KR100299725B1 - Character recognition apparatus and character recognition method - Google Patents

Abstract

The present invention relates to a handwritten character recognition method and apparatus for recognizing handwritten characters inputted on-line in a high-precision and high-speed order, (Input pattern) is inputted to the input means 1, the stroke extracting means 2 extracts the feature of the stroke constituting the input hand-written character, and based on this feature, The classifying means 4 classifies the stroke into a linear stroke, a non-linear stroke, and a continuous character stroke, and characterizes each of the classified strokes by means of the linear stroke corresponding means 5, the non-linear stroke corresponding means 7, The stroke feature corresponding to the character to be recognized (standard pattern) stored in the stroke feature dictionary 3 is associated with the stroke feature by the stroke correspondence adding means 8, And the character string evaluating means (10) is adapted to recognize the handwritten character and the continuous character stroke correspondence adding means (8) corresponds to the linear stroke correspondence adding means (5) or the nonlinear stroke correspondence adding means And performs correspondence addition with respect to the added recognition target character.

Description

Character recognition apparatus and character recognition method

The present invention relates to hand-written character recognition that reads handwritten characters written with a pen on a tablet of the present invention. The handwritten characters are classified into respective strokes constituting a character, and based on the characteristics of the classified strokes To a character recognition apparatus and a character recognition method for recognizing a character to be recognized.

An on-line handwriting (handwritten) character recognition apparatus that absorbs both of the conventional handwriting fluctuation and the stroke number fluctuation describes the strokes of handwritten characters as feature point arrays, and based on the assumption that stroking occurs in the order of strokes, The two complementary optimal correspondence addition algorithms, such as solving type and lacking correspondence solving type, allow strokes of a handwritten character pattern including a stroke and a stroke count variation, strokes of a standard character pattern by a handwritten stroke number 1, and the stroke corresponding to the stroke corresponding to the strokes is absorbed by the corresponding portion having the smallest distance in the case where the strokes are combined with the strokes before and after the strokes, On-line handwritten Chinese character recognition, NTT R & D Vol. 45 No.11 1996).

For example, FIG. 34 is a block diagram showing the configuration of a conventional apparatus for absorbing the fluctuation in the number of strokes and strokes exemplified in the conventional example. In Fig. 34, reference numeral 201 denotes input means for inputting an input pattern which is a handwritten character written on the tablet with a pen, reference numeral 202 denotes a preprocessing means for performing normalization such as sampling, position and size of an input pattern, 204 is a standard pattern dictionary storing a standard pattern which is an average character pattern written in a plurality of letters written in correct stroke order and stroke count, 205 is a standard pattern dictionary storing a standard pattern, A stroke calculation means 206 for calculating a distance between a stroke of the input pattern and a stroke of the standard pattern stroke, 206 denotes a one-to-one correspondence between the strokes of the input pattern and the strokes of the standard pattern by two types of complementary search algorithms One-stroke countermeasure adding means 207 for absorbing the fluctuation of the peeling order by the reference numeral 207, By selectively combining the stroke engagement means for absorbing the fluctuation of strokes, 208 is a pattern distance calculating means for calculating a distance between the pattern based on the corresponding additional results, 209 is a control means for controlling the respective means.

The operation of the conventional example will be described with reference to the block diagram of FIG. First, the control means 209 instructs the input means to obtain an input pattern.

Next, the control means 209 transfers the input pattern input by the input means 201 to the preprocessing means 202, and performs sampling, normalization of the position and size.

The control means 209 transfers the pre-processed input pattern to the feature point extracting means 203, and converts the input pattern into feature point rows formed at regular intervals on each stroke of the input pattern.

The control means 209 sends the feature point sequence of the input pattern to the one-to-one stroke correspondence adding means 206. The one-to-one stroke countermeasure adding means 206 performs the correspondence addition so that the input pattern and the standard pattern of the standard pattern dictionary 204 are such that there is not an excessive correspondence portion of each feature point on the basis of the stroke with the smallest stroke number (the over- A corresponding addition algorithm of FIG.

Next, the one-to-one stroke countermeasure adding means 206 performs counterparting so that the input pattern and the standard pattern of the standard pattern dictionary 204 have no corresponding portion lacking each characteristic point based on the stroke of the larger stroke number A corresponding solution type of the corresponding solution type).

A result obtained by the final one-to-one-stroke countermeasure adding means 206 as a result of the smallest distance among the corresponding addition results obtained by the two counterpart attachment algorithms.

Here, the distance calculated by the inter-stroke distance calculating means 205 at the time of the corresponding addition is calculated by dividing the sum of the distances between the start points and the end points by 2, that is, the end point matching distance, The partial matching is a value obtained by multiplying the point ratio by calculating the sum of the point-to-point distances corresponding to the starting point of the stroke on the side of the stroke.

Next, the selective stroke combining means 207 arranges the strokes corresponding to one another by the one-stroke corresponding means 206 in accordance with the stroke order of the many strokes. With respect to the strokes corresponding to each other by the one-to-one stroke countermeasure adding means 206, the strokes are combined based on the assumption that the engagement of the strokes occurs in the order of the strokes.

Concretely, the non-corresponding strokes written before the leading strokes of the strokes corresponding to each other by the one-to-one stroke countermeasure means 206 are joined in the order of the leading strokes. Similarly, with respect to unstable strokes written after the final strokes, the strokes are concatenated in the order of the final strokes. Since the other non-corresponding strokes exist between any one of the strokes corresponding to the one-to-one stroke corresponding means 206, it is possible to combine the non-corresponding strokes with the front and rear strokes once, Is cut to the minimum, thereby dividing into two strokes.

Here, the selective stroke combining means 207 calculates the sum of the distances between the points on the side having the higher score at regular intervals, corresponding to the score of the smaller one, and dividing the sum by the score Use the entire matching distance. However, the endpoint matching distance is used in the large classification process for high speed.

Finally, with respect to the combined stroke, a proper distance is normalized to calculate the final distance, and the character having the smallest distance is determined as the recognition result.

As described above, in the conventional character recognition apparatus for absorbing fluctuation in the order of strokes and the number of strokes, the distance between strokes is obtained by using the feature point row in which each stroke of the handwritten character is equally spaced. Therefore, A distortion of a stroke or a noise component such as a character or a distortion of a stroke caused by a positional shift, and a shift of a coordinate value of a minutiae due to a positional shift affect the stroke distance.

In the case of a character with many strokes such as Chinese characters, individual strokes are often composed of simple straight strokes. Therefore, even though the minutiae in the stroke do not contain important information, And it is easily affected by the deviation of the coordinate values.

In order to reduce the amount of calculation in absorbing the stroke variation, there is a problem in that the assumption that "the combination of strokes occurs in the order of stroke" is used, and that the stroke variation can not cope with fluctuations that do not occur in the order of the strokes.

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above-mentioned problems, and it is an object of the present invention to provide a stroke control device and a stroke control method, It is an object of the present invention to provide a character recognition apparatus and a character recognition method capable of achieving high-precision errors that are difficult to be affected and capable of character recognition at high speed.

A stroke feature dictionary in which features of straight stroke and non-straight stroke constituting respective recognition target characters for a plurality of recognition target characters are already stored; A stroke feature extraction means for extracting a feature of a stroke constituting the handwritten character from the handwritten character inputted by the input means, and a stroke feature extraction means for extracting, based on the feature extracted by the stroke feature extraction means, Stroke classifying means for classifying each stroke of the written character into one of the straight stroke, the non-linear stroke, or a continuous character stroke in which the corresponding character can not be matched with the stroke constituting the recognition target character; And the stroke characteristics are stored in advance. Linear stroke correspondence adding means for adding a feature of a linear stroke constituting a character to be expressed to a character string to be recognized and a nonlinear stroke characterizing means for characterizing a nonlinear stroke characterized by the stroke classifying means and a non- Linear strokes corresponding to the characteristics of the straight strokes; and a nonlinear stroke corresponding addition means for assigning the characteristics of the linear strokes so as to correspond to the characteristics of the straight stroke or the nonlinear stroke, Continuous stroke correspondence adding means for adding a feature of a stroke and a feature of a straight stroke or a non-straight stroke constituting a recognition target character stored in advance in the stroke feature, Additional means or phase And a character evaluation means for recognizing the handwritten character based on a result of the correspondence result obtained by the consecutive character stroke correspondence adding means, wherein the continuous character stroke correspondence adding means comprises: And the correspondence is added to the character to be recognized corresponding thereto by the linear stroke correspondence adding means.

A second aspect of the present invention is a stroke characterizing method comprising the steps of: inputting a handwritten character on-line; extracting a stroke character constituting the handwritten character from the handwritten character input by the inputting step; Wherein each stroke of the handwritten character is divided into a plurality of consecutive character strokes that can not be matched with the stroke constituting the straight stroke, the non-straight stroke, or the character to be recognized, based on the feature extracted by the stroke feature extracting step A linear stroke correspondence adding step of performing correspondence between the characteristics of the straight stroke classified by the stroke classification step and the characteristics of the straight stroke constituting the recognition target character stored in the stroke characteristic dictionary; Wherein the stroke classification step A nonlinear stroke correspondence addition step of performing a correspondence between a characteristic of a straight stroke and a characteristic of a nonlinear stroke constituting a recognition target character stored in advance in the stroke characteristic; Stroke strokes and the nonlinear strokes, and the character string is divided into a non-linear stroke and a non-linear stroke, and a continuous character or a non-linear stroke constituting the recognition target character stored in the stroke characteristic dictionary, And a character evaluation step of recognizing the handwritten character based on the stroke corresponding addition step, the linear stroke corresponding addition step, the nonlinear stroke corresponding addition step or the corresponding addition result obtained by the continuous character stroke corresponding addition stepA.

According to a third aspect of the present invention, the step of adding the continuous character strokes corresponds to the step of responding to the character to be recognized, to which the step corresponding to the straight stroke corresponds or the step corresponding to the nonlinear stroke.

1 is a configuration diagram of a character recognition apparatus according to a first embodiment;

2 is a flowchart showing a process flow of the character recognition apparatus according to the first embodiment;

Fig. 3 shows an input pattern " region " written in the input means 1 of Embodiment 1. Fig.

4 is a diagram showing stroke characteristics of an input pattern extracted by the stroke feature extracting means 2 of the first embodiment;

Fig. 5 is a view showing allocation of directions of strokes in the stroke characteristics of Embodiment 1, and directions for quantizing the directions of virtual strokes in 16 directions. Fig.

6 is a flow chart showing the flow of processing of the stroke classifying means according to the first embodiment;

Fig. 7 is a view showing an 8-way component distribution used when the stroke classifying means of the first embodiment obtains a direction component distribution; Fig.

8 is a flowchart showing the processing flow of the linear stroke corresponding means 5 of the first embodiment;

Figs. 9A and 9B are diagrams showing area division used when the stroke correspondence attachable area determining means 6 of the first embodiment is a corresponding portion of a straight stroke. Fig.

10 is a view showing information for determining a corresponding supplementary area used by the stroke correspondence addition possible area determining means 6 of the first embodiment;

11 is a diagram showing a correspondence attachable area of the input stroke of the first embodiment;

12 is a diagram showing a stroke column of a standard pattern of the character " a " in the stroke feature dictionary of the embodiment 1. Fig.

Fig. 13 is a diagram showing stroke characteristics in a standard pattern of the character " a " in the stroke feature dictionary of the first embodiment; Fig.

Fig. 14 is a diagram showing the corresponding portion according to the stroke shape used in the linear stroke corresponding means 5, the non-linear stroke corresponding means 7 and the continuous character stroke corresponding means 8 of the first embodiment;

15 is a diagram showing a form of compression of candidate strokes in the linear stroke corresponding means 5 of the first embodiment;

16 is a flowchart showing the processing flow of the nonlinear stroke countermeasure adding means 7 of the first embodiment;

17 is a flowchart showing the flow of processing of the continuous character stroke corresponding means 8 of the first embodiment;

18 is a diagram showing a region division used in the case where the stroke corresponding area determining means 6 of the first embodiment is a corresponding portion other than the linear stroke;

Fig. 19 is a diagram showing where the respective strokes of the input pattern of the first embodiment exist in the area used by the stroke-area-addition possible region determining means 6. Fig.

20 is a flowchart showing a flow of detailed processing in step S54 in the first embodiment;

Fig. 21 is a diagram showing characteristic points in the stroke of the standard pattern of the first embodiment; Fig.

22 is a view showing sampling points obtained by sampling the strokes of the input pattern according to the first embodiment at regular intervals;

23 is a diagram showing the result of adding the minutiae point of the standard pattern of the first embodiment and the sampling point of the input pattern.

24 is a diagram showing a stroke in which the input pattern of Embodiment 1 is divided into strokes of a standard pattern.

25 is a diagram showing each stroke of a character " WORD " written in the input means 1 of the first embodiment;

26 is a diagram showing where each stroke of the input pattern of the first embodiment exists in the area used by the stroke-corresponding addition area determining means 6. Fig.

27 is a view showing a standard pattern of a character " word " in the stroke feature dictionary (3) according to the first embodiment;

28 is a flow chart showing the flow of detailed processing of the corresponding portion of partial strokes of Embodiment 1. Fig.

29 is a diagram showing division candidate points of consecutive character strokes of the input pattern according to the first embodiment;

30 is a view showing a partial stroke of the first embodiment;

FIG. 31 is a flowchart showing a detailed flow of the process of step S72 of FIG. 28; FIG.

32 is a view showing the division stroke obtained in the partial stroke of FIG. 30;

33 is a view showing the division stroke obtained in the partial stroke;

34 is a configuration diagram of a conventional on-line character recognition apparatus.

Description of the Related Art [0002]

1: Input means 2: Stroke feature extracting means

3: stroke characteristic dictionary 4: stroke classification means

5: means for adding a linear stroke

6: stroke correspondence attachable area determining means

7: Non-linear stroke countermeasure attachment means

8: Continuous character stroke countermeasure adding means

9: partial stroke extracting means 10: character evaluation means

11: Control means

(Example 1)

Hereinafter, Embodiment 1 will be described with reference to the drawings.

1 is a configuration diagram of a character recognition apparatus according to the present invention,

In the figure, reference numeral 1 denotes an input means for inputting a character by hand to the tablet type input device, 2 denotes a feature extracting means for extracting a feature of a stroke constituting the handwritten character from the handwritten character input by the input means 1 3 is a stroke feature dictionary in which stroke characteristics of linear strokes and non-linear strokes constituting a standard pattern, which are each character to be recognized, recorded in correct pilt order, are stored in advance in a plurality of recognition target characters; Stroke stroke classifying means for classifying each stroke of a handwritten character into one of a linear stroke, a non-linear stroke, and a continuous character stroke based on the characteristic extracted by the stroke characteristic extracting means (2).

Reference numeral 5 denotes a linear stroke correspondence adding means for performing correspondence between the features of the linear stroke classified by the stroke classifying means 4 and the features of the linear strokes constituting the standard pattern stored in the stroke feature dictionary 3, A stroke corresponding additional area determining means 7 for determining a present area of each stroke of the handwritten character based on the feature extracted by the extracting means 2, A nonlinear stroke correspondence adding means for performing correspondence of the characteristics of the nonlinear stroke constituting the standard pattern stored in the stroke characteristic dictionary 3 and the nonlinear stroke correspondence means constituting the standard pattern stored in the dictionary dictionary 3, 8 a continuous character stroke classified by the stroke classifying means, It is divided into a straight stroke or a non-straight stroke, and the characteristics of this divided straight stroke or non- A character string correspondence adding means 9 for performing correspondence of characteristics of a linear stroke or a non-linear stroke constituting the standard pattern stored in the lock feature dictionary 3, and 9 is a partial stroke creating means for creating a partial stroke from a continuous character stroke .

Reference numeral 10 denotes a character recognition unit which obtains an evaluation value of a character from the result of the correspondence result obtained by the linear stroke correspondence adding unit 5, the nonlinear stroke correspondence adding unit 7, and the continuous character stroke correspondence adding unit 8, Character evaluation means, and 11 is control means for controlling each means.

Fig. 2 is a flowchart showing the flow of processing of the character recognition apparatus shown in Fig. 1. Fig.

An example of the processing procedure will be described with reference to a flowchart showing the processing flow of the character recognition apparatus in Fig.

First, in step S1, the input means 1 writes characters to the tablet type input device with a pen.

Fig. 3 shows an input pattern " region " of a handwritten character inputted to the input means 1, and numerals 60 to 65 are strokes inputted in order in time. In this example, the handwriting is not regular, with stroke and stroke count, and consists of six strokes.

The flow advances to step S2 and the stroke feature extracting means 2 extracts the feature of each stroke constituting the input pattern of handwritten characters. In this example, the stroke characteristics of the strokes 60 to 65 of the input pattern of Fig. 3 are extracted.

4 is a stroke characteristic of the input pattern extracted by the stroke characteristic extracting means 2 and shows the stroke characteristic obtained from the strokes 60 to 65 of the input pattern of Fig. Among the above stroke characteristics, the directions of the strokes and the directions of the virtual strokes that are actually untouched, which connect one stroke to the next, are a 16-way distribution diagram in which the directions are assigned to quantize the stroke shown in Fig. .

Next, proceeding to step S3, the stroke classifying means 4 classifies the stroke of the input pattern according to the shape thereof.

Fig. 6 is a flowchart showing the flow of the processing amount of the stroke classifying means 4. Fig. The operation of the stroke classifying means 4 will be described in detail with reference to the flowchart of FIG.

First, in step S10 of FIG. 6, the stroke classification means 4 determines whether the current stroke is a linear stroke in order from the first stroke. If the stroke is a straight stroke, the stroke classification means 4 proceeds to Step S11. Otherwise, it proceeds to Step S13 do. In the input pattern of Fig. 3, five strokes (60, 61, 62, 63, 65) are determined as straight strokes from the shape of the stroke feature of Fig.

In step S11, the stroke classification means 4 extracts the direction component distribution of the stroke. The direction component distribution of the stroke is obtained by the direction distribution of each straight line connecting points sampled at appropriate intervals of the input stroke. Here, the 8-way component distribution shown in Fig. 7 is used when the stroke classification means 4 obtains the direction distribution.

The flow advances to step S12. In step S12, the stroke classifying means 4 determines whether the directional component obtained is all in a predetermined direction. If the directional component is only a predetermined direction, the stroke classification means 4 proceeds to step S15 and determines that the stroke is a straight stroke. In the example of the input pattern in Fig. 3, since all the linear strokes have only components in a certain direction, the process proceeds to step S15, where the linear strokes are determined.

If the direction component is not the only direction, the stroke classification means 4 checks whether or not the stroke shape exists in the stroke feature dictionary 3 with respect to the straight stroke having the multi-directional component and the stroke other than the straight line in step S13 If the stroke is not present in the dictionary, the process advances to step S16 to determine a continuous character stroke. If the stroke is present in the dictionary, the flow advances to step S14 to determine a non-linear stroke. In the example of the input pattern in Fig. 3, since the unknown stroke 64 is a stroke that does not exist in the dictionary, the process proceeds to step S16 to determine a continuous character stroke.

2, the control means 11 transfers the stroke determined by the stroke classification means 4 to the linear stroke to the linear stroke corresponding means 5, and the linear stroke of the input pattern The strok feature dictionary 3 corresponding to the linear strokes of the standard pattern. In the example of the input pattern shown in Fig. 3, the linear strokes 60, 61, 62, 63 and 65 are transferred to the linear stroke corresponding means 5.

8 is a flowchart showing the flow of the process of the linear stroke correspondence adding means. The operation of the linear stroke corresponding means 5 will be described with reference to the flowchart of Fig.

First, the linear stroke correspondence adding means 5 classifies each linear stroke into three kinds (long, medium, and short) according to the length in step S21. The stroke length is determined by the Euclidean distance of the start point and the end point of the stroke. In this example, a long stroke is defined as 70 or more, a short stroke is defined as 30 or less, and an intermediate stroke is defined as an intermediate stroke. The strokes 60, 61, 62 and 65 are classified as long strokes, and the strokes 63 are classified as short strokes from the input stroke length information illustrated in the stroke characteristic of the input pattern of Fig.

The routine proceeds to step S22 where the linear stroke correspondence adding means 5 determines the existence area of the stroke from the stroke direction, the start and end point coordinates of the stroke, and the stroke length in the stroke correspondence region determining means 6 do.

9A and 9B show an example of area division used in the corresponding portion of the linear stroke. Here, there is shown an example of the case where the area where the strokes exist is determined by four divided areas in the horizontal direction and the vertical direction, respectively, in which the character is an externally connected rectangle.

Fig. 10 shows information for area determination used in the stroke correspondence addition region determining means 6. Fig. The stroke direction in Fig. 10 represents a direction obtained by quantizing the direction from the start point to the end point of the stroke to the 16-direction distribution diagram in Fig. For example, with respect to the horizontal straight stroke of FIG. 10, the stroke direction indicates that there is a possibility of corresponding to the horizontal stroke of the standard pattern in the directions of 4, 5, 6, 12, 13, The horizontal straight stroke existing in the region 1 indicates that there is a possibility of corresponding to the horizontal straight stroke existing in the region 1 and the region 2. [

As shown in Fig. 10, the stroke corresponding area determining means 6 determines that the stroke of the input pattern is relatively stable because the variation in the distance in the vertical direction is relatively small, and the area up to the adjacent area is the existing area. However, in a case where the data is spread over a plurality of regions, the adjacent region for the plurality of regions is defined as the existing region.

Fig. 11 shows the existence area of the stroke and the correspondence addition area with respect to the input pattern of Fig. 3. Fig.

The routine proceeds to step S23 where the linear stroke correspondence adding means 5 determines the stroke of the standard pattern of the stroke feature dictionary 3 to which the correspondence is to be added, with respect to the stroke of the input pattern.

12 shows a stroke of a standard pattern of the character " A " stored in the stroke feature dictionary 3. [ 13 is a stroke characteristic of the standard pattern of Fig. 12 stored in the stroke feature dictionary 3. Fig.

The corresponding conditions of the correspondence are determined based on three types of stroke conditions: 1) stroke condition; 2) stroke length condition; and 3) stroke condition correspondence addition condition.

Fig. 14 shows whether or not the corresponding portion is added in accordance with the stroke shape. The stroke shape used for the linear stroke corresponding means 5, the non-linear stroke corresponding means 7, and the continuous character stroke corresponding means 8 And determines whether or not the corresponding part according to the present invention is appropriate.

More specifically, the shape conditions of the strokes are such that the corresponding portions to which a circle is added in the stroke shape of the input pattern and the stroke pattern of the standard pattern in the correspondence relationship shown in Fig. 14, .

The length of the stroke can be set to correspond to the long stroke or the intermediate stroke, the intermediate stroke can correspond to the long stroke and the short stroke, and the short stroke makes it possible to correspond to the intermediate stroke or the short stroke . In addition, the condition of the region in which the stroke can be corresponded uses the result of the stroke correspondence attachable region determining means 6.

In this example, the stroke characteristic shown in Fig. 13 is stored in the stroke characteristic dictionary 3 with respect to the standard pattern shown in Fig. Based on this stroke characteristic, a stroke of a standard pattern corresponding to each stroke of the input pattern is determined.

In this embodiment, for each stroke of the input pattern shown in Fig. 3, the strokes of the standard patterns corresponding to each other are determined.

Fig. 15 shows a state in which a stroke of a standard pattern corresponding to a stroke shape condition, a stroke correspondence attachable area condition, and a stroke length condition is compressed. Here, compression based on each condition is sequentially performed on only those compressed to that extent.

15, the stroke 60 corresponds to the strokes 70 and 73, the stroke 61 corresponds to the strokes 74 and 75, the stroke 62 corresponds to the stroke 75, the stroke 63 corresponds to the stroke 71, , The stroke 65 is compressed to the strokes 73 and 76. As shown in Fig. Here, since there is no other correspondence with respect to the stroke compressed by one, the coordinate stroke compressed by one can not be used for the corresponding addition of the other input strokes. Here, since the strokes 71 and 75 are compressed to one, the strokes 71 and 75 are excluded from the candidate strokes of other input strokes. Thereby, the stroke 61 is compressed to the stroke 74, and finally, there are only two strokes candidates in the stroke 60 and the stroke 65 only.

Next, in step S24, the linear stroke correspondence adding means 5 performs in detail the correspondence of the stroke of the input pattern to the candidate stroke of the standard pattern, As an addition result.

Here, in the check of the corresponding counterpart, it is checked whether or not the correspondence is possible by comparing the stroke direction, the stroke width, the stroke height, and the coordinate value of the stroke stop point with the threshold value of the predetermined counterpart. At this time, when the input strokes correspond to the consecutive strokes of the standard pattern, the corresponding portions of the virtual strokes are also used to perform the verification of the corresponding portions.

In addition, in order to absorb the input pattern peeling order variation, it is preferable that all the candidate strokes of the strokes of the standard pattern, which correspond to the strokes of the input patterns, so that the input patterns can be assigned to the strokes even when the input patterns are different from those of the standard patterns A countermeasure is added.

More specifically, the stroke characteristic of the input pattern of FIG. 4 is compared with the stroke characteristic of the standard pattern of FIG.

In this example, as for the strokes 61, 62, and 63 of the input pattern, the candidates for the corresponding portions are already compressed to one, and since both strokes are correct correspondent portions, the corresponding portions succeed. Since there are two candidates for the strokes 60 and 65, the candidate strokes of the standard pattern are corresponded to the strokes of the respective input patterns. That is, the stroke 70 of the standard pattern and the stroke 73 correspond to the stroke 60 of the input pattern, and the stroke 73 and the stroke 76 correspond to the stroke 65, respectively.

First, the correspondence processing of the stroke 60 of the input pattern will be described. The strokes 60 and the strokes 70 and 73 of the standard pattern are made to correspond to each other. Here, when the stroke 60 and the stroke 73 correspond to each other, the difference between the Y coordinate values of the continuous point is increased. If the result of the correspondence is checked by using the virtual stroke indicating the positional relationship between the strokes, the direction of the virtual stroke to the next stroke 74 in the case of the stroke 73 becomes "16" from FIG. The direction of the virtual stroke of the stroke 60 is the same as the direction of the stroke 61 at the end point of the stroke 60 because the stroke of the input pattern corresponding to the standard stroke 74 in the case of the stroke 60 is the stroke 61. [ , And becomes " 13 ". Therefore, the difference in the virtual stroke direction becomes 16-13 = 3.

On the other hand, in the corresponding part of the stroke 60 and the stroke 70, the difference in the point coordinate values is small, and the difference in the direction of the virtual stroke is small (the distance of the corresponding result is small).

Then, the strokes 65 and the strokes 73 and 76 of the standard pattern are corresponded to each other. When the stroke 65 corresponds to the stroke 73, the difference in the coordinate value of the continuous point (the difference in the stroke characteristic) increases. On the other hand, when the stroke 65 and the stroke 76 are added to each other, the difference in the stroke characteristic is small and an accurate counterpart.

As a result, the corresponding portion of the linear strokes corresponds to the stroke 60 of the input pattern, the stroke 61 of the standard pattern, the stroke 62, the stroke 75, 63 are added to the stroke 71 and the stroke 65 is added to the stroke 76 to be compressed in a one-to-one correspondence.

As in the above example, even if the stroke order of the input pattern is different from the stroke order of the standard pattern by the corresponding addition processing, the corresponding addition can be performed.

When the linear stroke of the input pattern corresponds to any one of the strokes of the standard pattern, the process proceeds to step S25, where the linear stroke correspondence adding means 5 proceeds to step S27 to make the corresponding addition possible (OK) If there is a non-conforming stroke, the process proceeds to step S28, and the addition of the correspondence with the current standard pattern is made impossible (NG).

In this example, the linear strokes of all the input patterns were exactly matched with the strokes of the standard pattern.

Next, the process proceeds to step S5 of FIG. 2, where the control means 11 checks the result of adding the linear stroke, and if the corresponding addition is impossible (NG), the process proceeds to step S8, . If the counterpart attachment is successful, the process proceeds to step S6, and the non-linear strokes are added.

FIG. 16 is a flowchart showing the flow of processing of the nonlinear stroke countermeasure adding means 7; FIG. The operation of the nonlinear stroke countermeasure adding means 7 will be described in detail with reference to the flowchart of FIG.

In step S30, the nonlinear stroke correspondence adding means 7, as in the case of the linear stroke, sets the stroke correspondence addition possible region determining means 6 with respect to the nonlinear stroke, . This correspondence section refers to the information diagram for determining the area in Fig. 10 like the portion corresponding to the linear stroke.

Here, the correspondence between the non-linear strokes of the input pattern and the strokes of the standard pattern is made only for the standard pattern in which the linear strokes corresponding to the linear stroke corresponding means 5 are present.

The routine proceeds to step S31 where the nonlinear stroke countermeasure adding means 7 performs the correspondence with the corresponding stroke of the standard pattern for each stroke shape of the input pattern.

First, a check is made to determine whether or not the stroke corresponding to the standard pattern corresponds to each stroke of the input pattern obtained in the process of step S30 by the stroke shape. The rule of the corresponding part of the stroke shape refers to the diagram showing the corresponding part of the corresponding part in accordance with the stroke shape in Fig.

Next, when the counterpart can be attached, the read stroke is checked for the corresponding portion of the detailed stroke in the same manner as the linear stroke.

The routine proceeds to step S32. In the case where it is determined that the stroke of the input pattern corresponds to any one of the strokes of the standard pattern, the nonlinear stroke correspondence adding means 7 proceeds to step S34 and determines that the corresponding addition succeeds. If no corresponding portion is found, it is regarded that there is no match with the stroke of the standard pattern, and the process proceeds to Step S33 to determine the continuous character stroke.

The routine proceeds to step S35. In step S35, the nonlinear stroke countermeasure adding means 7 checks whether all nonlinear standard strokes have corresponded to each other. If not, the routine proceeds to step S30, and the next nonlinear stroke counter is added. In addition, when all the corresponding portions have been terminated, the corresponding portion of the non-linear stroke corresponding means 7 is terminated.

Next, the process proceeds to step S7 in Fig. 2, and the control means 11 performs the correspondence of the continuous character strokes to the continuous character stroke correspondence adding means 8.

17 is a flowchart showing the flow of processing of the continuous character stroke correspondence adding means 8; Hereinafter, the operation of the continuous character stroke correspondence adding means 8 will be described with reference to the flowchart of Fig.

First, in step S50, the continuous character stroke correspondence adding means 8 determines the correspondence addition capable area of the input character string of the input pattern by the stroke correspondence addition possible region determining means 6, and corresponds to the continuous character stroke of the input pattern A stroke candidate of a standard pattern that can be added is obtained.

Here, the continuous character stroke correspondence portion of the input pattern is determined only for the standard pattern in which the linear stroke correspondence means 5, the linear stroke corresponding to the nonlinear stroke correspondence means 7, and the nonlinear stroke exist All.

18 is a diagram showing an example of area division used when the stroke correspondence addition possible region determination means 6 is a corresponding portion other than the linear stroke. In the case other than the stroke shown in the information diagram for determining the correspondence addition possible region in Fig. 10, the stroke correspondence attachable region determining means 6 determines the stroke correspondence attachable region using the region dividing in Fig. The stroke correspondence attachable area is intended to extend to one area outside the area including the external connection rectangle of the stroke of the input pattern. In the example of the input pattern of Fig. 3, since the stroke 64 is determined as a continuous character stroke and is not the stroke shape shown in Fig. 10, the region in which the stroke 64 is included is determined first.

Fig. 19 is an example in which the input pattern of Fig. 3 is divided into regions of Fig. 18 to obtain the external connection rectangle 80 of the stroke 64, and each stroke of the input pattern corresponds to the stroke- Which is used in FIG. In this drawing, it can be seen that the external connection rectangle 80 of the stroke 64 is included in the areas A to H, and the areas in which strokes can be corresponded to are areas A to L, respectively.

Next, the continuous character stroke correspondence adding means 8 obtains a stroke of the standard pattern existing in the determined correspondence addition enabling area. In the case of the stroke example of the standard pattern shown in Fig. 12, in the case of four examples of the strokes 70, 71, 72 and 73 included in the areas A to L, the linear stroke corresponding means 5 and the non- The strokes of the standard patterns corresponding to the input strokes in one-to-one correspondence by the stroke corresponding means 7 are excluded because the corresponding portions are determined. Therefore, in this example, the strokes 70 and 71 are excluded because they are fixed by the linear stroke corresponding means. As a result, the strokes 72 and 73 are selected as the candidate strokes.

The process proceeds to step S57, and the continuous character stroke correspondence adding unit 8 checks whether or not the correspondence addition possible area has been determined for all consecutive character strokes. If the determination result is YES, the process proceeds to step S51. If not, .

Since there is only one continuous character stroke in the input pattern of Fig. 3, the process proceeds to step S51, and the candidate strokes of the standard pattern which are the start strokes and the end strokes of the continuous character strokes of the input pattern are determined.

Concretely, first, the region where the dot is present for the consecutive points of the continuous character strokes of the input pattern is determined by region dividing in FIG. 18, and the area up to one outside of the existing region is set as a correspondence attachable region.

Next, a stroke in which the start point is included in the stroke of the standard pattern corresponding to the correspondence possible area where the start point of the continuous character stroke of the input pattern is present is selected as the candidate of the start stroke. Similarly, a stroke whose end point is included in the stroke of the standard pattern corresponding to the correspondence addition area in which the end point of the continuous character stroke of the input is present is selected as the candidate of the end stroke.

In the example shown in Fig. 19, since the start point of the continuous character stroke 64 of the input pattern exists in the area A, the area in which the corresponding part of the start point can correspond is the area A, B, E, F. Since all the time points of the strokes 72 and 73, which are candidates of the standard pattern obtained in the process of step S50, are included in this area, two strokes 72 and 73 are candidates for the start strokes. Similarly, since the end point of the continuous character stroke 64 exists in the area H, the corresponding addable area becomes the area C, D, G, H, K, L. As a result, the candidates of the end point strokes become strokes 72 and 73 as well.

The process proceeds to step S52, and the continuous character stroke correspondence adding unit 8 checks whether the sum of the number of strokes of the continuous point candidates obtained in step S51 (however, the same strokes are not repeatedly counted) becomes three or less, If the number is three or less, the process proceeds to step S54, and if the number is three or more, the process proceeds to step S53. In the case of the continuous character stroke in Fig. 3, since the number of candidate strokes is two, the flow proceeds to step S54.

The correspondence between the continuous character strokes of the input pattern and the candidate strokes of the standard pattern is made in step S54. 20 is a flowchart showing the flow of the counterpart attachment process in step S54. With reference to the flowchart of Fig. 20, corresponding portions of continuous character strokes will be described in detail.

First, in step S60, the continuous character stroke corresponding means 8 determines the order of the candidate strokes of the standard pattern. Here, in step S60, the number of candidate strokes of the standard pattern is necessarily compressed to three or less. Since the consecutive character strokes are composed of at least two strokes, the combination of the strokes of the standard pattern in the case of three strokes is 12 types of 3C2 x P2 + P3.

Likewise, when there are two candidate strokes of the standard pattern, there are two types.

In the worst case, the number of combinations is limited to twelve, so that the combination can be sufficiently processed in real time.

1) The constraint by the candidates of the start stroke and the end stroke determined in step S51

2) Restriction by non-erasable stroke

Here, erasable non-stroking refers to a stroke of a standard pattern not corresponding to the stroke candidate of the current input pattern. Since the candidates of the standard pattern strokes corresponding to the strokes of the input patterns are determined in the process of step S50, the erase strokes are determined by this information.

In the case of the stroke 64 of the input pattern, there are two candidate strokes, and any of the strokes is a candidate for the start stroke and the end stroke, and the strokes do not exist in the erase stroke. In this example, first, the strokes 72 and 73 are made to correspond to each other in the same order as the strokes.

The flow advances to step S61, and the continuous character stroke correspondence adding unit 8 performs the correspondence between the continuous character stroke of the input pattern and the stroke of the standard pattern. Since the stroke feature point information of the standard pattern is a start point and end point only when the stroke is a standard stroke, and the feature point exists around the start point, end point, and bending point of the stroke other than the linear stroke, The feature point stored in the memory 3 is also used. Also, the stroke minutia information of the input pattern is a sampling point expressed at an appropriate interval after the preprocessing.

13, there are three points of the starting point, the minutiae point and the ending point in the stroke, and the minutiae point information of the stroke 73 includes two points of the starting point and the ending point Able to know. Therefore, there are five stroke feature points of the standard pattern. Stroke minutiae points of the obtained standard patterns are indicated by marks in Fig. The characteristic points of the continuous character strokes 64 of the input pattern are indicated by a circle in Fig.

Next, the continuous character stroke correspondence adding means 8 performs a correspondence addition where the distance between the stroke feature point of the input pattern and the stroke feature point of the standard pattern becomes minimum. In general, the correspondence between the stroke A of the input pattern and the stroke B of the standard pattern corresponds to the set of elementary columns of the feature point of the stroke A of the input pattern {0, 1, ... , I}, the set of elementary columns of the feature points of the stroke (B) of the standard pattern {0, 1, ... , J}

ω: {0, 1, 2, ... , I} → {0, 1, 2, ... , J}

≪ / RTI >

Here, ω is an event in which both ends of the stroke (A) and the stroke (B) are coincident with each other and satisfies ω (0) = 0 and ω (I) = j , The order of the components is called a stretch (contraction) thought because it is a monotonous idea that does not allow a reverse rotation. The i-th element of the stroke A is ai, the j-th element of the stroke B is bj, the distance of ai and bj is d (ai, bj): the Euclidean distance between the feature points (ai and bj) Assuming that the cumulative distance from the i-th element of the stroke A to the j-th element of the stroke B is g (i, j)

1? I? I, 0? J? J,

g (i, j) = 0 (k, j)

However, g (0, 0) = d (a0, b0) and g (0, j) = ∞ (when j> 0) are established.

Thus, the information of the corresponding part of the stroke A of the input pattern and the stroke B of the standard pattern is obtained. Here, if the value of k providing min {g (i-1, k) | 0? K? J) is recorded, information of the optimum corresponding portion between the minutiae points is obtained. That is, whether or not each feature point of the stroke of the standard pattern corresponds to a feature point such as a continuous character stroke of the input pattern.

When the feature point of the stroke of the standard pattern in Fig. 21 and the corresponding feature of the stroke feature point (sampling point) of the input pattern in Fig. 22 are performed by the dynamic programming method, do.

The flow advances to step S62, and the continuous character stroke corresponding section 8 obtains the stroke characteristic of the division stroke from the obtained result of the corresponding addition part by the stroke feature extraction section 2. [ In order to obtain the stroke characteristic, first, the continuous character stroke of the input pattern is divided by the feature point added to the stroke point of the standard pattern, and the stroke sandwiched by the return point is extracted as the division stroke.

Next, the stroke characteristic extracting means 2 extracts the stroke characteristic of the division stroke. 24 shows an example in which the division strokes 130 and 131 are obtained from the stroke 64 of the input pattern.

The process proceeds to step S63, and the continuous character stroke correspondence adding unit 8 checks whether or not the correspondence addition with the continuous character stroke of the input pattern is performed with respect to the current candidate stroke. As a result of the check, if the corresponding addition is not performed in all the strokes, the flow advances to step S60 to perform the correspondence addition for the other stroke orders.

In the case of the stroke 64 of the input pattern, the flow advances to step S60 to perform the correspondence in the order of strokes 73 and 72 in the order of the strokes. Since the corresponding addition process is the same as that in the case of the strokes 72 and 73, description thereof is omitted here.

In this case, since the distances between the stroke minutiae points of the input pattern and the stroke minutiae points of the standard pattern become smaller in the case of the strokes 72 and 73, the corresponding partial strokes are selected as a result.

The continuous character stroke correspondence adding means 8 proceeds to step S56 in FIG. 17 when the correspondence of all the consecutive character strokes to the continuous character strokes of the current input pattern is completed, and judges whether or not the correspondence is added for all the consecutive character strokes If it is determined that the correspondence has been completed, the process of the correspondence adding unit 8 for the continuous character stroke is terminated and the process proceeds to step S8 in Fig. When the counterpart attachment has not been completed, the process returns to step S51 of FIG. In the input pattern of Fig. 3, since the number of consecutive character strokes is one, the condition is satisfied and the process proceeds to step S8.

In step S8, the control means 11 checks whether all the standard patterns of the stroke feature dictionary 3 have been matched, and if so, the process proceeds to step S9. If not, the process proceeds to step S4 To perform the correspondence with the next standard pattern.

With respect to the input pattern in Fig. 3, since the correspondence with all the standard patterns has been completed, the process proceeds to step S9.

The control means 11 sets the correspondence result obtained in the linear stroke correspondence adding means 5, the nonlinear stroke correspondence adding means 7 and the continuous character stroke correspondence adding means 8 to the character evaluation means 10, Lt; / RTI >

The character evaluation means 10 obtains the evaluation value of the character based on the result of the transferred correspondence portion, and selects the recognition target character (standard pattern) having the smallest distance as the final recognition result.

The evaluation value (D) of the character is obtained by the following expression (3).

D = (Wd x Dd + Ww x Dw + Wh x Dh

+ Wvw x Dvw + Wvh x Dvh + Wvd x Dvd) / k

Here, the distance Dd after normalization in the stroke direction, the distance Dw after normalization of the stroke width, the distance Dh after normalization of the stroke height, the distance Dvw after the normalization of the virtual stroke width, Wv is the weight of the stroke width, Wv is the weight of the stroke width, Wvh is the weight of the stroke width, Wh is the weight of the stroke width, Wv is the weight of the stroke width, Wvh is the weight of the virtual stroke width, Wvh is the virtual stroke height Wvd represents the weight of the virtual stroke direction, and k represents the number of strokes of the current standard pattern. Also,

Wd + Ww + Wh + Wvw + Wvh + Wvd = 1

.

However, when there are a plurality of corresponding addition results of the linear strokes, the non-linear strokes, and the continuous character strokes, it is determined that there is a non-corresponding standard pattern stroke without any corresponding portion in any of the input patterns, So that the contradiction of the strokes of the plurality of input patterns corresponding to the strokes of the plurality of input patterns does not occur.

Next, the partial stroke extracting means 9 will be described. 25 is an example of an input pattern including consecutive character strokes written with four strokes of the character "word" of seven strokes.

Hereinafter, the operation of the partial stroke extracting means 9 will be described taking the input pattern of Fig. 25 as an example. In this example, it is assumed that the strokes 141, 142, and 143 correspond to each other by the linear stroke corresponding means 5 and the non-linear stroke corresponding means 7, and a description thereof will be omitted.

Since the stroke 140 is a continuous character stroke, the process advances to step S7 in Fig. 2, and the corresponding portion is added to the continuous character stroke correspondence adding means 8. Fig.

17, the continuous character stroke correspondence adding means 8 determines the correspondence addition region of the input character string to the stroke correspondence addition possible region determining means 6, A stroke candidate of a standard pattern corresponding to the stroke candidate is obtained.

The stroke correspondence attachable area determining means 6 uses the area segment information shown in Fig. 18 to determine a correspondence attachable area. The correspondence attachable area is intended to extend to one area outside the area including the external connection rectangle of the stroke of the input pattern. Fig. 26 shows an example in which the area division shown in Fig. 18 is performed for the input pattern in Fig. It can be seen in Figure 26 that the outer connecting rectangle 144 of the stroke 140 includes the areas A through L. [ As a result, the stroke-corresponding area of the stroke 140 becomes all the areas A to P.

The continuous character stroke correspondence adding means 8 obtains a stroke of a standard pattern present in the determined correspondence addition enabling area. 27 is a diagram showing the relationship between the input pattern and the input pattern in the standard pattern of the character " word " The strokes of the standard patterns corresponding to each other in a one-to-one correspondence are excluded because the corresponding portions are determined. This example is excluded because the strokes 154 and 156 are determined by the linear stroke corresponding means 5 and the stroke 155 is determined by the non-linear stroke corresponding means 7. As a result, four of the strokes 50, 151, 152, and 153 are selected as the candidate strokes.

Next, the continuous character stroke correspondence adding means 8 proceeds to step S57 in FIG. 17 to check whether or not the correspondence addition possible area has been determined for all continuous character strokes. If the determination result is YES, If not, the flow advances to step S50. Since there is only one continuous character stroke in the input pattern of Fig. 25, the flow advances to step S51 to determine the candidate strokes of the standard pattern which are the start strokes and the end strokes of the continuous character strokes.

26, since the start point of the stroke 140 of the input pattern exists in the region B, the corresponding addable region becomes the region A, B, C, E, F, G, Strokes 150, 151, and 152 of the standard pattern become candidates for the start stroke. Likewise, the strokes 152 and 153 are candidates for the end stroke.

Proceeding to step S52, the continuous character stroke correspondence adding means 8 checks the number of strokes of the standard pattern. In this example, the number of candidate strokes is four, and thus the process proceeds to step S53.

In step S53, the continuous character stroke correspondence adding means 8 extracts the partial strokes from the continuous character strokes of the input pattern by the partial stroke extracting means 9. [

Fig. 28 is a flowchart showing the flow of the process of adding the partial strokes in step S53.

The operation of the partial stroke extracting means 9 will be described with reference to the flowchart shown in Fig.

In step S70, the partial stroke extracting means 9 extracts the inflection point of the stroke of the input pattern as a partial stroke dividing candidate point, and sets the starting point of the partial stroke as the starting point of the stroke of the input pattern.

FIG. 29 is a diagram showing candidate points of division of consecutive character strokes of input patterns. FIG. 29, reference numerals 160 to 165 denote partial stroke division candidate points of the stroke 140 of the input pattern, 166 denotes the start point of the partial stroke, and 167 denotes the end point of the partial stroke.

The process proceeds to step S71 so that the partial stroke extracting means 9 follows the dividing point backward at the end point of the stroke of the input pattern and determines that the number of candidate strokes of the stroke from the starting point of the current partial stroke to the dividing point is 3 or less , And the dividing point is determined as the end point of the partial stroke. The partial strokes from the starting point 166 to the division candidate points 160, 161, 162, 163, 164 and 165 are selected in the example of FIG. 29, The number of candidate strokes is four. Therefore, the number of candidate strokes is four (4), and therefore, the number of candidate strokes is four Go ahead. Likewise, the sequential dividing point is moved in the starting direction, a partial stroke is obtained, and the candidate stroke is extracted. In this example, since the number of candidate strokes of the partial strokes from the starting point 166 to the division candidate points 160, 161, and 162 becomes three, this stroke is selected as the partial stroke.

The process proceeds to step S72, and the continuous character stroke correspondence adding unit 8 performs the correspondence between the obtained partial strokes and the candidate strokes.

31 is a flowchart showing the flow of the process of step S72. The processing of step S72 will be described with reference to the flowchart of Fig.

In step S80, candidates capable of corresponding to each other are compressed by the external connection rectangle of the partial stroke and the external connection rectangle of each candidate stroke. More specifically, a candidate stroke including an external connection rectangle is selected in an area extending from the external connection rectangle of the partial stroke to a predetermined limit value. 180 in Fig. 30 is an externally connected rectangle of an extended partial stroke. In this example, the stroke 152 of the candidate strokes 150, 151, 152 is excluded from the candidate.

The process proceeds to step S81, and the continuous character stroke correspondence adding unit 8 determines the order of the candidate strokes of the standard pattern as in the case where the number of candidate strokes is three or less.

Here, the determination of the combination of the paylines is also a candidate of the start stroke if the current partial stroke includes the start point of the consecutive character strokes of the input pattern, or information that is the candidate of the end stroke when the end portion includes the end point. There are two candidate strokes of the current partial stroke, and since both strokes satisfy the condition of the start stroke, two types of strokes are selected.

Here, the strokes 150 and 151 are first subjected to the corresponding strokes in the order of stroke.

The flow proceeds to step S82, in which the continuous character stroke correspondence adding means 8 performs the correspondence between the partial strokes and the candidate strokes of the standard pattern. The method of counterparting is the same as that of step S61 in Fig. 20 (the same as in the case of the corresponding portion of the continuous character stroke). In the step S61 in Fig. 20, since the strokes 150 and 151 are common linear strokes, each starting point end point becomes a minutiae point, and a portion corresponding to the minutiae point of the partial stroke is obtained by the dynamic programming method.

The flow advances to step S83, and the continuous character stroke correspondence adding unit 8 divides the partial stroke from the obtained result of the corresponding addition, and obtains the stroke characteristic of the divided stroke by the stroke characteristic extracting unit 2. Fig. 32 is an example in which the dividing strokes 190 and 191 are obtained from the partial strokes of Fig.

The process proceeds to step S84, and the continuous character stroke correspondence adding unit 8 checks whether or not the correspondence addition with the partial strokes of the input pattern concerning all the writing order is performed for the current candidate strokes. As a result of the check, if all the combinations are not performed, the flow advances to step S81 to perform the correspondence with respect to the different writing order.

In this example, the flow advances to step S81 to perform a correspondence in another stroke order such as the strokes 151 and 150. Since the correspondence addition process is the same as that in the case of the strokes 151 and 150, a description thereof will be omitted. In this case, the case where the strokes 150 and 151 are pincushions is adopted as a result of the correspondence between the stroke characteristic points of the input pattern and the strokes of the standard pattern.

Next, the continuous character stroke correspondence adding means 8 proceeds to step S73 in Fig. 28 and judges whether or not the correspondence of the current continuous character strokes is completed. If it is judged that the correspondence of the current continuous character strokes has ended, Proceed to S74.

Since it has not been completed in the example of FIG. 29, the process proceeds to step S74. In step S74, the end point of the current partial stroke is set as the start point of the next partial stroke. Here, the division candidate point 162 in Fig. 29 becomes the start point of the next partial stroke.

The process proceeds to step S71, and the continuous character stroke correspondence adding means 8 uses the partial stroke extracting means 9 to create the next partial stroke. Specifically, the partial stroke extracting means 9 refers to the new partial stroke in the reverse direction from the end point of the partial stroke, and determines whether the number of candidate strokes of the stroke from the starting point of the current partial stroke to the dividing point is 3 or less , And the dividing point is defined as the end point of the partial stroke.

In the example of FIG. 29, since the already determined strokes 150 and 151 are excluded from the candidates, the candidates are any of the strokes 152 and 153. Therefore, all the remaining partial strokes constituted by the division candidate points 162, 163, 164, and 165 and the end point 167 of the partial stroke are selected.

The process proceeds to step S72, and the continuous character stroke correspondence adding unit 8 performs the correspondence between the obtained partial strokes and the candidate strokes.

The operation of the corresponding addition of partial strokes will be described with reference to the flowchart of Fig.

In step S80, candidates capable of corresponding to each other are compressed by the external connection rectangle of the partial stroke and the external connection rectangle of each candidate stroke. In this example, both of the candidate strokes 152 and 153 are the objects of the corresponding part satisfying the condition.

Next, proceeding to step S81, the continuous character stroke correspondence adding means 8 determines the order of writing. Since there are two candidate strokes of the present partial stroke, and both strokes satisfy the condition of the ending stroke, there are two order of strokes.

First, the strokes 152 and 153 are assigned in the same order as the strokes.

The flow advances to step S82, and the continuous character stroke correspondence adding unit 8 performs the correspondence between the partial stroke and the candidate stroke of the standard pattern. 20, and since the strokes 52 and 153 are linear strokes at the same time, the respective turning points become the characteristic points, and the corresponding portions of the partial strokes are matched by the dynamic programming method.

Here, since the starting point 162 of the current partial stroke coincides with the end point of the partial stroke up to that point, there may actually be an untwisted virtual stroke component between the stroke and the stroke during that time. Therefore, both the corresponding portion with the division candidate point 162 as the starting point and the corresponding portion with the next division candidate point 163 as the starting point are performed, and the smaller the rectangle is determined as the final result. In this example, the corresponding part with the division candidate point 163 as the starting point is selected.

The process proceeds to step S83, in which the continuous character stroke corresponding section 8 divides the partial stroke from the obtained result of the corresponding part and obtains the stroke characteristic from the stroke characteristic extracting section 2. 33 shows an example in which the dividing strokes 192 and 193 are obtained from the partial strokes.

The process proceeds to step S84 and the continuous character stroke correspondence adding unit 8 checks whether or not the correspondence addition with the partial strokes of the input pattern is performed with respect to the current candidate strokes in all the writing order. As a result of the check, if all the combinations are not performed, the flow advances to step S81 to perform the correspondence with respect to the different writing order.

In the above example, the process proceeds to step S81, and the corresponding parts are added in the same order as the strokes 153 and 152. Corresponding addition processing is the same as in the case of the strokes 152 and 153, and a description thereof will be omitted. In this case, since the distances between the stroke minutiae points of the input patterns and the stroke minutiae points of the standard pattern are small in the case of the strokes 152 and 153, the corresponding portion of the right partial strokes is adopted as a result.

Next, the process proceeds to step S73 in Fig. 28, and the continuation character stroke correspondence adding unit 8 determines whether or not the correspondence of the current continuous character stroke is completed. If it is finished, the step S75, Proceed to S74. Since the process has been completed in the above example, the process proceeds to step S75.

In step S75, the continuous character stroke correspondence adding means 8 obtains the result of the corresponding addition of the continuous character strokes from the result of the corresponding addition of each partial stroke. In this case, since two strokes of the standard pattern correspond to each of the partial strokes, the arrangement in which the two strokes are arranged in the order of the partial strokes is adopted as the final continuous character stroke.

In the case of this example, there is no corresponding stroke in the stroke of the standard pattern, but when there is a stroke not corresponding to the stroke, the portion corresponding to the stroke connected with the continuous partial stroke is referred to as a stroke not corresponding to the stroke , And when the distance of the corresponding portion becomes small, the stroke is adopted. At this time, the number of strokes of the standard pattern inserted between the partial strokes is only one.

With the above processing, the partial stroke is extracted from the continuous character stroke, the stroke corresponding portion is ended, and the character evaluation means 10 calculates the final character evaluation value, and the recognition result is determined.

In the present embodiment, the coordinate information of the corresponding feature point of the continuous character stroke is used, but the stroke may be sampled at regular intervals and the corresponding addition may be performed using the direction code between the sample points.

In the present embodiment, in order to determine an area capable of adding strokes, an area obtained by dividing a character into 4 × 1, 1 × 4, and 4 × 4 (each representing a divided area in the horizontal direction × vertical direction) , And the number of divisions may be any number.

In the present embodiment, the division candidate points of the continuous character strokes are taken as the bending points, but it is also possible to use points sampled at regular intervals.

As described above, according to the present embodiment, with respect to the strokes that are not yet written yet, only the strok characteristic points are subjected to the corresponding addition, and the evaluation is performed using only the stroke characteristic finally, , It is possible to realize a recognition method which is resistant to deformation of characters.

In addition, not only the strokes that have not been continuously written but also the written strokes that are subsequently written correspond to the strokes in the order in which they are written. Therefore, the pattern in which both the strokes and the strokes are varied can be recognized with high accuracy.

In addition, since the extremely long strokes are divided into partial strokes so as to correspond to each other, it is possible not to increase the processing time but to add the strokes in consideration of the strokes.

In addition, input strokes that are handwritten characters are classified into a linear stroke, a non-linear stroke, or a continuous character stroke, and correspondence is performed in the order of a portion corresponding to a straight stroke, a portion corresponding to a nonlinear stroke, It is possible to efficiently carry out the correspondence addition processing without performing the correspondence addition between the correct answer and the other standard pattern, Recognition is possible.

In addition, since the correspondence of the continuous character strokes is made at the feature points of the start and end points of the stroke, it is possible to cope with high speed and to suppress the capacity of the stroke feature dictionary to a small degree.

Since the present invention is configured as described above, the following effects are exhibited.

In the first or second invention, strokes are classified into a linear stroke, a non-linear stroke, and a continuous character stroke from the shape of a handwritten character (input pattern) stroke written on the tablet, and each stroke characteristic is classified into a standard pattern Characters), it is possible to perform character recognition at high speed and with high accuracy with respect to input patterns whose stroke and stroke counts have changed.

The first or third invention is characterized in that the corresponding portion of the stroke of the standard pattern of each stroke of the input pattern is assigned to the standard stroke corresponding to the linear stroke, Do.

Claims (3)

An input means for inputting handwritten characters on-line, A stroke feature dictionary in which the characteristics of straight stroke and non-straight stroke constituting each recognition target character are stored in advance for a plurality of recognition target characters, Stroke feature extraction means for extracting a feature of a stroke constituting the handwritten character from the handwritten character inputted by the input means, Wherein each stroke of the handwritten character is divided into a continuous character stroke which can not be matched with the stroke constituting the linear stroke, the nonlinear stroke, or the recognition target character, based on the characteristic extracted by the stroke characteristic extraction means A stroke classifying means for classifying the stroke into any one of the stroke classifying means, Linear stroke correspondence adding means for performing correspondence between the characteristics of the straight stroke classified by the stroke classification means and the characteristics of the straight stroke constituting the recognition target character stored in the stroke characteristic dictionary; Linear stroke correspondence adding means for performing correspondence between characteristics of nonlinear strokes classified by said stroke classifying means and characteristics of nonlinear strokes constituting recognition target characters stored in said stroke characteristic dictionary; Wherein the continuous character stroke classified by the stroke classifying means is divided into the linear stroke or the non-linear stroke, and the characteristic of the divided linear stroke or non-linear stroke and the character to be recognized stored in the stroke characteristic dictionary A continuous character stroke correspondence adding means for performing correspondence between a feature of a straight line stroke and a nonlinear stroke, And character evaluation means for recognizing the handwritten character based on the result of the corresponding portion obtained by the linear stroke corresponding means, the nonlinear stroke corresponding means or the continuous character stroke corresponding means, Wherein said continuous character stroke correspondence adding means performs a correspondence addition to said character to be recognized corresponding thereto by said linear stroke corresponding means or said non-linear stroke corresponding means. A character recognition method provided with a stroke feature dictionary in which features of straight stroke and non-straight stroke constituting each recognition target character for a plurality of recognition target characters are stored in advance, An input step of inputting a handwritten character online, A stroke feature extraction step of extracting a feature of a stroke constituting the handwritten character from the handwritten character input by the input step; Wherein each of the strokes of the handwritten character is divided into a plurality of stroke strokes in a continuous character stroke in which the corresponding strokes constituting the rectilinear stroke, A stroke classification step of classifying the stroke classification into any one, A linear stroke corresponding step of performing a correspondence between the characteristics of the linear stroke classified by the stroke classification step and the characteristics of the linear stroke constituting the recognition target character stored in the stroke characteristic dictionary; A nonlinear stroke correspondence adding step of performing a correspondence between the characteristics of the nonlinear strokes classified by the stroke classification step and the characteristics of the nonlinear strokes constituting the recognition target character stored in the stroke characteristic dictionary; Wherein the stroke characterizing step divides the continuous character stroke classified by the stroke classification step into the linear stroke or the non-linear stroke, and forms the recognition target character stored in the stroke characteristic dictionary and the character of the divided linear stroke or non- A continuous character stroke correspondence addition step of performing correspondence between the feature of the straight line stroke and the nonlinear stroke, And a character evaluation step of recognizing the handwritten character based on the result of the addition corresponding to the linear stroke, the nonlinear stroke corresponding addition step or the result of the corresponding addition obtained by the continuous character stroke corresponding addition step Character recognition method. 3. The method of claim 2, Wherein the step of adding the continuous character strokes corresponds to the step of performing the correspondence addition to the character to be recognized corresponding to by the step of adding the linear stroke or the step of adding to the non-linear stroke.