corner, reduces (flexibility derivative step-down) away from the corner flexibility, thereby, as desired, a stroke border can be arranged in the corner or near the corner.
In a preferred embodiment of the invention, in case be each sampling spot acquisition flexibility again, the flexibility array of sampling spot just can be smoothed again, so that any known artifacts that digitized process is introduced reduces to minimum.Performed level and smooth type should be a kind of standard method, can select according to existing concrete digitizing characteristic.This can comprise a point and its neighbor point are averaged (weighting to this point itself and closest approach gets higher), replaces the flexibility value of the point of being analyzed with the mean value that calculates.The size of used here smooth window ideally should be in that to be wider than hand-written low flexibility place wideer, and narrower in high flexibility part, so that losing of important information reduces to minimum in smoothing process.Because of this is that flexibility itself is smoothed, so a kind of level and smooth optimal way is to calculate the initial bending degree, carry out smoothly according to those flexibility, carry out again level and smooth according to the new flexibility that calculates again then.
In a preferred embodiment of the invention, for each sampling spot again, utilize any hogging bending degree on duty with-1 absolute value that calculates flexibility.In calculating the flexibility derivative, preferably use its absolute value and without actual flexion degree value, because the preferred embodiment of the inventive method is only considered the strong row that turn round in the handwriting input rather than for a given crooked direction of turning round.
As shown in Figure 1, this method is then calculated each flexibility derivative at sampling spot place again in step 140.Referring to Fig. 7, the flexibility derivative at some R place is defined as an absolute value of R+1 place flexibility and deducts behind the absolute value of R-1 place flexibility again divided by 2 (also being drawn flexibility value slope of a curve).Fig. 9 illustrates the curve map of the derivative of the flexibility that obtains on the every bit shown in Figure 7.Be similar to above-mentioned application more than two point try to achieve the tolerance of flexibility more accurately, the flexibility derivative should adopt the window (to 3 o'clock) of broad to calculate at 5 o'clock when having reference point, and also adopts narrower window (to 3 o'clock) at 2 o'clock where necessary.Because of the derivative of flexibility can not calculate at the end points place of ink marks section, the flexibility derivative of Gu Duandianchu can be defined as the flexibility derivative at the neighbor point place that equals corresponding simply.
Referring to Fig. 1 and Fig. 9, next step the processing of the preferred embodiment of this method is in step 150, check the array of the new flexibility derivative value of calculating, limit the position at the some place that comprises that inflection end and maximal value trend reduce with the local maximum (910) of seeking the flexibility derivative, or the local minimum of flexibility derivative (920) limits the position at the some place that comprises that inflection end minimum value trend increases.For after the local maximum and each part of (in time) handwriting input before the local minimum, seek out the mid point (930) mid point of part arc length (should) of this part.This mid point is defined as M point (930).In step 150,, then put M and be selected as a stroke border if the flexibility absolute value that the difference of local maximum and local minimum surpasses a threshold value T (940) and M point place for certain part surpasses certain threshold value T ' (820).
Must be to parameter T ' and T " valuation, they are relevant with the unit that flexibility and flexibility derivative are measured.As long as the Character recognizer of use error tolerance limit, T ' and T " exact value not strict.For " or in any experimental coordination of any other parameter, desirable target is all its segmentations of the multiple handwriting segmentation of unanimity as much as possible that reaches concrete character class in order to produce performed T ' and the T of a specific embodiment of the present invention.This should be experimentally how segmentation realizes to the sample of the various reality of the handwriting input that will discern by checking this program.
Except above-mentioned selected stroke frontier point, the point that every nib is mentioned or nib falls also all is chosen to be the stroke frontier point.In a preferred embodiment of the invention, in step 160,, be on the peaked point so that they drop on the flexibility absolute value for many like that based on removable two points of the frontier point of flexibility derivative.Utilize the frontier point of mobile flexibility derivative, flexibility is measured produced identical stroke border, improved the location of stroke frontier point with the flexibility derivative for the basis.Yet advantageous applications of the present invention just could be used a kind of given point, but the stroke that is less than at 3 that produces does not have.
According to the present invention, stroke frontier point group defines the group of strokes of a correspondence.These strokes are sent to a Character recognizer based on stroke, so that discern this handwriting input.
That invention and preferred embodiments thereof relate to is novel, stroke segmentation method more accurately.According to the present invention, under the multiple handwriting situation of handwriting input, this input repeatedly is divided into similar group of strokes each time.For example, if this handwriting input letter is with the different letters that writes out " L " by different writers, then invention and preferred embodiments thereof can be more at every turn should letter " L " input be divided into similar stroke segmentation frontier point, and irrelevant with different Writer's difference.This stroke segmentation helps stroke is explained for based recognizer provides more accurate stroke.
Those skilled in the art finds that many embodiment of the present invention are useful.A kind of tangible expansion is to write to cursive script from the handwritten form of printing described herein.The practical methods of stroke segmentation is irrelevant with the method for character segmentation, thereby can handle the technology that cursive script writes and can utilize stroke segmentation method described here easily.Another embodiment makes segmentation scanning or that " off line " writes become stroke.With using the straightforward method of the present invention in this task is to carry out the finishing that attenuates of clerical type, to obtain the ink marks curve of constant width.So, make the point of crossing seem similar because of lacking temporary transient information, so can be to set the stroke border on the point on basis and the point of crossing at the flexibility derivative with tangent turning round.
Claims (8)
1. one kind contains the method for the hand-written character of a plurality of ink marks pixels in order to identification, it is characterized in that may further comprise the steps:
Each pixel to a plurality of ink marks pixels is calculated the flexibility derivative value, and each flexibility derivative value is represented the rate of change of the absolute flexibility at respective pixel place;
Select one group of stroke border, so that each stroke boundary is between an ink marks pixel with high flexibility derivative value and ink marks pixel with low flexibility derivative value subsequently;
Seek the position of one group of stroke. so that each stroke border is positioned at the end of a stroke;
Calculate at least a stroke feature value of each stroke, to produce a character feature collection;
Use this character feature collection and judge identification described hand-written character.
2. according to claim 1 method, it is characterized in that the ink marks pixel with height flexibility derivative value has the maximum flexion derivative value of a part, and the ink marks pixel subsequently with low flexibility derivative value has the minimum bend degree derivative value of a part.
3. according to the method for claim 2, it is characterized in that each stroke border is in the ink marks pixel with local maximum flexion derivative value and have on the mid point between the ink marks pixel of local minimum bend degree derivative value.
4. according to the method for claim 1, it is characterized in that each stroke border is positioned on the point with local maximum absolute flexibility value.
5. the method for a hand-written character of forming by a sequence of points in order to identification, wherein each point comprises three air coordinates values, it is characterized in that this method may further comprise the steps:
Calculate the flexibility derivative value of each point of a plurality of points, each flexibility derivative value is expressed the rate of change of absolute flexibility on respective point;
Select one group of stroke border, so that each stroke border is between a point with high flexibility derivative value and one 's the point that has low flexibility derivative value subsequently;
Seek the position of one group of stroke, so that each stroke border is positioned at the end of a stroke;
Calculate at least one stroke feature value of each stroke, to produce a character feature collection;
Use this character feature collection and judge identification described hand-written character.
6. according to the method for claim 5, it is characterized in that the point with high flexibility derivative value has local maximum flexion derivative value, the low flexibility derivative value point that has subsequently has local minimum bend degree derivative value.
7. according to the method for claim 6, it is characterized in that each stroke border is on the mid point between a point with local maximum flexion derivative value and the point with local minimum bend degree derivative value.
8. according to the method for claim 5, it is characterized in that each stroke border is positioned at one to have on the local maximum absolute point that becomes curvature value.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US24040794A | 1994-05-10 | 1994-05-10 | |
| US08/240,407 | 1994-05-20 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN1128074A true CN1128074A (en) | 1996-07-31 |
Family
ID=22906392
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN95190347A Pending CN1128074A (en) | 1994-05-10 | 1995-05-03 | Method of stroke segmentation for handwritten input |
Country Status (14)
| Country | Link |
|---|---|
| EP (1) | EP0710384A4 (en) |
| JP (1) | JP2002515144A (en) |
| CN (1) | CN1128074A (en) |
| AU (1) | AU2431695A (en) |
| BR (1) | BR9506197A (en) |
| CA (1) | CA2162489A1 (en) |
| CZ (1) | CZ6196A3 (en) |
| FI (1) | FI960110A (en) |
| HU (1) | HUT75820A (en) |
| IL (1) | IL113659A0 (en) |
| NO (1) | NO955064D0 (en) |
| PL (1) | PL312469A1 (en) |
| SK (1) | SK3096A3 (en) |
| WO (1) | WO1995032485A1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1317664C (en) * | 2004-01-17 | 2007-05-23 | 中国科学院计算技术研究所 | Confused stroke order library establishing method and on-line hand-writing Chinese character identifying and evaluating system |
| CN100338621C (en) * | 2005-04-07 | 2007-09-19 | 上海交通大学 | Eigenvalue error compensation on limited sample collection and parameter distribution correcting method |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB9701793D0 (en) | 1997-01-29 | 1997-03-19 | Gay Geoffrey N W | Means for inputting characters or commands into a computer |
| US6563515B1 (en) | 1998-05-19 | 2003-05-13 | United Video Properties, Inc. | Program guide system with video window browsing |
| US7603684B1 (en) | 1998-05-19 | 2009-10-13 | United Video Properties, Inc. | Program guide system with video-on-demand browsing |
| EP1562138B1 (en) * | 2004-02-06 | 2009-08-19 | Dassault Systèmes | A process for drafting a curve in a computer-aided design system |
| EP1562130A1 (en) * | 2004-02-06 | 2005-08-10 | Dassault Systèmes | A process for modifying a curve in a computer-aided design system |
| CN100405389C (en) * | 2004-08-06 | 2008-07-23 | 摩托罗拉公司 | Identifying character from stroke mark |
| JP2006162692A (en) * | 2004-12-02 | 2006-06-22 | Hosei Univ | Automatic lecture content creating system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2658137B2 (en) * | 1988-03-11 | 1997-09-30 | 沖電気工業株式会社 | Character recognition method |
| JP3017740B2 (en) * | 1988-08-23 | 2000-03-13 | ソニー株式会社 | Online character recognition device and online character recognition method |
| US5590220A (en) * | 1992-08-12 | 1996-12-31 | International Business Machines Corporation | Bending point extraction method for optical character recognition system |
-
1995
- 1995-05-03 HU HU9503882A patent/HUT75820A/en unknown
- 1995-05-03 CA CA002162489A patent/CA2162489A1/en not_active Abandoned
- 1995-05-03 SK SK30-96A patent/SK3096A3/en unknown
- 1995-05-03 EP EP95918350A patent/EP0710384A4/en not_active Withdrawn
- 1995-05-03 CZ CZ9661A patent/CZ6196A3/en unknown
- 1995-05-03 BR BR9506197A patent/BR9506197A/en unknown
- 1995-05-03 CN CN95190347A patent/CN1128074A/en active Pending
- 1995-05-03 JP JP53030695A patent/JP2002515144A/en active Pending
- 1995-05-03 AU AU24316/95A patent/AU2431695A/en not_active Abandoned
- 1995-05-03 PL PL95312469A patent/PL312469A1/en unknown
- 1995-05-03 WO PCT/US1995/005409 patent/WO1995032485A1/en not_active Application Discontinuation
- 1995-05-08 IL IL11365995A patent/IL113659A0/en unknown
- 1995-12-14 NO NO955064A patent/NO955064D0/en unknown
-
1996
- 1996-01-10 FI FI960110A patent/FI960110A/en not_active Application Discontinuation
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1317664C (en) * | 2004-01-17 | 2007-05-23 | 中国科学院计算技术研究所 | Confused stroke order library establishing method and on-line hand-writing Chinese character identifying and evaluating system |
| CN100338621C (en) * | 2005-04-07 | 2007-09-19 | 上海交通大学 | Eigenvalue error compensation on limited sample collection and parameter distribution correcting method |
Also Published As
| Publication number | Publication date |
|---|---|
| BR9506197A (en) | 1996-04-24 |
| IL113659A0 (en) | 1995-08-31 |
| HUT75820A (en) | 1997-05-28 |
| WO1995032485A1 (en) | 1995-11-30 |
| AU2431695A (en) | 1995-12-18 |
| EP0710384A4 (en) | 1997-05-02 |
| EP0710384A1 (en) | 1996-05-08 |
| FI960110A0 (en) | 1996-01-10 |
| MX9600189A (en) | 1998-11-29 |
| PL312469A1 (en) | 1996-04-29 |
| CZ6196A3 (en) | 1996-06-12 |
| NO955064L (en) | 1995-12-14 |
| HU9503882D0 (en) | 1996-02-28 |
| NO955064D0 (en) | 1995-12-14 |
| CA2162489A1 (en) | 1998-06-01 |
| JP2002515144A (en) | 2002-05-21 |
| SK3096A3 (en) | 1996-10-02 |
| FI960110A (en) | 1996-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1121662C (en) | Method and microprocessor for preprocessing handwriting having characters compesed of preponderance of straight line segments | |
| CN1035904C (en) | Estimation of baseline, line spacing and character height for handwriting recognition | |
| JP4787275B2 (en) | Segmentation-based recognition | |
| US7302099B2 (en) | Stroke segmentation for template-based cursive handwriting recognition | |
| US6603881B2 (en) | Spatial sorting and formatting for handwriting recognition | |
| US7394934B2 (en) | Recognition of electronic ink with late strokes | |
| US7256773B2 (en) | Detection of a dwell gesture by examining parameters associated with pen motion | |
| US7349576B2 (en) | Method, device and computer program for recognition of a handwritten character | |
| US7437001B2 (en) | Method and device for recognition of a handwritten pattern | |
| Feldbach et al. | Line detection and segmentation in historical church registers | |
| US20080240569A1 (en) | Character input apparatus and method and computer readable storage medium | |
| US7369702B2 (en) | Template-based cursive handwriting recognition | |
| CN1128073A (en) | Method for recognizing handwritten input | |
| US20110229038A1 (en) | Feature Design for HMM Based Eastern Asian Character Recognition | |
| El Abed et al. | Comparison of different preprocessing and feature extraction methods for offline recognition of handwritten arabicwords | |
| EP1854048A1 (en) | Recognition graph | |
| JPH0844826A (en) | Collating method of handwritten input | |
| EP1553522A2 (en) | Determining positions of images of a stroke | |
| CN1128074A (en) | Method of stroke segmentation for handwritten input | |
| Mandal et al. | Slant Estimation and Correction for Online Handwritten Bengali Words | |
| JP2000251013A (en) | Method and device for character recognition and storage medium | |
| MXPA96000189A (en) | Method of segmentation of traces for the entry of characters manuscri | |
| JP2633523B2 (en) | Handwriting input device | |
| JP6437208B2 (en) | Handwritten music symbol recognition apparatus and handwritten music symbol recognition program | |
| JPH11167606A (en) | Handwritten character recognizing device and program storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C01 | Deemed withdrawal of patent application (patent law 1993) | ||
| WD01 | Invention patent application deemed withdrawn after publication |