CN1093965C - On-line character recognition method and apparatus thereof - Google Patents

Abstract

The present invention adopts the input unit with the writing information of the input characters, and selects the font recognition unit of a plurality of candidate characters with high similarity from the standard characters stored in the font library according to the general shape characteristics of the written characters. Stroke feature is carried out with regard to the candidate characters selected by the above-mentioned font recognition unit. Consider the corresponding naming of the stroke order. A stroke number name recognition unit for recognizing the corresponding name of the stroke number, and a display unit for displaying the recognition result of the stroke number corresponding name recognition unit or the stroke number corresponding name recognition unit.

Description

Online character recognition method and online character recognition device

The invention relates to an online character recognition method and device for high-precision recognition of handwritten characters.

Conventional online character recognition devices include an online character recognition device that recognizes characters by absorbing stroke-order changes, and an online character recognition device that recognizes characters by absorbing stroke-order changes. Various online character recognition devices are described below.

Existing technology example 1:

In the existing online character recognition device that absorbs and recognizes characters by absorbing stroke order changes, it is between all the strokes of the standard font of the font and all the strokes of the input font to obtain the distances of the respective combinations of the strokes. Among the strokes of the font, the strokes of the input font with the smallest distance are named accordingly, and the stroke changes of the input font are processed accordingly (Japanese patent application Sho 54-061146), or information such as the method of changing the stroke order is stored in the standard In the font, according to the stored content, the corresponding naming of the strokes of the input font and the standard font is limited, so as to reduce the processing amount of the corresponding naming (Japanese Patent Publication Sho 57-178579).

For example, FIG. 26 is a block diagram showing the structure of a conventional online character recognition device disclosed in Japanese Patent Laid-Open No. Sho 57-178579 for absorbing stroke order changes.

In Fig. 26, 1 is the font matching device that calculates the distance between the strokes of the input font and the strokes of the standard font, 2 is the standard font storage device that pre-stores the stroke order of the standard font of each character, and 3 is the combination of strokes stored Stroke-order combination storage devices such as methods, 4 is a standard font reconstruction device for changing the stroke order of a standard font stored in the standard font storage device 2 according to the information of the stroke-order combination storage device 3, and 5 is a reforming device according to the font The distance output by the matching device 1 determines the recognition device of the recognition result.

Hereinafter, the operation of the online character recognition device of the conventional example using FIG. 26 will be described.

First, the input font information X is input in the font matching device 1 . The input font information X is composed of n strokes in the case of n-stroke characters. If the stroke order of each stroke is set as X1, X2, ... Xn, it will be expressed in the multivariate vector space as

X = (X1, X2, ... Xn).

In addition, if the standard font formed by n strokes stored in the standard font storage device 2 is SK (K represents the classification name of the standard font), the stroke order of the various strokes of SK is SK1, SK2, ... SKn , then the standard font SK is represented in the multivariate vector space as

      SK = (SK1, SK2, ... SKn)

Next, the font matching device 1 calculates the distance between the standard font SK and the input font X, but because the input font X is not necessarily recorded in the correct stroke order, it simply compares each element of the input font X with the The element order of the standard font corresponds to the naming, and it is impossible to correctly identify characters written in a stroke order different from that of the standard font.

In addition, the standard font reconstruction device 4 converts the stroke order to form the standard font SKm (m=1, 2, ..., p) according to the stroke combination information stored in the stroke combination storage device 3, and sequentially sends them to the font matching device 1. Here, the total number of combinations of stroke order changes of the standard font of p.

Next, the font matching device 1 calculates the distance Pxkm from the input font X for the standard fonts sequentially sent from the standard font reforming device 4 using the following formula.

Pxkm＝|X1-Skm1|+|X2-Skm2|+...+|Xn-Skmn|

Then, the recognition device 5 recognizes the minimum distance within the P distances sent by the font matching device 1, and uses the minimum distance as the distance Pxk between the input font X and the standard font Sk. Perform the above processing on the standard fonts with n strokes in all categories stored in the standard font storage device, and finally output the category with the smallest distance as the recognition result.

Existing technology example 2

In the existing online character recognition device that recognizes characters by absorbing changes in the number of strokes, there are multiple feature points of the strokes of the input font and the stroke feature points of the standard font to perform DP matching, and the corresponding naming of the features is performed so as to form a minimum Distance naming (Patent Publication No. 62-62394), or, in order to improve the recognition accuracy of fonts with stroke count changes, stroke connection methods are described in standard fonts (Patent Publication No. 60-79485, Patent Publication No. 60-136783) etc.

For example, FIG. 28 is a block diagram of an existing online character recognition device that absorbs changes in the number of strokes described in Patent Publication No. 62-62394. FIG. 27 is a diagram for explaining the recognition sequence of a conventional online character recognition device for absorbing changes in the number of strokes.

Fig. 27(a) shows the standard font, and Fig. 27(b) shows the input font. Fig. 27(c)(D) illustrates the corresponding naming of each stroke of the standard font and each stroke of the input font.

In Fig. 27, 30-39 represent the strokes of the input font, and 10-21 represent the strokes of the standard font.

And in Fig. 28, 40 is the stroke number detection part, 41 is font address generating part, 42 is font, 43 is feature extracting part, 44 is DP matching part, 46 is judging part, 47 is range specifying part.

Hereinafter, the operation of the conventional online character recognition device for absorbing fluctuations in the number of strokes will be described with reference to FIGS. 27 and 28 .

First, the input font is input to the stroke count detecting unit 40 and the feature extracting unit 43, and the stroke count and the feature points of each stroke are extracted. The number of strokes (n+α) to (n−β) to be compared is output to the font library 42 by the stroke count detection unit 40 with respect to the stroke count n of the input font. In the case of the "系" character shown in Fig. 27(b), because n=5, the stroke number range of α=2, β=1 is 7 strokes～4 strokes.

The input font like Fig. 27 (b) is 5 strokes, forms 10 characteristic points, and standard font is 6 strokes, forms 12 characteristic points. However, if the corresponding naming is simply carried out from the beginning, the feature numbers which have not been correspondingly named as shown in Fig. 27(c) are left over, and the resulting distance value becomes larger because the corresponding strokes are not correctly named.

Therefore, in this prior art example, DP (Dynamic Programming) matching is adopted, and those with different numbers of strokes are correspondingly named according to the smallest distance. Here, specific DP matching is performed as follows.

Each feature point of the input font is expressed as follows:

A＝a1，a2，……，ai，…aI

The characteristic points of the standard font of the standard font library are expressed as follows:

B = b1, b2, ..., bj, ..., bJ

In this case, the distance d(i, j) between ai and bj can be obtained as follows. Let d(i, j)=|ai-bj| then (1) When i=1, j=1,

When g(1,1)=d(1,1)(2) i=1, j≠1

g(1, j)=g(1, j-1)+d(1, j)(3) when i≠1, j=1

g(i,1)=g(i-1,1)+d(i,1)(4) Other cases

Calculate g(I, J) (initial condition) from the above-mentioned recursive formula, and use the following formula to calculate the distance S(A, B) between the feature point sequence of A and the feature point sequence of B. Then, take the one with the smallest distance as the recognition result.

S(A, B)＝1/(I+J)*g(I, J)

The DP matching unit 44 sequentially reads the standard fonts in the stroke count range from the font library 42, uses the above-mentioned DP matching method to find the distance between the feature points of the input font and the feature points of the standard font, and sends it to the judging unit 46. The determination part 46 outputs the character whose distance is the smallest among the said distances as a recognition result.

In this example, even if it is an input font like FIG. 27(b), the corresponding naming of the feature points is correctly performed as shown in FIG. 27(d), and the corresponding naming distance is smaller than that of FIG. 27(c).

As explained above, with respect to all combination calculation distances of standard fonts and input font styles, the online character recognition device of example 1 of the prior art example 1 that absorbs stroke order changes, will calculate distances for all combinations, and the amount of calculation is large, and it is difficult to carry out real-time recognition. difficulty.

In addition, the on-line character recognition device described in the prior art example 1 of the corresponding naming method of the stroke order in the standard font can only process the stroke order changes prepared in advance in the standard font. In order to improve the accuracy of the corresponding naming, corresponding Setting many strokes has the problem of making the font size larger.

In addition, in the device for responding to stroke order changes in the prior art example 1, it can cope with stroke order changes but not stroke number changes, and there is a problem that the recognition accuracy of connected characters and the like is reduced.

Equally, in the on-line character recognition device of prior art example 2 such as prior art example 2 that carries out the correspondence naming of different fonts with DP matching, in order to obtain the distance of 1 standard font n strokes and input font m strokes, must carry out The calculation of the distance of n×m should be performed on all standard fonts in the range of strokes (m+d)～(n-β). Here, if considering the occasion of inputting characters with 11 strokes as the input characters, then m=11, for example, if α=1 and β=1 are set, then the distance calculation must be carried out with respect to standard fonts with 10 strokes to 13 strokes. Even if it is limited to JIS level 1 characters, there are 289 characters in 10 strokes and 289×10×11=31790 distance calculations, and 300 characters in 11 strokes need to be calculated 300×11×11=36300 times. Calculate, 12 strokes have 297 words to do the distance calculation of 297×12×11=39204 times, and 13 strokes have 242 words to do the distance calculation of 242×13×11=34606 times. For this reason, a total of 141,900 distance calculations are required, and the amount of calculation is very large, making it difficult to recognize in real time.

Furthermore, an online character recognition device such as prior art example 2, which has information such as a combination method in the font library, has difficulty in forming a font library with information on the connection method of each stroke of each character. And the font capacity must be very large. Furthermore, there is a problem that the recognition accuracy is lowered because it cannot correspond to consecutive characters that have not been prepared in advance.

In addition, the character recognition device corresponding to the variation of the number of strokes in the prior art example 2 can only respond to the variation of the number of strokes but not the variation of the stroke order, and there is a problem that the recognition accuracy of characters with variation in the stroke order is reduced.

Moreover, with the above-mentioned character recognition devices of prior art examples 1 and 2, the stroke information of characters is actively used for recognition, but for some characters with continuous strokes or changes in the stroke order, although the overall shape of the characters is beautiful, it is difficult to read with high precision. .

Regardless of the quality of the characters, for recognition processing, even in the case of writing in relatively small places, the same processing time is required as for characters written with pollution. long time problem.

The object of the present invention is to solve related problems and obtain an online character recognition method and an online character recognition device that can realize high-precision recognition and high-speed recognition.

Specifically, the first purpose is to perform high-speed and high-precision recognition by reducing the amount of calculations for recognition of stroke-order-corresponding names.

In addition, it is the second object to perform high-speed and high-precision recognition by reducing the amount of calculation for recognition of stroke-number-corresponding names.

In addition, in addition to high-speed recognition, it is also possible to perform corresponding processing on stroke-order changes and stroke-number changes to perform high-precision recognition. This is the third purpose.

A fourth object is to perform recognition with excellent accuracy even when the stroke order of written characters is different from that of standard characters.

The fifth object is to allow characters with a high probability of correct interpretation to respond to soft stroke changes, and characters with a low probability of correct interpretation to be recognized at high speed and high accuracy without excessive stroke changes.

The sixth is to output the recognition result at high speed when the recognition result of font recognition is good.

In addition, it is the seventh object to prevent the recognition rate from deteriorating because the corresponding name cannot be realized in actual writing.

In order to achieve the above purpose, the online character recognition method invented in the first point includes:

A writing information input step of inputting the written information with the writing action of the characters;

Extracting the feature representing the overall shape of the above-mentioned written character from the above-mentioned writing information, and selecting a plurality of font recognition steps similar to the standard characters of the above-mentioned written character from the font library according to the feature representing the above-mentioned overall shape;

Extract the first stroke feature that represents the stroke feature that constitutes the above-mentioned written character from the above-mentioned writing information, consider the stroke order of the above-mentioned written character and perform the stroke and the above-mentioned font recognition step of the above-mentioned written character according to the above-mentioned first stroke feature. The corresponding naming of the strokes of the standard characters, calculate the corresponding naming recognition step of the stroke sequence of a plurality of standard characters selected in the above-mentioned font recognition step relative to the similarity of the above-mentioned written characters, and,

The first output step of outputting the standard characters selected in the above-mentioned font recognition step according to the said similarity calculated in the above-mentioned stroke-order correspondence naming recognition step.

Here, the written information input step corresponds to the step of writing characters on a sign or the like in an embodiment described later (S1 in FIG. 2 ). The first stroke feature is equivalent to the direction from the start point to the end point of the character stroke, the width and height of the circumscribed rectangle of the stroke in the following embodiments. The degree of similarity between the plurality of standard characters selected in the above-mentioned font recognition step and the above-mentioned written characters is equivalent to the distance between the standard characters and the written characters in the embodiment described later.

The online text recognition method of the second invention includes:

The writing information input step of inputting the written information along with the writing action;

Extracting the features representing the overall shape of the above-mentioned written characters from the above-mentioned writing information, and selecting a plurality of standard characters similar to the above-mentioned written characters from the font library according to the features representing the above-mentioned overall shape;

Extract the second stroke feature representing the stroke feature that constitutes the above-mentioned written character from the above-mentioned writing information, and according to the above-mentioned second stroke feature, consider the number of strokes of the above-mentioned written character to perform the stroke of the above-mentioned written character and the stroke of the standard character selected in the above-mentioned font recognition step The corresponding naming, calculate the number of strokes corresponding to the naming recognition step of a plurality of standard characters selected in the above-mentioned font recognition step relative to the similarity of the above-mentioned written characters, and

The second output step of outputting the standard character selected in the above-mentioned font recognition step based on the above-mentioned similarity calculated in the above-mentioned stroke-number-corresponding-name recognition step.

Wherein, the second stroke feature is equivalent to the stroke width, height, etc. in the embodiments described later.

The online text recognition method invented by the third point includes:

The writing information input step of inputting the writing information accompanying the writing action;

Extracting the features representing the overall shape of the above-mentioned written characters from the above-mentioned written information, and selecting a plurality of standard characters similar to the above-mentioned written characters from the font library according to the above-mentioned overall features;

Extract the feature representing the first stroke that constitutes the stroke feature of the above-mentioned written character from the above-mentioned written information, and according to the above-mentioned first stroke feature, consider the stroke order of the above-mentioned written character to carry out the comparison between the stroke of the above-mentioned written character and the standard character selected in the above-mentioned font recognition step The corresponding naming of strokes, calculating the corresponding naming recognition step of the stroke sequence of a plurality of standard characters selected by the above-mentioned font recognition step relative to the similarity of the above-mentioned written characters;

A first discrimination step of judging whether to output the above-mentioned standard characters by comparing the similarity calculated in the above-mentioned stroke-order correspondence recognition step with a predetermined value;

When being judged as output in the above-mentioned first discrimination step, the first output step of outputting the standard characters selected in the above-mentioned font recognition step according to the above-mentioned degree of similarity calculated in the above-mentioned stroke-corresponding naming recognition step;

When it is judged not to output in the above-mentioned first judging step, a second stroke feature representing a stroke feature constituting the above-mentioned written character is extracted from the above-mentioned written information, and the above-mentioned written character stroke and The corresponding naming of the relative strokes of the standard characters selected in the above-mentioned font recognition step, and the stroke number corresponding to the naming recognition step of calculating the similarity of the above-mentioned written characters of a plurality of standard characters selected in the above-mentioned font recognition step; and

The second output step of outputting the standard characters selected in the above-mentioned font recognition step according to the above-mentioned similarity calculated in the above-mentioned stroke-number correspondence naming recognition step.

Among them, the first judging step corresponds to the judging whether or not to output by the stroke order priority output means 55 in the later embodiment (S4 in FIG. 2 ).

In the online character recognition method of the fourth invention, in the above-mentioned stroke-order-corresponding naming recognition step of the online character recognition method of the first or third invention, according to the above-mentioned second stroke feature, the stroke order of the above-mentioned written characters is performed in the above-mentioned The corresponding naming of the strokes of the written characters and the standard character strokes selected in the above-mentioned font recognition step, when the above-mentioned corresponding naming cannot be performed, the order of the corresponding naming strokes can be changed to carry out corresponding naming again.

In the online character recognition method of the fifth invention, in the online character recognition method of the fourth invention, the above-mentioned font recognition step is used to calculate the similarity of the above-mentioned selected standard characters relative to the above-mentioned written characters, and the above-mentioned stroke order is used to correspond to the naming recognition step According to the similarity of the above-mentioned standard characters calculated in the above-mentioned font recognition step, the corresponding naming of the above-mentioned stroke sequence is changed multiple times.

The online character recognition method of the sixth invention is in the online character recognition method of the third invention, and after the above-mentioned font recognition step, the similarity of the standard characters calculated in the above-mentioned font recognition step to the above-mentioned written characters Compared with a predetermined value, it is judged whether to output the second judging step of the above-mentioned standard characters; The third output step of the standard text selected in

When it is judged not to output in the above-mentioned second judging step, the above-mentioned stroke-order-corresponding naming recognition step is carried out.

Here, the second judging step corresponds to judging the presence or absence of output by the font-priority output means 110 in the later embodiment (S3 in FIG. 18).

In addition, the online character recognition method of the seventh invention is in the online character recognition method of the second invention,

The above-mentioned stroke number corresponding naming recognition step is made of DP matching step and similarity calculation step,

The DP matching step extracts the second stroke features representing the features of the strokes constituting the written characters from the written information, and uses DP matching to compare the strokes of the written characters with the standard character strokes selected in the font recognition step based on the second stroke features. The corresponding name of

The above-mentioned similarity calculation step extracts the stable strokes corresponding to the strokes of the above-mentioned written characters and the strokes of the above-mentioned standard characters from the strokes correspondingly named in the above-mentioned DP matching step, and calculates the above-mentioned stable strokes relative to the strokes selected in the above-mentioned font recognition step. the similarity of the above-mentioned written characters of the standard characters,

The second output step outputs the standard character selected in the font recognition step based on the similarity calculated in the similarity calculation step.

Here, the similarity calculation step corresponds to the detection by the stable stroke detection means 132 in the later embodiment (S51 in FIG. 22).

The online character recognition device of the eighth invention includes:

A written information input unit for inputting the written information accompanying the writing action;

Extracting the feature representing the overall shape of the above-mentioned written character from the above-mentioned writing information, and selecting a plurality of font recognition units similar to the standard characters of the above-mentioned written character from the font library according to the above-mentioned feature representing the overall shape;

Extracting a first stroke feature representing the features of the strokes constituting the written character from the written information, and performing an identification of the strokes of the written character and the strokes of the standard character selected by the font recognition unit in consideration of the stroke order of the written character based on the first stroke feature. Corresponding naming, calculating the corresponding naming recognition unit of the stroke order corresponding to the similarity of a plurality of standard characters selected by the above-mentioned font recognition unit with respect to the above-mentioned written characters; and

A first output unit for outputting the standard characters selected by the font recognition unit based on the stroke order corresponding to the similarity calculated by the name recognition unit.

Here, the written information input unit corresponds to the input unit 50 of the later embodiment. And the first output unit is equivalent to the display unit 58 in the later embodiment.

The online character recognition device of the ninth invention includes:

A written information input unit for inputting written information accompanying the writing action;

Extracting the feature representing the overall shape of the above-mentioned written character from the above-mentioned writing information, and selecting a plurality of font recognition units similar to the standard characters of the above-mentioned written character from the font library according to the feature representing the above-mentioned overall shape;

Extracting the feature representing the second stroke that constitutes the stroke feature of the above-mentioned written character from the above-mentioned writing information, and performing the strokes of the above-mentioned written character and the strokes of the standard character selected by the above-mentioned font recognition unit in consideration of the number of strokes of the above-mentioned written character according to the above-mentioned second stroke feature The corresponding naming, calculating the number of strokes corresponding to the naming recognition unit of a plurality of standard characters selected in the above-mentioned font recognition unit relative to the similarity of the above-mentioned written characters; and

A second output unit for outputting the standard characters selected in the font recognition unit according to the similarity calculated in the stroke number corresponding to the naming recognition unit.

Wherein, the second output unit is equivalent to the display unit 58 in the following embodiments.

The online character recognition device of the tenth invention includes:

A written information input unit for inputting written information accompanying the writing action;

Extracting features representing the overall shape of the above-mentioned written characters from the above-mentioned writing information, and selecting a plurality of font recognition units similar to the above-mentioned written characters from the font library according to the above-mentioned features representing the overall shape;

Extract the first stroke feature representing the features of the strokes constituting the written character from the written information, consider the stroke order of the written character according to the first stroke feature, and perform the comparison between the strokes of the written character and the standard character selected in the font recognition unit The corresponding naming of strokes, calculating the stroke order corresponding naming recognition unit of a plurality of standard characters selected in the above-mentioned font recognition unit relative to the similarity of the above-mentioned written characters;

A first discrimination unit that judges whether to output the above-mentioned standard text by comparing the similarity calculated in the above-mentioned stroke-order corresponding naming recognition unit with a predetermined value;

When it is judged to be an output in the above-mentioned first judging unit, the first output unit that outputs the standard characters selected in the above-mentioned font recognition unit according to the above-mentioned similarity calculated in the above-mentioned stroke order corresponding to the naming recognition unit; and

When it is judged as not to be output by the first judging unit, a second stroke feature representing a stroke feature constituting the written character is extracted from the written information, and the strokes of the written character are compared with each other in consideration of the number of strokes of the written character based on the second stroke feature. The corresponding naming of the standard character strokes selected in the above-mentioned font recognition unit, calculate the number of strokes corresponding to the similarity of the plurality of standard characters selected in the above-mentioned font recognition unit with respect to the above-mentioned written characters, and the corresponding naming recognition unit; and

A second output unit that outputs the standard characters selected by the font recognition unit according to the similarity calculated by the stroke number corresponding to the naming recognition unit.

Wherein, the first judging unit is equivalent to the stroke order priority output unit 55 in the following embodiments.

In addition, the online character recognition of the eleventh invention is in the online character recognition device of the ninth invention,

The above-mentioned stroke number corresponding naming unit is composed of a DP matching unit and a similarity calculation unit;

The above-mentioned DP matching unit extracts from the above-mentioned writing information the second stroke features representing the stroke features constituting the above-mentioned written characters, and performs DP matching on the above-mentioned written character strokes and the standard character strokes selected by the above-mentioned font recognition unit according to the above-mentioned second stroke features. Corresponding naming;

The above-mentioned similarity calculation unit extracts the stable strokes corresponding one-to-one between the strokes of the above-mentioned written characters and the strokes of the above-mentioned standard characters through the named strokes from the above-mentioned DP matching unit, and calculates the standard strokes selected in the above-mentioned font recognition unit with respect to the above-mentioned stable strokes. The degree of similarity of the characters to the above-mentioned written characters;

The second output unit outputs the standard character selected by the font recognition unit based on the similarity calculated by the similarity calculation unit.

Wherein, the similarity calculation unit is equivalent to the stable stroke detection unit 132 in the following embodiments.

The online character recognition device of the 12th invention is in the online character recognition device of the 10th invention, including:

A second judging unit that judges whether to output the above-mentioned standard characters by comparing the similarity of the standard characters calculated in the above-mentioned font recognition unit of the above-mentioned written characters with a predetermined value;

A third output unit that outputs the standard characters selected in the above-mentioned font recognition unit according to the above-mentioned similarity calculated in the above-mentioned font recognition unit when it is judged as output in the above-mentioned second judging unit;

The above-mentioned stroke order corresponds to the naming recognition unit, and is processed when it is judged not to be output by the second judging unit.

Embodiments of the invention will be described below with reference to the drawings. In the attached picture:

FIG. 1 is a block diagram showing an online character recognition device according to Embodiment 1;

FIG. 2 is a flow chart showing the operation of the control unit 59 in Embodiment 1;

FIG. 3 is a diagram showing an input font written in the input unit 50 of Embodiment 1;

Fig. 4 represents the flow chart of the font recognition unit 52 action of embodiment 1;

FIG. 5 is a diagram showing the types of allocation direction codes of the font recognition unit 52 in Embodiment 1;

Fig. 6 is the input glyph diagram after carrying out direction code allocation processing of embodiment 1;

Fig. 7 is the division of direction code area of embodiment 1, the figure of the code count of each direction in each area;

Fig. 8 is a flow chart showing the actions of the stroke-order corresponding naming recognition unit 54 in Embodiment 1;

FIG. 9 is a diagram showing directions when the stroke-order-corresponding naming recognition unit 54 assigns stroke directions and virtual stroke directions in Embodiment 1;

Fig. 10 is a figure representing the stroke features of the first stroke of the input font with the stroke-order corresponding naming recognition unit 54 of embodiment 1;

FIG. 11 is a diagram showing the rule of stroke-corresponding naming by the stroke-order corresponding naming recognition unit 54 in Embodiment 1;

12 is a diagram showing the stroke arrangement of the standard glyph of the character " " in the stroke feature font library of Embodiment 1;

Fig. 13 is a diagram showing the input fonts of the character " " in Embodiment 1 with different stroke orders;

FIG. 14 is a diagram showing the process of changing the stroke order of the input font of the stroke-order corresponding naming recognition unit 54 in Embodiment 1;

FIG. 15 is a diagram showing input fonts of ligatures in order to illustrate the operation of the stroke number-corresponding naming recognition unit 57 in Embodiment 1;

FIG. 16 is a diagram showing the corresponding naming results of standard fonts and input fonts of the stroke number corresponding naming recognition unit 57 in Embodiment 1;

17 is a structural diagram of an online character recognition device in Embodiment 2;

FIG. 18 is a flow chart showing the flow of operations of the control unit 59 of the second embodiment;

FIG. 19 shows a configuration diagram of an online character recognition device according to Embodiment 3;

FIG. 20 is a flowchart illustrating the flow of operations of the control unit 59 according to Embodiment 3;

FIG. 21 is a structural diagram of an online character recognition device according to Embodiment 4;

FIG. 22 is a flow chart illustrating the operation flow of the control unit 59 in Embodiment 4;

FIG. 23 is a diagram illustrating the path where the distance between the input glyph of the DP matching unit 130 of Embodiment 4 and the DP matching result of the standard glyph becomes the minimum;

FIG. 24 is a diagram illustrating the path when the actual writing of the input glyph and the standard glyph in the DP matching unit 130 of Embodiment 4 cannot be named correspondingly;

FIG. 25 is a diagram showing the corresponding names of the stable strokes detected by the stable stroke detection unit 132 of Embodiment 4;

Fig. 26 is a structural diagram showing the online character recognition device of prior art example 1;

Fig. 27 is a figure representing the input font, standard font and its corresponding naming result of prior art example 2;

Fig. 28 is a structural diagram showing an online character recognition device of prior art example 2;

Embodiment 1

The online character recognition device of this embodiment extracts spare characters through font recognition, performs stroke-order-corresponding naming recognition on the spare characters, and then performs stroke-number-corresponding naming recognition, which will be described below with reference to FIGS. 1-16 .

FIG. 1 is a configuration diagram of an online character recognition device according to this embodiment.

In FIG. 1 , 50 is an input unit for inputting written information such as signs, and 51 is a font library that stores in advance the general shape characteristics of characters independent of the number of strokes and stroke order of each character.

52 is a font recognition unit that recognizes by comparing with the aforementioned font library 51 that extracts overall shape features independent of stroke number and stroke order from the written information obtained from the aforementioned input unit 50 . The overall shape feature of so-called characters independent of stroke number and stroke order will be described later.

53 is a stroke feature character library that stores in advance the number of strokes of each character, the stroke order, and the stroke features of each stroke that constitutes the character.

In this specification, the so-called strokes refer to various character strokes that constitute a certain character.

54 is the stroke feature extracted from the writing information obtained by the input unit 50, and the stroke sequence corresponding name recognition unit is used to identify the corresponding name of each character stored in the stroke feature font library 53 by considering the stroke order.

55 is a stroke order priority output unit for determining whether to output the recognition result as the final recognition result according to the recognition result information obtained by the recognition unit 54 corresponding to the stroke order.

57 is a stroke-number-corresponding-naming recognition unit that recognizes the stroke features of the input font and the strokes of the standard font in the aforementioned stroke feature font library 53 in consideration of the number of strokes.

58 is a display unit for displaying the recognition result of the stroke-number corresponding naming recognition unit 54 or the recognition result of the stroke-number corresponding naming recognition unit 57 .

59 is a control unit for controlling the processing operation of each unit 50-57 in the character recognition device, which is connected to each unit 50-57, respectively.

Hereinafter, the operation of the online character recognition device according to this embodiment will be described with reference to FIG. 2 .

FIG. 2 is a flowchart showing the processing procedure of the control unit 59 of the online character recognition device according to this embodiment.

As a description of this embodiment, for input text The character recognition program at that time will be described.

First, at step S1, the control unit 59 instructs the input unit 50 to obtain an input font. The input glyphs contain input process information following actions written into the input unit 50 . Such information is written information.

的图，60～66是按时间先后输入的笔划顺序。60～66的顺序表示写入笔划的输入顺序也即笔序。本例中用七划书写，所以是由7条笔划构成。Fig. 3 is to illustrate input unit 50 writing input font

60～66 are the stroke order of inputting in succession according to time. The order of 60-66 indicates the input order of written strokes, that is, the stroke order. Write with seven strokes in this example, so be made of 7 strokes.

Therefore, in S1, seven strokes 60 to 66 are obtained as writing information.

Then enter step S2, the control unit 59 issues an instruction, and the stroke sequence of the input font is sent to the font recognition unit 52 for font recognition.

In this example, the strokes 60-66 of the input font in Figure 3 are sent to the font recognition unit 52, and the font recognition unit 52 extracts the overall shape features of the characters that have nothing to do with the number of strokes and stroke order from the strokes 60-66 of the input font for recognition .

The procedure of step S2 of the recognition processing by the font recognition unit 52 will be described in detail using FIG. 4 .

FIG. 4 is a flowchart showing the processing procedure of the font recognition unit 52 .

First, in step S11 of FIG. 4, the font recognition unit 52 assigns 4 kinds of direction codes to each point of multiple samples for each stroke of the input font.

4 kinds of direction codes are shown in Fig. 5 4 kinds of direction codes, specifically vertical (V) 70, horizontal (H) 71, oblique upper right (R) 72, oblique upper left (L) 73 direction codes.

In each direction, the range from the direction of the arrow to the dotted line is regarded as the same direction.

Multi-point sampling is performed on one stroke, and each point is assigned to any one of the aforementioned 4 direction codes.

However, the usual sampling points are not necessarily closely spaced due to writing speed deviation and sampling rate.

Therefore, if the direction codes are assigned to the original dots, the number of direction codes will vary depending on the writing speed even for characters with the same font. In addition, the number of dots varies depending on the size of the characters, so the size should be properly normalized and interpolation between dots should be performed to normalize the dots.

The general direction from the current point to the next is obtained for each normalized point, and the direction code is assigned to each point by judging whether the direction corresponds to any direction in FIG. 5 . Here, since the next point does not exist at the end point of the stroke, the same direction as that of the previous point is assigned. The results of normalizing the size of each stroke 60-66 of the input font, interpolation processing, and assigning direction codes to each point are shown in FIG. 6 .

Fig. 6(a) is the result of assigning the horizontal direction code (H), and Fig. 6(b) is the result of assigning the vertical direction code (H). The font of this example, because there is not oblique upper right (R), the direction of oblique upper left (L), omitted these directions.

Next, enter step S12 of Fig. 4, divide the whole word into 4*4 areas, and count each direction code in each area.

Fig. 7 is an illustration of dividing the entire character into 4×4 regions and calculating the number of direction codes in each region.

Fig. 7(c) is the result of calculating the code in the horizontal direction in Fig. 6(a), and Fig. 7(d) is the result of calculating the code in the vertical direction in Fig. 6(b).

Therefore, the directional code feature F of four directions is extracted from the input font and expressed as a 4×4 16-dimensional vector along each direction.

      F = (FH, FV, FR, FL)

       FH = FH1, FH2, ..., FH16

       FV = FV1, FV2, ..., FV16

FR＝FR1，FR2，…，FR16

      FL = FL1, FL2, ..., FL16

FH1, FH2, . . . , FH16 indicate the number of calculations of direction codes in the 4×4 area. It has the same meaning for FV, FR, FL.

Then enter into S13 of FIG. 4 , and obtain the distance D between the direction code distribution feature of each character stored in the font library in advance and the feature of the input font.

The distribution characteristic of the direction code of standard font can be expressed as follows:

S = (SH, SV, SR, SL)

     SH=SH1, SH2,...,SH16

SV＝SV1, SV2,..., SV16

SR＝SR1, SR2,..., SR16

SL=SL1, SL2,..., SL16

Here, the distance between the input font and the standard font can be expressed as follows.

D = (DH, DV, DR, DL)

DH＝|FH1-SH1|+|FH2-SH2|+……+|FH16-SH16|

        DV = |FV1-SV1|+|FV2-SV2|+...+|FV16-SV16|

DR＝|FR1-SR1|+|FR2-SR2|+...+|FR16-SR16|

DL＝|FL1-SL1|+|FL2-SL2|+...+|FL16-SL16|

From the above formula, the distance between the standard glyph in the font library and the input glyph is obtained, and 50 candidate characters as recognition results are output in sequence from the smaller distance.

As described above, the processing of steps S11 to S13 in FIG. 4 is a recognition processing that does not depend on the character's overall shape characteristics such as the number of strokes and the order of strokes.

After selecting 50 candidate characters, move to step S3 of FIG. 2 .

Then enter S2 of FIG. 2 , the control unit 59 issues an instruction to send the 50 candidate characters obtained by the font recognition unit 52 to the stroke-order corresponding name recognition unit 54, so that the stroke-order corresponding name recognition is performed.

The operation of the stroke-order correspondence naming recognition unit 54 will be described in detail with reference to FIG. 8 .

FIG. 8 is a flow chart illustrating the operation of the stroke-order corresponding naming recognition unit 54 .

First, the stroke-order corresponding naming recognition unit 54 enters step S20 of FIG. 8 , and extracts the stroke features of each stroke from the input font.

Each stroke shape feature of each stroke is extracted as a stroke feature, the direction from the start point to the end point of the stroke, the width of the stroke circumscribing rectangle, the height of the stroke circumscribing rectangle, the direction of the imaginary stroke, the width of the imaginary stroke circumscribing rectangle, and the height of the imaginary stroke circumscribing rectangle. .

The circumscribed rectangle of the stroke is the rectangle circumscribed relative to the stroke.

An imaginary stroke is exactly a straight line connecting the end point of the current stroke and the starting point of the next stroke. When the final stroke, the next stroke is the original stroke. The circumscribing rectangle of the imaginary stroke is a rectangle having the imaginary stroke as a diagonal.

Also, the direction of each stroke is quantized using the eight directions shown in FIG. 9 . In addition, the width and height are normalized values in which the width and height of the circumscribing rectangle of the entire character are set to 100, respectively. FIG. 10 shows an example of extracting seven stroke features from the first stroke 66 of an input font.

Then, compare the strokes of the input font with the standard font strokes in the stroke feature font library of 50 candidate characters obtained by the font recognition unit 52, because the corresponding naming of the stroke order is only carried out in the stroke order corresponding naming recognition unit, so if the input If the number of strokes of the font is inconsistent with the number of strokes of the standard font, it cannot be named accordingly.

Here, since the stroke number information is not used in the font recognition unit 52, characters of various stroke numbers are included in the character candidates. Therefore, at first S21 compares the number of strokes between the input font and the standard font. When the number of strokes is inconsistent, S21 is "No", then the naming corresponding to the standard font of the current candidate character is not carried out, and the corresponding naming with the next character is carried out. When S21 matches, it becomes "YES" and proceeds to S22.

In S22, the stroke features of the standard letters in the stroke feature font library 53 and the stroke features of the input font are used for corresponding naming. Here, the standard font is the correct stroke order as shown in FIG. 12 or the stroke characteristics of the stroke order written by most people. In addition, to determine whether the strokes correspond to each other in naming, a corresponding naming test is performed with reference to the corresponding naming rule described later.

Fig. 11 illustrates an example of corresponding naming rules. In the example, the characteristics of various stroke shapes are described, and the input font and the standard font are verified based on the characteristics. For example, in the test of imitating the corresponding names of each other, only the stroke width of the important feature of imitating strokes is tested.

Here, in the direction test, when the difference in direction is +-2 or more, it is determined that the name does not correspond.

The width and height of the circumscribed rectangle of the whole text are 100, and when the difference between the width and height is more than 70, the verification of the width and height will be judged as not corresponding to the naming. However, the standard is different, and the direction is unstable when short strokes are named in correspondence with each other compared to when long strokes are named in correspondence with each other. Therefore, the allowable range of the correspondence name should be expanded.

In addition, since the direction, width, and height of virtual strokes represent the positional relationship between strokes as a scale, the stroke shape is not tested. Here, the above-mentioned corresponding naming is performed when the corresponding naming is successful, and when the corresponding naming fails or when all strokes have been correspondingly named, the process proceeds to S23.

Then in S23, within the strokes of the input font, check whether the corresponding naming is successful with respect to all strokes, enter S26 when all corresponding, and then enter S24 when the corresponding naming fails in the middle. In this example, because the input font is written in the correct stroke order, all strokes are named correspondingly, and just enter S26.

Then at S26, calculate the distance between the stroke feature of the input font and the stroke feature of the standard font with respect to all strokes, after each feature is normalized, after weighting the distance of each feature, finally find the distance with the existing standard font. The weighting can be pre-set to improve the recognition rate, or the weight can be changed by the number of strokes, etc.

However, the stroke shape is not used in distance calculations.

Here, the normalized distance Dd from the stroke direction, the normalized distance Dw from the stroke width, the normalized distance Dn from the stroke height, the normalized distance Dvw from the virtual stroke width, and the normalized distance Dvw from the normalized virtual stroke height The distance D to the calibration glyph is obtained from the back distance Dvh and the normalized distance Dvd of the false stroke direction.

D=(Wd×Dd+Ww×Dw+Wn×Dh+Wvw×Dvw+Wvh×Dvh+Wvd×Dvd) Here, Wd represents stroke direction weight, Ww represents stroke width weight, Wh is stroke height weight, Wvw is imaginary Stroke width weight, Wvh is the weight of imaginary stroke height, Wvd is the weight of imaginary stroke direction. These weights have the following relationship.

   Wd+Ww+Wh+Wvw+Wvh+Wvd＝1

This weight is preset to improve the recognition rate. In addition, such weighting can also be changed according to the number of character strokes to be recognized.

Hereinafter, the operation in the case of failing to correspond to the naming in S23 will be described. Set the text as shown in Figure 13 Write in a different stroke order than the standard glyph. At this time, the second stroke 91 of the input font and the second stroke 81 of the standard font are both vertical strokes, and it is determined that there is no corresponding naming in the verification of the height of the stroke. Therefore, at the stage where the corresponding naming of the second stroke is completed, the process proceeds to S24.

Then at S24, it is judged that the stroke order of the input font is different, the stroke of the input font that has failed in corresponding naming, by the stroke of the input font that is not used in the corresponding naming successively (if the Kth stroke with the existing standard font is to be carried out The corresponding naming of the stroke, then in the strokes other than the input strokes used in the corresponding naming of the K-1 stroke to the standard font, the corresponding naming of the K stroke is carried out) exchange, and the stroke order arrangement is replaced.

But only when the final stroke does not correspond to the naming, or when there is no unused stroke in the stroke of the standard font that is carrying out the corresponding naming now, it is possible to judge whether the corresponding naming so far has any errors. The unused input strokes can be named correspondingly only if the naming is not corresponding to the naming.

In this example, as shown in FIG. 14( a ), the stroke 91 that does not correspond to the name is replaced with the next stroke 92 .

Then enter into S25, compare the number of times of recalculation (stroke order exchange times) of the input stroke so far with the upper limit of the number of times of recalculation of the predetermined number of times of recalculation, become "No" when not exceeding, get back to S22, and again never correspond to the name The strokes of standard glyphs start to be named accordingly.

The input stroke 92 arranged in S22 is not named correspondingly with the stroke 81 of the second stroke of the standard font. Same as the previous time, the corresponding naming of the second stroke of the standard font fails, and enters S24.

In S24, the next unused input stroke 93 is just in the second stroke, and the input strokes 91, 92 after the failure of the corresponding naming of the second stroke of the standard font return to the stroke order of the original input font.

Fig. 14(b) shows the arrangement of input strokes after shifting.

Repeat the above process until finally all the strokes correspond to the named stroke sequence exchange process shown in Figure 14(c)-(f). In this example, after 6 recalculations, the corresponding naming is successful, and enters into S26 together with the occasion of writing in the correct stroke order to calculate the distance from the standard font. In addition, in accordance with the above-mentioned stroke order exchange, among the input strokes whose stroke order has been changed, the direction, width, and height of virtual strokes representing the positional relationship between strokes are recalculated.

After completing the above corresponding naming of all the recognition candidate characters obtained by the font recognition unit 52, go to S4 in FIG. 2 .

In S4 , the control unit 59 sends the distance of the character with the shortest distance among the characters successfully named by the stroke-order corresponding naming recognition unit 54 to the stroke-order priority output unit 55 .

In the stroke priority output unit 55, the distance corresponding to the successfully named character is compared with a predetermined threshold, and if the distance is determined to be smaller than the threshold, the result is output, and if the distance is determined to be greater than the threshold, the result is not output.

Control unit 59, when judging that the stroke-order priority output unit 55 is outputting, in the 50 characters selected in the character recognition unit 52, start from the little character of the distance calculated by the stroke-order corresponding naming recognition unit 54 in order A certain number is sent to the display unit 58, and the text is displayed on the display unit 58.

In the case of the input font of Fig. 3 and the input font of Fig. 13, the corresponding naming of the common stroke order is successful, and the distance obtained is much smaller than the threshold value, and it is determined that the stroke order priority output unit 55 outputs, and the corresponding naming of the stroke order is identified The recognition result of unit 54 is output as a final result.

If it is judged that the stroke order priority output unit 55 does not output, then go to S5 in FIG. 2 .

用六划书写成字形的场合进行讨论。Next, in order to explain the operation of S5 in Figure 2, the characters in Figure 15 will be

Discuss the occasions where the characters are written with six strokes.

字的标准字形，不能用笔序对应命名识别单元54作与图15的输入字形的对应命名。With the input font of Figure 15 written in six strokes, because of only having 7 strokes in the stroke feature font library 53

The standard font of word can't be done with the corresponding naming of the input font of Fig. 15 with stroke order correspondence naming recognition unit 54.

Therefore, even assuming that the corresponding naming with other six-stroke characters is successful, because it becomes a different character corresponding naming, and the distance from the input font is very large, the stroke order priority output unit 55 judges because it is not a value smaller than the threshold value. S4 in Fig. 2 has no output, and enters S5.

In S5, the stroke number-corresponding naming recognition unit 57 uses the same DP matching as in the prior art example 2 for the 50 characters selected by the character recognition unit 52, and performs corresponding naming of those with different stroke numbers to minimize the distance.

The strokes of the input glyph are expressed as follows,

A=a1, a2, ..., ai, ..., the stroke of the standard font of aI (I is the stroke number of input font) font storehouse represents as follows,

B=b1, b2, ..., bj, ..., bJ (J is the stroke number of a standard font) then ai, the distance d (i, j) of bj can be calculated as follows.

d(i, j)＝|wai-wbj|+|Hai-Hbj|

Here, W is the width of the stroke, and H is the height of the stroke.

Set d(i,j)=|ai-bj|

(1) When i=1, j=1

g(1,1)=d(1,1)

(2) When i=1, j≠1

g(1,j)=g(1,j-1)+d(1,j)

(3) When i≠1, j=1

g(i,1)=g(i-1,1)+d(i,1)

由上面表示的递推式求出g(i，j)(初始条件)，用下式求出A的笔划列与B的笔划列的距离S(A，B)。(4) Other situations

g(i, j) (initial condition) is obtained from the recursive expression shown above, and the distance S(A, B) between the stroke sequence of A and the stroke sequence of B is obtained by the following equation.

S (A, B)=I/(I+J)*g (I, J) carries out above-mentioned DP matching with regard to the candidate text of 50 recognition results that font recognition unit 52 gains, and distance S (A, B) from minimum is sequentially output to the display unit 58 as recognition results.

The display unit 58 will display the characters sequentially starting from the character with the smallest distance calculated by the DP matching.

In this example, through DP matching, the stroke arrangement of the input glyph in FIG. 15 and the standard glyph in FIG. 12 are named correspondingly as in FIG. 16 , which can be correctly identified.

In addition, in this embodiment, although the feature of the glyph recognition unit 52 is the use of the direction code feature, it may also be the surrounding distribution feature that takes the area from the circumscribed rectangle of the character to the area that touches the stroke of the character as a feature, or the line of the stroke. Density, etc. reflect the overall glyph characteristics of the text.

Also, in the present embodiment, the substituting code feature is divided into 4×4 regions, but the number of divisions of the regions may also be changed according to the capacity of the font library and the recognition ability.

In addition, in this embodiment, 50 characters are fixed as the number of candidate characters output from the font recognition unit 52 , but this value may be changed according to the recognition ability of the font recognition unit 52 .

In addition, in this embodiment, the same number of candidate characters are output from the font recognition unit 52 to the stroke order change corresponding naming recognition unit 54 and the stroke number corresponding naming recognition unit 57, but the output candidates can also be changed according to the recognition ability of each recognition unit. number of characters.

The effects of the online character recognition device of this embodiment will be described below.

If adopt above-mentioned present embodiment, only with regard to the candidate character of recognition result by the general feature of character, carry out by stroke order correspondence out-of-name recognition, by the recognition of stroke number corresponding naming, do not carry out corresponding naming with the feature of the dissimilar character of font, Therefore, high-speed and high-precision identification can be realized.

Specifically, when performing handwritten character recognition, in the case of a part of the characters of the second level of JIS, it is necessary to perform high-speed recognition of 4,000 characters. The number of strokes corresponds to the naming, and it is actually difficult in terms of the processing time it takes. Here, in this embodiment, the character recognition unit 52 is used for the first step of character recognition, and the general characteristics of the characters are used to quickly compress the characters that are recognized in the second stage of character recognition or stroke number recognition. .

Furthermore, since characters with similar font styles are output as candidate characters, the user's impression of misrecognition is enhanced, and the ratio of correct characters included in the recognition candidate characters is also increased.

In addition, since the corresponding naming of strokes is strictly carried out through multiple features in the stroke-order corresponding naming recognition unit 54, it is possible to recognize with high precision the written characters that follow the number of strokes, that is, the number of strokes that do not exist in standard fonts such as ligatures. The changed input font can be correctly sent into the corresponding naming recognition unit 57 of the number of strokes, so that the ligatures can also be recognized with high precision.

As the character display in the present embodiment, there are cases where the recognition result of the output unit 55 is displayed with stroke order priority and the recognition result displayed by the stroke number corresponding to the name recognition unit 57 .

Thus, the output to the display unit 58 is formed in two stages. Therefore, when it is determined to be output by the stroke order priority output unit 55, it can be displayed at the previous stage, so the recognition time can be advanced.

In addition, in the present embodiment, although the first item is to carry out font recognition, the second item is to carry out stroke order corresponding naming recognition, and the third item is to carry out stroke number corresponding naming recognition, but in the occasion that pays more attention to stroke order than ligature , where the sequence of strokes places more emphasis on continuous strokes, etc., the recognition sequence after the second item can also be changed according to the situation.

Therefore, the following forms can be considered as the character recognition procedure.

A: Font recognition → stroke number corresponds to naming recognition → stroke order corresponds to naming recognition

B: font recognition → stroke order corresponding naming recognition

C: font recognition → stroke number corresponds to naming recognition. Embodiment 2

The online character recognition device of the present embodiment judges whether to output the candidate characters extracted by the shape recognition through font recognition, and performs name recognition corresponding to stroke order and naming recognition corresponding to stroke number when it is judged not to output. The following description will be made with reference to FIG. 17 and FIG. 18 .

In FIG. 17 , 50 to 50 denote the same devices as those in Embodiment 1, and description thereof is omitted.

110 is a font priority output unit for judging whether to output the recognition candidate characters of the font recognition unit 52 based on the information of the recognition candidate characters obtained by the font recognition unit 52 .

Hereinafter, the operation of the online character recognition device according to the embodiment will be described with reference to FIG. 18 .

FIG. 18 is a flowchart showing the processing procedure of the control unit 59 of the online character recognition device according to this embodiment.

The description of the processing up to S2 is the same as that in the first embodiment, and the description thereof is omitted.

After the processing of S1 and S2 is completed, the control unit 59 sends the character with the smallest distance among the 50 recognition candidate characters obtained by the font recognition unit 52 to the font priority output unit 110 .

Next, in S30 of FIG. 18 , the font priority output unit 110 determines whether to output the recognition result of the font recognition unit 52 . Judgment, for the character of the minimum distance sent by the font recognition unit 52, obtain the similarity (for example, normalized distance) between the output font and the standard font, when the similarity is larger than the threshold value prepared for each character in advance, it is judged as Output the result, if it is small, it is judged not to output the result.

Assuming that the input font vector is f, and the standard font vector is g, then the similarity is calculated according to the following formula.

S(f, g)=(f, g)/‖f‖·‖g‖

Here, the threshold value of character A obtains the degree of similarity according to the standard font in the character font library 51 of character A and the direction code feature of the written font other than character A in the written font including all the character fonts of the recognition object, and the maximum similarity (similarity of glyphs other than the standard glyph closest to character A) is set. In this way, the threshold value of the character A in which no character with a similar font exists becomes a small value, and the rate at which the result of the font recognition section 52 is output as it is increases.

In addition, since the threshold value of characters with similar fonts becomes larger, the output part becomes smaller, and can be identified by detailed stroke features.

If it is determined in S30 that the font priority output unit 110 outputs the result, the 50 recognition candidate characters obtained by the font recognition unit 52 are displayed in order from the one whose distance calculated in S2 is smaller.

If it is determined in S30 that the font priority output unit 110 does not output the result, the processing from S3 onwards is performed similarly to the first embodiment for the 50 recognition candidate characters obtained by the font recognition unit 52 .

In addition, in the present embodiment, the output determination of the font priority output unit 110 uses the normalized distance as the similarity, but the Euclidean distance, the Mahalinobis distance, etc. may also be used.

The effect of the online character recognition device of this embodiment will be described below.

In the online character recognition device of the present embodiment, when there are no characters with similar fonts, the ratio of the recognition results of the output font recognition unit 52 increases, and it is possible to obtain a character that is not affected by stroke order changes, stroke number changes, and local deformation of characters. high-precision recognition. In addition, when outputting the recognition result of the fixed form recognition unit 52, it is possible to obtain very high-speed recognition.

And if this embodiment is adopted, for very carefully written characters, the high recognition accuracy is obtained by the font recognition unit, and the stroke order corresponding naming recognition is not performed and displayed at a high speed. Accurately recognize carefully written text and improve the impression on users.

In addition, in the online character recognition device of this embodiment, when the reliability of the result of the first-stage character recognition is high, since the recognition result of this stage is displayed, it is possible to recognize relatively carefully written characters at high speed.

Embodiment mode 3

In this embodiment, the online character recognition device sets multiple times the number of times of recalculation of the stroke sequence corresponding to the name recognition for the characters with a high possibility of correctly analyzing the characters among the candidate characters extracted through font recognition, and the following is performed with reference to FIGS. illustrate.

In FIG. 19 , 50 to 59 , and 110 are the same devices as those in the previous embodiment, and their descriptions are omitted. 120 is a stroke sequence corresponding naming identification unit for weighting the corresponding naming of the strokes of the input font and the standard font.

The operation of the online character recognition device of this embodiment will be described with reference to FIG. 20 .

FIG. 20 is a flowchart of the processing procedure of the control unit 59 of the online character recognition device according to this embodiment.

In FIG. 20 , the processing up to S30 is the same as that in Embodiment 2, and description thereof will be omitted.

Carry out the processing of S1, S2, and S3, and proceed to S40. In S40, the control unit 59 sends the 50 candidate characters obtained by the font recognition unit 52 into the weighted stroke sequence corresponding naming recognition unit 120, and the weighted stroke sequence corresponding naming recognition unit 120 performs input font and standard font The corresponding named recognition of the strokes.

The detailed stroke-corresponding naming operation is the same as the operation in Embodiment 1 of FIG. 8 , but the number of permutations (recalculation times) of the stroke order of S25 is exactly the same in Embodiment 1 with respect to the candidate characters obtained by the font recognition unit 52 , on the contrary, in this embodiment, it is changed according to the order of the candidate characters of the font recognition unit 52 (the order of the characters with the smallest distance).

Specifically, among the candidate characters of the recognition result of the font recognition unit 52, the characters with the highest rank are the characters with the highest probability of correctly answering the character, the larger the threshold value of the recalculation times, and the lower the rank That is, the lower the possibility of correctly answering a character, the smaller the threshold value of the number of times of recalculation.

If this embodiment is adopted, for the characters with low probability of correct answer, the number of strokes arranged and replaced is reduced, and the above-mentioned corresponding naming is not necessary, so it can be recognized at a high speed, and the need for characters with low possibility of correct answer can be reduced Misrecognition caused by the above-mentioned stroke order exchange.

Embodiment 4

The online character recognition device of this embodiment is to extract the stable point corresponding to the input font and the standard font one-to-one when the stroke number is named and recognized, and carry out the corresponding naming of each stroke on the stable point. Below, refer to FIGS. 21 to 25. illustrate.

In FIG. 21 , 50 to 59 are the same devices as Embodiment 1, 110 and Embodiment 2, and 120 and Embodiment 3, and the description thereof is omitted.

130 is a DP matching unit, 131 is a verification unit corresponding to a stroke, and 132 is a naming unit corresponding to a stable stroke.

The operation of the online character recognition device according to this embodiment will be described with reference to FIG. 22 .

FIG. 22 is a flowchart showing the processing procedure of the control unit 59 of the online character recognition device according to this embodiment.

In FIG. 22 , the operations up to S4 are the same as those in Embodiments 1 to 3, and the description thereof will be omitted. Only the operations after S50 will be described.

First, when it is determined in S4 that the stroke order priority output unit 55 has no output, the control unit 59 sends all 50 candidate characters of the recognition result obtained by the font recognition unit 52 to the DP matching unit 130 .

Then, at S50 , the control unit 59 instructs the DP matching unit 130 to name the input font and the standard font for the 50 candidate characters of the recognition result obtained by the font recognition unit 52 .

Here, the stroke corresponding naming verification unit 131 verifies the corresponding command path of the DP matching unit 130, and compensates for the non-corresponding naming, so as to prevent the path from being selected.

Regarding the corresponding naming path of the DP matching unit 130, the corresponding naming of the input glyph in FIG. 15 and the standard glyph in FIG. 12 is taken as an example for illustration. The input glyphs in FIG. 15 and the standard glyphs in FIG. 12 are named as in FIG. 16 , and the corresponding naming paths are presented in FIG. 23 in two dimensions.

In Fig. 23, the horizontal axis is the stroke arrangement of the standard font, and the vertical axis is the stroke arrangement of the input font. The thick line in the figure is the corresponding named path of DP matching.

Of course, there is no problem in the case of correct corresponding naming in FIG. 24 , but there is still a place where corresponding naming like that in FIG. 24 is required.

FIG. 24 shows the path of uncorresponding naming places in the actual writing of the input glyph and the standard glyph by the DP matching unit 130 .

The corresponding naming situation of Fig. 24 is that the stroke of the 3rd stroke of the input font corresponds to the 3rd, the 4th, the 5th stroke of the standard font. Such a corresponding state is called a combination of strokes, which means that all the strokes originally recorded in two strokes are recorded into a state of one-stroke continuous strokes.

Moreover, the fifth stroke of the standard font is named correspondingly to the third stroke and the fourth stroke of the input font. Such a state is called stroke separation, which means that the entire stroke recorded by one stroke is now recorded by two divisions.

That is, the combination and separation of the fifth stroke of the standard font occur simultaneously. Because this phenomenon does not occur in general character writing, the corresponding naming test unit 131 is used to prevent non-corresponding naming.

That is to say, when the DP unit 130 performs DP matching, the corresponding naming test unit 131 of the strokes does not correspond to the names in normal writing such as horizontal to vertical paths and vertical to horizontal paths, and adds extreme values to the distance values of the selected paths. Large offsets prevent this path from being chosen.

Then enter S51, the control unit 59 sends the path information obtained on the DP matching unit 130 to the stable stroke checking unit 132. The so-called path information is information that expresses the correspondence of each stroke with lines as shown in FIGS. 23 to 25 .

The stable stroke checking unit 132 extracts each point on the path as a stable point when the forward and backward paths are oblique paths. However, when the path directly behind the starting point of the path is an oblique path, and when the oblique path is directly in front of the end point of the path, it is a stable point.

FIG. 25 shows the stable points extracted by the stable stroke detection unit 132. In FIG.

In FIG. 25 , the stable points extracted by the stable stroke extraction unit 132 are denoted by O in the path of the corresponding naming result of the input font in FIG. 15 and the standard font in FIG. 12 . In this example, 5 stable points are drawn.

Next, the stable stroke verification unit 132 performs a one-to-one corresponding naming of the input font and the standard font strokes for the extracted stable points, and verifies the stroke naming with the 7 special pictures used by the stroke sequence corresponding naming recognition unit 54 .

This corresponding naming, when judging that there is no corresponding naming among the strokes of the stable point, then it is judged that the corresponding naming in the DP matching fails, and enters the corresponding naming of the candidate characters of the next font recognition unit 52.

In the example of FIG. 25, the strokes other than the stroke 102 are combined within the strokes of the input font, and the test of corresponding naming with the strokes of the standard font is performed. Since the corresponding naming is performed correctly, it is determined that the corresponding naming is successful.

The above processing is performed on all 50 candidate characters of the character recognition unit 52, and a certain number of characters starting from the character with the smallest distance within the corresponding successfully named characters is output as the final recognition result.

Now, the effects of the online character recognition device of this embodiment will be described.

According to the above-mentioned present embodiment, regarding the corresponding naming at the time of DP matching, since the path is controlled in actual writing so that non-corresponding naming does not occur, high-precision recognition can be realized.

In addition, according to this embodiment, with respect to the strokes corresponding to the DP matching results one-to-one, the corresponding names of the strokes are verified by using multiple features, so wrong corresponding names can be prevented and high-precision recognition can be realized.

The present invention has the following effects due to the configuration as described above.

The first invention can reduce the number of standard characters for name recognition according to the stroke order and can perform high-speed and high-precision recognition because the standard characters selected in the glyph recognition step are identified according to the stroke order corresponding to the name.

The second invention can reduce the number of standard characters for stroke number corresponding naming recognition because of the standard characters selected in the font recognition step, and can perform high-speed and high-precision recognition.

The third invention is because there is a first judging step for judging whether to output standard characters by comparing the similarity calculated in the stroke-corresponding naming recognition step with a predetermined value, and when judging as output in the first judging step, The first output step of standard characters selected in the font recognition step according to the above-mentioned similarity calculated in the above-mentioned stroke-corresponding naming recognition step, so when the recognition result of the stroke-corresponding naming recognition step is good, the recognition result can be output at a high speed .

The fourth point of the invention is to carry out the corresponding naming of the strokes of the written characters and the strokes of the above-mentioned standard characters due to the stroke order of the above-mentioned written characters. The character stroke order is different from the standard character stroke order and can be recognized with high precision.

In the fifth point of invention, the corresponding naming of the stroke order corresponding to the stroke order in the naming recognition step is changed by the number of times corresponding to the similarity calculated in the font recognition step, so characters with a high possibility of correct analysis can be sequenced more smoothly. Change, so that the characters with low probability of correct analysis can be recognized at high speed and high precision without changing the stroke order.

The sixth invention is because after the font recognition step, there is a second judging step of judging whether to output the standard characters by comparing the similarity of the standard characters selected in the font recognition step with respect to the written characters with a predetermined value. ,

And, when it is judged to be output in the above-mentioned second judging step, the third output step of outputting the above-mentioned standard characters according to the above-mentioned similarity calculated in the above-mentioned font recognition step, so when the recognition result of the font recognition step is good, the Identify high-speed outputs.

The seventh point of the invention is that from the strokes correspondingly named in the DP matching, the stable strokes corresponding to the strokes of the written characters and the strokes of the standard characters are extracted, and the similarity of the above-mentioned standard characters with respect to the above-mentioned written characters can be calculated with respect to the above-mentioned stable strokes The stable stroke detection step can prevent the deterioration of the recognition rate caused by non-corresponding naming in actual writing.

The eighth invention can reduce the number of standard characters for stroke-corresponding naming recognition on the standard characters selected in the glyph recognition unit, enabling high-speed and high-precision recognition.

The ninth aspect of the invention can reduce the number of standard characters to be recognized for stroke-number-corresponding naming by performing stroke-number-corresponding naming recognition on the standard characters selected in the font recognition unit, thereby enabling high-speed and high-precision recognition.

The tenth invention has a first judging unit for judging whether to output a standard character by comparing the similarity calculated in the stroke-order corresponding naming recognition unit with a predetermined value, and when judging as output in the first judging unit, Corresponding to the first output unit of the standard character selected in the font recognition unit according to the above-mentioned degree of similarity calculated in the above-mentioned stroke-order corresponding naming recognition unit, so in the occasion that the recognition result is good in the stroke-corresponding naming recognition unit, the recognition result can be transferred at a high speed output.

In the eleventh invention, in the DP matching unit, the stable strokes corresponding to the strokes of the written characters and the strokes of the standard characters are extracted from the strokes of the corresponding names, and the similarity of the above-mentioned standard characters relative to the above-mentioned written characters is calculated based on the above-mentioned stable strokes. High-speed stable stroke detection unit, so it can prevent the deterioration of recognition rate caused by non-corresponding naming in actual writing.

The twelfth invention has a second judging unit for judging whether to output the above-mentioned standard characters by comparing the similarity between the standard characters selected in the character recognition unit and the above-mentioned written characters with a predetermined value, and

When it is judged as output in the above-mentioned second judging unit, the third output unit that outputs the above-mentioned standard characters according to the above-mentioned similarity calculated in the above-mentioned font recognition unit, therefore, in the occasion where the recognition result of the above-mentioned font recognition unit is good, can use The recognition result is output at high speed.

Claims (12)

1. An online character recognition method, having a writing information input step of inputting the writing information with the writing action of the writing, It is characterized in that it also has the following steps: Extracting the features representing the overall shape of the above-mentioned written characters from the above-mentioned written information, and selecting a plurality of standard characters similar to the above-mentioned written characters from the font library according to the features representing the above-mentioned overall shape. A first stroke feature describing the characteristics of the strokes constituting the written character is extracted from the written information, and based on the first stroke feature, the strokes of the written character and the character recognition step are selected in consideration of the order of the strokes of the written character. The corresponding naming of the strokes of the standard characters, calculate the corresponding naming recognition step of the stroke order corresponding to the similarity of a plurality of standard characters selected in the above-mentioned font recognition step with respect to the similarity of the above-mentioned written characters, and The first output step of outputting the standard characters selected in the above-mentioned font recognition step according to the above-mentioned similarity calculated in the above-mentioned stroke-corresponding name recognition step. 2. A kind of online character recognition method, has the writing information input step of inputting its writing information with the writing action of writing, It is characterized in that it also has the following steps: Extracting and describing the overall shape features of the above-mentioned written characters from the above-mentioned writing information, and selecting a plurality of standard characters similar to the above-mentioned written characters from the font library according to the features describing the above-mentioned overall shape. A second stroke feature describing the characteristics of the strokes constituting the written character is extracted from the written information, and based on the second stroke feature, the strokes of the written character are compared with the criteria selected in the font recognition step in consideration of the number of strokes of the written character. The corresponding naming of the strokes of the characters, calculating the number of strokes corresponding to the naming recognition steps of a plurality of standard characters selected in the above-mentioned font recognition step relative to the similarity of the above-mentioned written characters, and The second output step of outputting the standard characters selected in the above-mentioned font recognition step according to the above-mentioned similarity calculated in the above-mentioned stroke-number corresponding naming recognition step. 3. An online character recognition method, which has a writing information input step of inputting its writing information along with the writing action of the characters, It is characterized in that it also includes the following steps: Extracting and describing the overall features of the above-mentioned written characters from the above-mentioned written information, and selecting a plurality of font recognition steps similar to the standard characters of the above-mentioned written characters from the font library according to the characteristics of the above-mentioned overall shape, Extract the feature of the first stroke that constitutes the stroke feature of the above-mentioned written character from the above-mentioned written information, and according to the above-mentioned first stroke feature, consider the stroke order of the above-mentioned written character to perform the stroke of the above-mentioned written character and the standard character selected in the above-mentioned font recognition step The corresponding naming of the stroke, calculate the corresponding naming recognition step of the stroke sequence of a plurality of standard characters selected in the above-mentioned font recognition step relative to the similarity of the above-mentioned written characters, By comparing the degree of similarity calculated in the recognition step corresponding to the above-mentioned stroke order with a predetermined value, the first judgment step of judging whether to output the above-mentioned standard characters, When being judged as output in the above-mentioned first discriminating step, the first output step of outputting the standard characters selected in the above-mentioned font recognition step according to the degree of similarity calculated in the above-mentioned stroke-corresponding naming recognition step, When it is judged not to output in the first judging step, a second stroke feature representing a feature of the strokes constituting the written character is extracted from the written information, and the writing is performed in consideration of the number of strokes of the written character based on the second stroke feature. Character strokes and the corresponding naming of the strokes of the standard characters selected in the above-mentioned font recognition step, calculate the number of strokes corresponding to the similarity of a plurality of standard characters selected in the above-mentioned font recognition step with respect to the above-mentioned written characters, and the corresponding naming recognition step, and A second output step of outputting the standard characters selected in the above-mentioned font recognition step according to the above-mentioned similarity calculated in the above-mentioned stroke number correspondence recognition step. 4. The character recognition method according to claim 1 or 3, characterized in that, the above-mentioned stroke sequence corresponds to the named recognition step according to the above-mentioned second stroke feature, carries out the above-mentioned written character stroke and in the above-mentioned font recognition step according to the above-mentioned written character stroke order The corresponding naming of the strokes of the selected standard characters, if the above-mentioned corresponding naming cannot be carried out, the stroke order of the corresponding naming is changed and the corresponding naming is carried out again. 5. The online character recognition method according to claim 4, characterized in that, the above-mentioned font recognition step calculates the degree of similarity of the above-mentioned selected standard characters with respect to the above-mentioned written characters, The stroke-order-corresponding naming recognition step is to change the corresponding naming of the stroke order according to the number of times of the similarity of the standard characters calculated in the font recognition step. 6. The online character recognition method according to claim 3, characterized in that, after the above-mentioned font recognition step, the similarity of the standard characters calculated in the above-mentioned font recognition step with respect to the above-mentioned written characters is compared with a predetermined value , the second judging step of judging whether to output the above-mentioned standard characters, and when judging as output in the above-mentioned second judging step, outputting the standard characters selected in the above-mentioned font recognition step according to the above-mentioned similarity calculated in the above-mentioned font recognition step The third output step, When it is judged not to output in the second judging step, the above-mentioned stroke-order-corresponding naming recognition step is performed. 7. by the online word recognition method of claim 2 record, it is characterized in that, above-mentioned stroke number corresponds to naming recognition step to be made of DP matching step and similarity calculating step, The DP matching step extracts the second stroke features representing the features of the strokes constituting the written characters from the written information, and performs DP matching between the strokes of the written characters and the standard characters selected in the font recognition step based on the second stroke features. The corresponding naming of the strokes, The above-mentioned similarity calculation step is to extract the stable strokes corresponding to the strokes of the above-mentioned written characters and the strokes of the above-mentioned standard characters one by one from the strokes correspondingly named in the above-mentioned DP matching step, and to calculate the above-mentioned stable strokes for the selected strokes in the above-mentioned font recognition step. The similarity between the standard characters of the above-mentioned written characters, The second output step outputs the standard character selected in the font recognition step based on the degree of similarity calculated in the degree of similarity calculation step. 8. An online character recognition device, having a writing information input unit for inputting its writing information along with the writing action of the character, It is characterized in that it also includes: Extract the features representing the overall shape of the above-mentioned written characters from the above-mentioned writing information, select a plurality of font recognition units similar to the standard characters of the above-mentioned written characters from the font library according to the above-mentioned overall shape features, A first stroke feature representing the characteristics of the strokes constituting the written character is extracted from the written character information, and based on the first stroke feature, the strokes of the written character are compared with the criteria selected by the font recognition unit in consideration of the order of strokes of the written character. The corresponding naming of character strokes, calculating the stroke order corresponding naming recognition unit of a plurality of standard characters selected by the above-mentioned font recognition unit relative to the similarity of the above-mentioned written characters, and A first output unit that outputs the standard characters selected by the above-mentioned font recognition unit according to the above-mentioned similarity calculated by the above-mentioned stroke order corresponding to the name recognition unit. 9. An online character recognition device, having a writing information input unit for inputting its writing information along with the writing action of the character, It is characterized in that it also includes: Extract the feature representing the overall shape of the above-mentioned written character from the above-mentioned writing information, select a plurality of font recognition units similar to the standard characters of the above-mentioned written character from the font library according to the above-mentioned feature representing the overall shape, Extract the second stroke feature representing the stroke feature constituting the above-mentioned written character from the above-mentioned handwriting information, and perform the above-mentioned written character stroke and the standard character stroke selected by the above-mentioned font recognition unit in consideration of the number of strokes of the above-mentioned written character based on the above-mentioned second stroke feature. The corresponding naming, calculate the number of strokes corresponding to the recognition unit of a plurality of standard characters selected in the above-mentioned font recognition unit with respect to the similarity of the above-mentioned written characters, A second output unit that outputs the standard characters selected by the above-mentioned font recognition unit according to the above-mentioned similarity calculated by the above-mentioned stroke number corresponding to the name recognition unit. 10. An online character recognition device, having a writing information input unit for inputting the writing information along with the writing action of the character, It is characterized in that it also includes: From the above-mentioned writing information, extract the features representing the overall shape of the above-mentioned written characters, and select a plurality of font recognition units similar to the standard characters of the above-mentioned written characters from the font library according to the above-mentioned features representing the overall shape, A first stroke feature representing the features of the strokes constituting the written character is extracted from the written information, and the strokes of the written character and the standard character selected by the font recognition unit are performed in consideration of the stroke order of the written character based on the first stroke feature. The corresponding naming of strokes calculates the stroke order corresponding naming recognition unit of the similarity of a plurality of standard characters selected in the above-mentioned font recognition unit with respect to the above-mentioned written characters, By comparing the similarity calculated in the above-mentioned stroke-order corresponding naming recognition unit with a predetermined value, it is judged whether to output the first discrimination unit of the above-mentioned standard characters, When discriminating and outputting in the above-mentioned first discriminating unit, the first output unit that outputs the standard characters selected in the above-mentioned font recognition unit according to the above-mentioned degree of similarity calculated in the above-mentioned stroke-order corresponding naming recognition unit, When it is judged not to be output by the first judging unit, a second stroke feature representing a feature of the strokes constituting the written character is extracted from the written information, and the writing is performed in consideration of the number of strokes of the written character based on the second stroke feature. Character strokes and the corresponding naming of the strokes of the standard characters selected in the above-mentioned font recognition unit, calculate the number of strokes corresponding to the similarity of the above-mentioned written characters of a plurality of standard characters selected in the above-mentioned font recognition unit, and name the recognition unit, A second output unit for outputting the standard characters selected by the above-mentioned font recognition unit according to the above-mentioned similarity calculated by the above-mentioned stroke number corresponding to the name recognition unit. 11. According to the online character recognition device recorded in claim 9, it is characterized in that: The above-mentioned stroke number corresponding naming recognition unit is composed of a DP matching unit and a similarity calculation unit; The DP matching unit extracts the second stroke features representing the features of the strokes constituting the written characters from the written information, and performs DP matching between the strokes of the written characters and the standard characters selected by the font recognition unit based on the second stroke features. The corresponding name of the stroke; The above-mentioned similarity calculation unit extracts the stable strokes corresponding to the above-mentioned written character strokes and the above-mentioned standard character strokes from the correspondingly named strokes in the above-mentioned DP matching unit, and calculates the value of the standard characters selected in the above-mentioned font recognition unit with respect to the above-mentioned stable strokes. The degree of similarity to the above-mentioned written characters; The second output unit outputs the standard characters selected by the font recognition unit of the above category according to the similarity calculated in the similarity calculation. 12. The online character recognition device according to claim 10, characterized in that it has A second judging unit for judging whether to output the standard character by comparing the similarity of the standard character calculated in the font recognition unit with respect to the written character with a predetermined value, and When it is judged as output in the above-mentioned second judging unit, the third output unit that outputs the standard characters selected in the above-mentioned font recognition unit according to the upper similarity degree calculated in the above-mentioned font recognition unit, The stroke-corresponding name recognition unit performs processing when it is judged that there is no output by the second judging unit.

Images