JPH1011537A - Character segmenting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character segmenting device which can reduce its processing amto and segments the characters touching with each other with the separate characters left as they are. SOLUTION: A coordinate input part 1 detects the stroke points at a fixed time interval based on the two-dimensional coordinate data against the character strings which are inputted in sequence. A basic segment division part 2 divides a stroke string included in the coordinate data into plural basic segments. When the number of strokes of a basic segment is larger than a prescribed number or the vertical-horizontal ratio of the basic segment is larger than a certain level, a basic segment compulsive division part 3 decides this basic segment as a division candidate. If the angle of vector set between the adjacent strokes is defined in a range of 20 to 80 degrees and also the vector length is lager than the diagonal length of a circumscribed rectangular of the basic segment by 0.43 times or more, these adjacent strokes are divided for generation of a new basic segment.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognizing apparatus for recognizing a handwritten character string, and more particularly to a character recognizing apparatus for performing character pre-processing as a preprocessing for character recognition.

[0002]

2. Description of the Related Art In order to realize a good man-machine interface by inputting handwritten characters, there is a demand for a method of recognizing a character string written on a blank sheet of paper without a character frame which does not burden a writer. For example, as such, in online handwritten character recognition using the temporal change of the movement of the pen, in order to cope with a character frame-free handwritten character, an online handwritten character described in Japanese Patent Application Laid-Open No. There was a character recognition system.

In this recognition method, a character string is input as a coordinate string from a coordinate input device such as a data tablet, and a handwritten character string is recognized using a candidate lattice method based on coordinate information of strokes constituting a character. . That is, in this method, a stroke string is divided into a plurality of basic segments, and the separated basic segments are combined to create a candidate character. Then, those candidate characters are compared with a standard character group, and an optimum character string is obtained and cut out.

[0004]

An important point in the above-described character string cutout processing is to accurately separate basic segments. If the basic segments are erroneously separated, the processing amount increases as character recognition processing, and the recognition rate decreases.

[0005] As a method of dividing the basic segment, in the case of horizontal writing Japanese, there is a method of projecting a stroke sequence on the horizontal axis and dividing the stroke sequence when the interval is larger than a certain threshold Th. In this method, the threshold is set to a negative value, for example, T
When h = −0.3, as shown in FIG. 2 to be described later, it is possible to divide even a portion where a shadow overlaps, but it also affects other non-contact portions, and the basic segment is affected. Will increase.

FIGS. 2A and 2B are explanatory diagrams of such a conventional example of dividing a basic segment. FIG. 2A shows an input character string, and FIG.
Is the case where the threshold value Th = 0, and (c) is the case where the threshold value Th = −0.3.

[0007] As shown in (a), in the input character string "Katsuno-shi, Osaka", "Sakafu" and "Katsuno" have a portion that overlaps the other character in the horizontal direction. Therefore, when Th = 0, as shown in (b), when the stroke row is projected on the horizontal axis, since there is no gap, the stroke row is divided as one basic segment. On the other hand, Th =
In the case of −0.3, as shown in (c), for example, “fu” of “Hanfu” is divided into two basic segments. In the candidate lattice method for recognizing characters from combinations of basic segments, if the number of basic segments increases, the number of combinations increases and the processing amount increases.
The number of basic segments should be as small as possible. Therefore, when the number of basic segments increases as in (c), the processing amount increases, which hinders speeding up.

[0008] In view of the above, a character cutout apparatus capable of effectively generating a basic segment by cutting out a contacting character while leaving a small amount of processing and a separated character as it is. Was desired.

[0009]

The present invention employs the following structure to solve the above-mentioned problems. <Structure of Claim 1> A basic segment dividing unit that divides an input character string into basic segments consisting of a stroke sequence that is a two-dimensional coordinate data sequence, and a basic segment divided by the basic segment dividing unit is a predetermined segment. If the number of strokes is larger than the number of strokes, the angle and length of the vector generated in the adjacent strokes are within a predetermined angle range and a predetermined ratio or more to the length corresponding to the size of the basic segment And a forced segmentation unit that divides an adjacent stroke satisfying the condition and generates a new basic segment.

<Explanation of Claim 1> The character cutout device according to the first aspect of the present invention is mainly a character cutout device for online handwritten characters that processes a handwritten character string that is sequentially written and input. Applicable to both vertical writing.

The basic segment dividing unit ideally divides the input character string into basic segments in units of characters, but two characters may be recognized as one basic segment. That is, the basic segment dividing section may have any configuration as long as it divides a character string into arbitrary basic segments.

The predetermined number of pictures determined by the basic segment forced division section is, for example, a value such as 25 pictures or more, but this can be set appropriately according to the use state. The vector generated by the adjacent stroke is, for example, a vector from the midpoint of the target stroke to the start point of the next stroke after the target stroke, but is not limited thereto, and (end point, start point), A vector such as (middle point, middle point) may be used. The range of the preset angle is
For example, the angle is in the range of 20 ° to 80 °, but this can also be appropriately selected. The length corresponding to the size of the basic segment is, for example, the length of a diagonal line of a circumscribed rectangle of the basic segment, but may be another value as long as the length corresponds to the size.

The basic segment compulsory division unit checks the inter-stroke vector of each basic segment output from the basic segment division unit based on the above-described angle and length conditions. As a result, it is possible to effectively divide only the basic segment composed of the touching characters and having a large number of strokes.

<Structure of Claim 2> The input character string is
A basic segment dividing unit that divides the basic segment into a basic segment composed of a stroke sequence that is a two-dimensional coordinate data sequence, and the length of the basic segment divided by the basic segment dividing unit in the character string direction is set in the vertical direction in the character string direction. If the length is greater than the preset ratio by more than the length, the angle and length of the vector generated in the adjacent stroke are within the range of the preset angle, and the length corresponding to the size of the basic segment. It is determined whether the ratio is equal to or greater than a predetermined ratio,
A character segmentation device comprising: a basic segment forced division unit that divides an adjacent stroke satisfying this condition and generates a new basic segment.

<Explanation of Claim 2> In the invention of claim 1 described above, the condition that the basic segment forced division unit is a division candidate is equal to or greater than a predetermined number of strokes. For example, it is an aspect ratio for a horizontally written character string. Here, the character string direction is the horizontal direction in the case of a horizontally written character string, and the vertical length in the character string direction means the height of the character string. It is equally applicable. The subsequent operation of the basic segment forced division unit is the same as that of the first aspect. With such a configuration, it is possible to effectively divide only the basic segment having a large aspect ratio and composed of the touching characters.

<Structure of Claim 3> In the character segmenting device according to Claim 1 or 2, when there are a plurality of vectors satisfying a preset condition, the character segment is set at an intermediate point of the distance in the character string direction in the basic segment. A character segmentation device comprising a basic segment forced division unit that uses a close vector as a division candidate vector.

<Explanation of Claim 3> The midpoint of the distance in the character string direction in the basic segment is, for example, in the case of a horizontally written character string, the midpoint of the horizontal distance of the basic segment. That is, this is a value obtained from the viewpoint that when two characters are included as the basic segment, there is a high possibility that the character is divided at the intermediate portion. With such a configuration, even when a plurality of candidate vectors for division exist, the division vector can be accurately and easily determined.

<Structure of Claim 4> A coordinate input unit for extracting a stroke sequence which is a two-dimensional coordinate data sequence from an input character string, and a plurality of strokes extracted by the coordinate input unit are adjacent to each other. The angle and length of the vector generated by the stroke are within a preset angle range, and
Determining whether the length of the character string in a direction perpendicular to the character string direction is equal to or greater than a predetermined value, and if the length is satisfied, an adjacent vector dividing unit that divides an adjacent stroke to generate a basic segment And a character segmenting device.

<Explanation of Claim 4> In the invention of claim 4, the invention of claims 1 to 3 described above extracts division candidates based on the number of strokes and the aspect ratio of the basic segment. In this configuration, the division candidates are extracted only by the feature amount of the length. For example, that the vector is equal to or more than a predetermined value with respect to the length in the direction perpendicular to the character string direction indicates that there is a length of 80% or more of the height of the character string in the case of a horizontal writing character string. However, this value can be changed as appropriate.

Therefore, regardless of the contactless or non-contact character string, the basic segment generated by the adjacent vector division unit is basically divided into partial and adjacent kanji characters. Has good affinity with the recognition process of
Character strings can be cut out effectively.

[0021]

Embodiments of the present invention will be described below in detail with reference to the drawings. << Specific Example 1 >><Structure> FIG. 1 is a structural diagram showing a specific example 1 of the character segmenting apparatus of the present invention. The device shown in the figure comprises a coordinate input unit 1, a basic segment division unit 2, a basic segment forced division unit 3, a candidate character generation device 4, a candidate character recognition device 5, and an optimum character string selection device 6.

The coordinate input unit 1 is composed of, for example, a tablet, and has a function of detecting, at a constant time interval, each writing point with two-dimensional coordinate data for a character string that is sequentially written and input. The basic segment dividing unit 2 has a function of dividing a stroke sequence into a plurality of basic segments from coordinate sequence data.

The basic segment forcible division unit 3 has a function of forcibly dividing contact characters that have not been divided by the basic segment division unit 2. That is, the basic segment compulsory division unit 3 determines whether the basic segment divided by the basic segment division unit 2 is larger than a predetermined number of strokes, or the length of the basic segment in the character string direction is vertical. If the length is greater than or equal to a predetermined ratio than the length in the direction, the angle and length of the vector generated in the adjacent stroke are within the predetermined angle and the length corresponding to the size of the basic segment. It has a function of determining whether the ratio is equal to or more than a predetermined ratio, dividing an adjacent stroke satisfying this condition, and generating a new basic segment.

The candidate character generation device 4 has a function of combining each basic segment sent from the basic segment forced division unit 3 to generate a candidate character. The candidate character recognition device 5 has a function of calculating the degree of difference between the candidate character and the standard character group, and detecting the name of the standard character with the minimum degree of difference and the degree of difference. The optimal character string selection device 6
It has a function of assigning a series of character names that minimizes the sum of the differences to the input stroke sequence. The basic segment dividing unit 2 to the optimum character string selecting device 6 are configured by a microcomputer having programs for realizing the respective functions and a CPU for executing the programs.

<Operation> The operation of this embodiment will be described below assuming that an online handwritten character string corresponding to Japanese in horizontal writing is cut out.

The position information of the N strokes str1 to strn input from the coordinate input unit 1 such as a tablet is input to the basic segment dividing unit 2. For a certain value K (1 ≦ K ≦ N−1), the basic segment dividing unit 2 calculates the maximum value X1 of the X coordinate from the first stroke to the Kth stroke and the X value from the K + 1st stroke to the Nth stroke. The minimum value X2 of the coordinates is detected, and (X
Only when 2-X1) is positive, the stroke train is divided between the K-th stroke and the K + 1-th stroke.
This operation is sequentially performed for K from 1 to N-1, and the input stroke sequence is divided into basic segments. That is, the basic segment dividing unit 2 calculates the distance of the gap between each stroke when the stroke sequence is projected on one coordinate axis,
When the distance of the gap does not exist, the strokes on both sides are combined. When the distance of the gap exists, the strokes on both sides are divided, and the input stroke sequence is divided into basic segments. In this division, if a gap is created in the writing direction in the opposite direction, it is regarded as one basic segment.

FIG. 3 is a diagram for explaining a gap generated in a direction opposite to the writing direction. As shown, the letter "small"
Has a gap in the opposite direction between the strokes of the stroke orders 1 and 2. In this case, the strokes of the stroke orders 1 and 2 are the same basic segment.

The processing result by the basic segment dividing unit 2 is, for example, as shown in FIG.
become that way. Therefore, in this specific example, the basic segment including the touching character, which has not been divided by the basic segment division unit 2, is divided by the basic segment forced division unit 3.

FIG. 4 is a flowchart of the basic segment forced division process. The basic segment forced division unit 3 sequentially processes the processing result of the basic segment division unit 2 from the first basic segment.

First, the number of strokes of the basic segment of interest is 25 or more, or the aspect ratio of the basic segment is 2
Is checked (step S1). If this condition is not satisfied, no processing is performed on the basic segment. On the other hand, if any of the conditions is satisfied, it is determined whether there is a divided vector that satisfies both the following (Equation 1) and (Equation 2) (Step S2). still,
The condition for the division candidate is that the number of strokes is 25 strokes or more or the aspect ratio exceeds 2, because in the case of ordinary Japanese kanji, it is considered that there is hardly any value exceeding these values. These values can be appropriately changed according to the use conditions and the like.

(Equation 1) 20 ° <kakudo <80 ° (Equation 2) d12 / Line> 0.43

FIG. 5 is an explanatory diagram of the feature amount used for division. FIG. 6 is an explanatory diagram of the calculation of the feature amount used for the division.
The example of FIG. 5 shows a case where the character string “Hanfu” is output as one basic segment as an example of the processing result in the basic segment dividing unit 2. Here, d12 corresponds to FIG.
As shown in the calculation formula in (a), it is the length from the midpoint of the stroke of interest to the start of the next stroke. That is, in FIG. 5, the distance from the middle point of str1 (which is the stroke to be focused) to the start point of str2 (the stroke next to the focused stroke) is shown.

Kakudo in (Equation 1) is the angle between d12 and the X axis, str2_sx and str1_mx (the X coordinate of the starting point of the next stroke, and the X coordinate of the midpoint of the stroke to be focused) and , Str2_sy and str1_my
Based on the relationship between (the Y coordinate of the start point of the next stroke and the Y coordinate of the middle point of the focused stroke), it is determined as shown in the arithmetic expression in FIG. Furthermore, of (Equation 2)
Line is the length of the diagonal line of the circumscribed rectangle of the group output as the basic segment, and the length is obtained by the arithmetic expression shown in FIG.

Returning to FIG. 4, in the above step S2,
If there is no division vector that satisfies both (Equation 1) and (Equation 2), no processing is performed on the basic segment, and if there is, the process proceeds to the next step S3 for performing forced division. That is, (Equation 1) in step S2,
The case where (Equation 2) is satisfied means that d12 is satisfied as in the example of FIG.
Is longer than a certain value with respect to the length of the diagonal line of the circumscribed rectangle,
And when the angle is between 20 ° and 80 °, it means that those strokes are likely to be another elementary segment.

In step S3, it is checked whether the number of divided vectors is only one. If there is only one, forced division is immediately performed (step S5). Select (Step S4). In the selection method in step S4, a divided vector having the smallest value in Expression 3 below is set as a candidate. (Equation 3) BX = | box_xm- (str1_mx + s
tr2_sx) / 2 |

In FIGS. 5 and 6, str1_sx
Is the X coordinate of the start point of the stroke of interest, str1_s
y is the Y coordinate of the start point of the stroke of interest, str1_
ex is the X coordinate of the end point of the stroke of interest, str1
_Ey is the Y coordinate of the end point of the focused stroke.
The X coordinate of the midpoint of the stroke of interest is str1
_Mx = (str1_sx + str1_ex) / 2, and the Y coordinate of the midpoint of the stroke of interest is str1_
my = (str1_sy + str1_ey) / 2. Further, box_xmin is the minimum X coordinate value of the basic segment, box_xmax is the maximum X coordinate value of the basic segment, box_ymin is the minimum Y coordinate value of the basic segment, and box_ymax is the maximum Y coordinate value of the basic segment.

FIG. 7 is an explanatory diagram when two candidate vectors exist. As shown in the drawing, the absolute value of the difference between the X coordinate of the midpoint of the candidate vector 1 and the X coordinate of the midpoint of the basic segment is BX1, and the X coordinate of the midpoint of the candidate vector 2 and the X coordinate of the midpoint of the basic segment are set. When the absolute value of the difference between BX1 and BX2 is BX2, the smaller of the values of BX1 and BX2 is set as the divided vector. Therefore, in the case of the example shown in FIG.
Becomes a division vector. That is, such a selection is determined from the viewpoint that a vector closer to the midpoint of the basic segment is more likely to be a vector to be divided.

FIG. 8 is an explanatory diagram showing the processing result of the basic segment forced division unit 3. As shown in the figure, by performing the above-described forced division processing, the basic segment division unit 2
As a result of the processing at, “Hanfu” and “Katsuno”, which were regarded as one basic segment, are also cut out as basic segments.

The operation after the candidate character generation device 4 for the basic segment divided by the basic segment forced division unit 3 is as follows. The candidate character generation device 4 generates a candidate character, and the candidate character recognition device 5 outputs the candidate character. Calculate the degree of difference between the character string and the standard character group.
A sequence of character names that minimizes the degree of difference is assigned. However, since these operations are known operations, a detailed description thereof will be omitted.

<Effect> As described above, in the first embodiment, since the basic segment forced division unit 3 is provided, the basic segment having a large number of strokes composed of touching characters,
Only the basic segment having a large aspect ratio can be effectively divided. As a result, it is possible to contribute to an improvement in the processing speed of the character cutout device.

<< Specific Example 2 >> In the above specific example 1, it is estimated that the basic segment is composed of two characters according to the horizontal length of the basic segment generated by blanks or the number of strokes. Does not use such a horizontal length or the number of strokes, but extracts a division candidate only with the feature amount of the angle and length of the inter-stroke vector.

<Structure> FIG. 9 is a view showing the structure of a character segmenting apparatus according to the second embodiment. This specific example 2 is the above specific example 1.
The difference is that an adjacent vector dividing unit 7 is provided instead of the configuration of the basic segment dividing unit 2 and the basic segment forced dividing unit 3. The other components are the same as those of the first embodiment, and thus the same portions are denoted by the same reference numerals and description thereof will be omitted.

The adjacent vector division unit 7 determines that the vector composed of adjacent strokes is within a predetermined angle and that the length of the vector is greater than or equal to an arbitrary percentage of the height of the input character string. Has a function of dividing adjacent strokes.

<Operation> The position information of N stroke sequences str1 to strn input from the coordinate input unit 1 such as a tablet is input to the adjacent vector dividing unit 7. In the adjacent vector dividing unit 7, first, the height (Heigh
t).

FIGS. 10A and 10B are diagrams for explaining the generation of a basic segment by using adjacent stroke vectors. FIG. 10A is a diagram for explaining the height of an input character string, FIG.
(C) is an explanatory diagram of a vector satisfied as a division candidate,
(D) is an explanatory diagram of the generated basic segment.
Note that also in this specific example 2, the target character string is “Katano-shi, Osaka” similar to the specific example 1.

Next, the adjacent vector dividing unit 7
As shown in (b), the middle point of the stroke str (k) and the start point of the stroke str (k + 1) (k = 1 to n−
Find the vector connecting 1). If the vector satisfies the following two conditions, the vector is divided before and after the vector in order to avoid erroneous division in the inner and lower parts. That is, it is considered that the relationship between the two strokes that satisfy the condition is the stroke at the end of one character and the stroke at the beginning of the next character, or the stroke at the end of the kanji bias and the stroke at the side beginning. . 20 ゜ <kakudo <80 ゜ d12> 0.8 * height

For example, in the case of (b), first, st
The vector between r1 and str2 does not satisfy the condition in both the angle and the length. The vector between str2 and str3 satisfies the angle but does not satisfy the length condition. On the other hand, in the vector between str3 and str4,
Since both the angle and the length satisfy the conditions, str1 to s
The segment up to tr3 is divided as a basic segment. Hereinafter, similarly, a vector satisfying as a division candidate is obtained {FIG. 10 (c)}, and a basic segment is generated {FIG.
(D)｝.

The basic segment generated in this manner is sent to the candidate character generation device 4, but the subsequent operation is the same as that of the first embodiment, and the description is omitted here.

<Effects> As described above, in the specific example 2, the generated basic segment is basically divided into partial and adjacent kanji characters regardless of the contactless or non-contact character string. It has good affinity with the candidate character recognition device 5 on the side, and it is possible to cut out a character string effectively.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a specific example 1 of a character cutout device of the present invention.

FIG. 2 is an explanatory diagram of a conventional example of dividing a basic segment.

FIG. 3 is a diagram illustrating a case of a gap generated in a direction opposite to a writing direction.

FIG. 4 is a flowchart of a basic segment forced division process in the specific example 1 of the character segmenting apparatus of the present invention.

FIG. 5 is an explanatory diagram of a feature amount used for division in a specific example 1 of the character cutout device of the present invention.

FIG. 6 is an explanatory diagram of a calculation of a feature amount used for division in a specific example 1 of the character cutout device of the present invention.

FIG. 7 is an explanatory diagram in a case where two division candidate vectors exist in the specific example 1 of the character segmenting apparatus of the present invention.

FIG. 8 is an explanatory diagram of a processing result of a basic segment forced division unit in the specific example 1 of the character cutout device of the present invention.

FIG. 9 is a configuration diagram of a specific example 2 of the character cutout device of the present invention.

FIG. 10 is an explanatory diagram of generation of a basic segment by an adjacent stroke vector in the specific example 2 of the character cutout device of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Coordinate input part 2 Basic segment division part 3 Basic segment forced division part 4 Candidate character generation device 5 Candidate character recognition device 6 Optimum character string selection device 7 Neighbor vector division part

Claims (4)

[Claims] 1. A basic segment dividing unit that divides an input character string into basic segments each composed of a stroke sequence, which is a two-dimensional coordinate data sequence, and the basic segment divided by the basic segment dividing unit has a predetermined number of strokes. If the number is larger, the angle and length of the vector generated in the adjacent strokes are within a predetermined angle range and a predetermined ratio or more to the length corresponding to the size of the basic segment. Is determined, and
A character segmentation device comprising: a basic segment forced division unit that divides an adjacent stroke satisfying this condition and generates a new basic segment. 2. A basic segment dividing unit that divides an input character string into basic segments each composed of a stroke sequence that is a two-dimensional coordinate data sequence, and a character string direction of the basic segment divided by the basic segment dividing unit. If the length is greater than or equal to a predetermined ratio than the length in the vertical direction in the character string direction, the angle and length of the vector generated by the adjacent strokes are within the range of the predetermined angle. Also, it is determined whether the length is equal to or more than a predetermined ratio with respect to the length corresponding to the size of the basic segment, and an adjacent stroke satisfying this condition is divided to generate a new basic segment. And a character segmenting device. 3. The character segmenting device according to claim 1, wherein when there are a plurality of vectors satisfying a preset condition, a vector close to an intermediate point of a distance in a character string direction in the basic segment is divided into candidate vectors. A character segmenting device, comprising a basic segment compulsory division unit that sets the vector as a vector. 4. A coordinate input unit for extracting a stroke sequence, which is a two-dimensional coordinate data sequence, from an input character string, and a plurality of strokes extracted by the coordinate input unit are generated by adjacent strokes. It is determined whether the angle and length of the vector are within a range of a preset angle and are equal to or greater than a predetermined value with respect to the length of the character string in a direction perpendicular to the character string direction. A character segmentation device comprising: an adjacent vector division unit that divides an adjacent stroke to generate a basic segment when the condition is satisfied.