JPH08212301A - Device and method for segmenting character - Google Patents

Abstract

PURPOSE: To precisely segment characters even when a character string of input characters which are not fixed in size or pitch is inputted by judging the kind of the inputted character set and changing segmentation parameters on the basis of the judged kind. CONSTITUTION: A user registers the kind of inputted characters by indicating the icon corresponding to the character kind with a pen. At a segmentation part 7, a character kind judging part 51 judges what character kind the input character string has. At a judging part 72' a criterion change 52 changes the segmentation parameters according to the character kind judged by the character kind judging part 51. A judging processing part 53 judges a character gap and a non-character gap. At an optimum path detection part 73', a segmentation feature evaluation part 54 changes the parameters according to the character kind judged by the character kind judging part 51. A segmentation position judging part 55 determines a segmentation position.

Description

Detailed Description of the Invention

The present invention relates to a clipping technique for clipping a plurality of characters in units. In particular, the present invention relates to a technique for detecting a cutout position (position between characters) for cutting out character data for each character, such as handwritten data input from a tablet or the like, or optical data input by OCR or the like.

[0002]

2. Description of the Related Art An electronic notepad or the like has been developed which recognizes handwritten input handwritten data as characters, converts the handwritten data into character data, and stores the character data. In this electronic memo pad, a plurality of writing frames for one character are provided, and one character is handwritten and input into the writing frame. Then, the character is recognized from the writing data for each character.

Further, in order to realize handwriting input with a higher degree of freedom, it has been considered to continuously write in a region where there is no frame for each character. However, in the latter case, it is necessary to perform a process of cutting out handwritten data continuously written for each character. This cut-out position is naturally between characters. For this reason, it is necessary to automatically detect the space between characters from the written data that is written continuously.
Various techniques for this have been proposed.

That is, individual writing data (stroke)
It is conceivable to judge the character cut-out position based on the interval. It is also conceivable to detect the character cut-out position by the time the pen is away from the writing surface (pen OFF time). It is also possible to determine the correct character cut-out position based on the evaluation value (character recognition reliability) when the cut-out character string is character-recognized.

Further, not only the character recognition reliability but also the evaluation value (language processing reliability) obtained by the language processing of the character recognition result may be used to determine the correct character cut-out position. It is also possible to combine the above determination means. In addition, such a thing is disclosed in Japanese Patent Laid-Open No. 6-1
No. 24364 (G06K9 / 34), JP-A-6-16226.
It is publicly known as shown in Japanese Patent Publication No. 9 (G06K9 / 62) and the like.

As described above, although the conventional technique is well known, an example of the conventional technique will be briefly described with reference to FIGS. 1 to 3. FIG. 1 is a block diagram showing a schematic configuration of a character cutting device. In the figure, 1 is an input unit. The input unit 1 is composed of an electromagnetic induction type tablet. Since the tablet is an electromagnetic induction type, it is possible to know not only the touch of the tablet surface of the electromagnetic induction type pen but also the separated state of the pen. The input unit 1 detects the coordinates (x, y) of the pen and the state of the pen (FAR of the pen, OFF of the pen, ON of the pen) at constant time intervals, and outputs the data. The pen FAR state is a state in which the pen tip is far away from the writing surface, the pen OFF state is a state in which the pen tip is far from the writing surface, and the pen ON state is a state in which the pen tip is in contact with the writing surface. It is in a state of being.

Reference numeral 2 denotes a central processing unit (CPU) for arithmetic processing which controls the entire apparatus. 3 is a ROM. The ROM 3 stores the control program of the CPU 2 and various data. Reference numeral 4 is a memory (RAM). R
The AM 4 is used as a work area for the CPU 2, and has an area for storing handwriting data input from the input unit 1, an area for storing character information as a result of the clipping unit 7, and an area for storing document information. ing.

Reference numeral 7 is a cutout portion. The cutout unit 7 includes a stroke sequence extraction unit 71, a determination unit 72, and an optimum route detection unit 73. The stroke sequence extraction unit 71 obtains a stroke sequence. The determination unit 72 first determines whether it is impossible to determine whether each stroke is between characters or non-characters. The stroke sequence extraction unit 71 uses the data from the input unit 1 to
Based on the pen state of the writing data, the coordinate data is divided and managed for each stroke, and the data necessary for cutting out, such as each stroke interval and pen OFF time, is calculated.

The determination unit 72 determines whether it is impossible to determine whether the strokes are between characters or non-characters. An example of this determination method will be described. The determination method in this conventional example is OF
The F time, the space width, and the FAR time (the time away from the writing surface so that the pen position cannot be detected) are used. Here, the OFF time (the time when the pen separation is about 1 cm or less)
However, if it exceeds the character-to-character fixed OFF time threshold value (about 0.5 sec), it is determined that there is character space, and the state value indicating the character space is set to "1".

Further, the FAR time (the time when the pen is separated by about 1 cm or more) is determined by the inter-character fixed FAR time threshold (about 0.0 s).
If it exceeds ec), it is determined that there is a space between characters and the state value is set to "1". When the space width (X direction) exceeds the inter-character fixed space width threshold (about 3.0 mm), it is determined that the space is between characters and the state value is set to "1".

On the other hand, the OFF time is a non-character fixed OFF
If it is smaller than the time (about 0.1 sec), it is determined that there is a space between characters and the state value is set to "0". In addition, the space width (X direction) is defined as the non-character space width threshold (about −
If it is smaller than 1.0 mm), the state value is set to "0". In other cases, the state value is set to "2" because the determination is impossible.

FIG. 2 shows an example of this writing. That is,
As shown by E1 in FIG. 2, if "Kyoga ga" is written, E3
Is converted into a stroke sequence. Then, as indicated by E4, it is assumed that the character spacing / non-character spacing between strokes is determined. This status value is "1"
Between strokes is always between characters, strokes with a state value of "0" are always between characters, and strokes with a state value of "2" can be either. In this case, when the stroke sequence is cut out finely, it can be divided into basic segments as shown by E5 in FIG. As shown by E6, there are nine stroke sets at this time.

A combination of a plurality of stroke sets (for example, 1-3-4-6-7-9, 1-3-4).
-6-8) can be considered in eight ways. The optimum route detection unit 73 detects the most likely combination of stroke sets. That is, the mutual relation of the stroke sets can be represented by a network as shown in FIG.

In FIG. 3, the nodes of the network are
The same numbers correspond to the nine stroke sets E6 shown in FIG. Further, adjacent nodes (stroke sets) are linked as characters. Then, the weight of the node represents the likelihood of one character, and the weight of the link represents the likelihood of the character. And for each node and link,
The optimum route is obtained by calculating the feature amount representing “one character likeness” and “character likeness” and obtaining the total of the routes. In this case, for example, 2-5-8 and 2-4-
It is assumed that two cases of 6-8 have a high score.

Reference numeral 5 in FIG. 1 is a recognition unit. This recognition unit 5
Is a well-known character recognition processor, performs character recognition on the stroke set input to the recognition unit 5 with reference to the recognition dictionary 10, and returns recognition result candidates (character codes) and their scores. In this embodiment, the stroke set of the combination 2-5-8 and the stroke set of the combination 2-4-6-8 are input and the characters are recognized.

Reference numeral 6 is a language processing unit. The recognition unit 6 is composed of a known language processor, and returns a word (character code string) and its score from a character string obtained by combining the character recognition result candidates obtained by the recognition unit 5. Reference numeral 8 is a character string recognition unit. The character string recognizing unit 8 is the optimum route detecting unit 1.
The score of 5, the score of the recognition unit 5, and the score of the language processing unit 6 are comprehensively determined to determine the correct character string.

Reference numeral 9 is a display section. The display unit 9 is a display, and displays handwriting data and the like input from the input unit 1. In this example, the input unit 1 and the display unit 9 are separately shown, but an integrated type such as a display-integrated tablet may be used. 12 is an internal bus. The internal bus 12 includes a data bus from the CPU 2, an address bus, a control signal bus, and the like.

By the way, when the character is cut out, the standard reference format data (standard character width, standard character aspect ratio, character pitch, character space width) etc. of the input character are known in advance. The determination unit 72 and the optimum route detection unit 73 can easily cut out. That is, when the standard character format data is as shown in FIG. 4A, the clipping is performed according to this format data.

For example, the optimum route detection unit 73 determines that the set of the basic segment is "likely one character" as it is closer to the character size of the reference format data. In addition, it is determined that "a space between characters" is determined as a blank space existing closer to the character pitch of the standard format data. Also, the larger the blank space is than the space width of the character,
It is judged that "it seems to be between characters".

That is, in the case of the input of FIG.
As in (c), there is a high possibility that it will be cut out. In addition, when the standard character format data is as shown in FIG.
Is likely to be cut out as shown in FIG. If the size of the input character is known, there is little possibility of erroneous clipping.

The size of the character differs depending on the character type. However, changing the clipping process depending on the character type has not been considered in the conventional system. Note that it has been conventionally performed to change the character recognition process instead of the cutout process depending on the character type. Further, it is generally known that characters of the same character type in a character string tend to be continuous. This is applied to language processing to selectively output a result in which characters of the same character type continue in character recognition result candidates. However, it is not performed to determine what kind of character string the input character string is composed of and to change the cutout process according to the result.

[0022]

In a character string or the like in which characters of a plurality of character types are mixed, the size of characters and the space width of characters are likely to be uneven, and the cutout accuracy is likely to decrease. In this way, if the size of the input character deviates from the standard size, the clipping process will be incorrect. If characters of different sizes are mixed and input, the cutout process will be erroneous.

It is an object of claims 1 to 5 and 10 of the present application to improve the character segmentation accuracy by determining the character type configuration of the input character string and performing the character segmentation processing according to the determination result. . Further, claims 6 to 9 and 11 of the present application,
An object of the present invention is to improve the character cutting accuracy for character data in which the character size and the character space width are not uniform.

That is, when cutting out irregular character data, a plurality of standard character widths to be a reference are required. However, it is difficult to cut out by providing a plurality of standard character widths. By the way, the standard character pitch is naturally set smaller than the standard character width, and (standard character pitch-standard character width) is the standard character space width.

As described above, the format data of various characters are related to each other, and by this data, character-likeness and
The one-character likeness was evaluated to detect the cutout position. However, this is the case when the sizes of the input characters are uniform. Claims 6 to 9, 11, and 12 of the present application,
For character data with uneven character size and character space width, in order to improve the character cutout accuracy, the related cutout process was performed by the related character format data. The evaluation of 1 character likeness is performed as an individual process to deal with uneven character data.

[0026]

The character cutting device of the present application is
An input means (1) for inputting character data, and a character cutout device for cutting out a character from the character data input by the input means (1), a character type determination means for determining a character type of an input character set ( 51, 51 ') and changing means (52, 54) for changing the cutout parameter based on the character type determined by the character type determining means (51, 51').

The character slicing device according to the present application includes an input means (1) for inputting character data, and a character slicing device for slicing characters from the character data input by the input means (1). Character type determination means (51,51 ')
And a changing means (52, 51) that selectively uses a plurality of cutting methods based on the character type determined by the character type determining means (51, 51 ′).
54) and are provided.

Further, the character cutting device of the present invention is characterized in that the judgment in the character type judging means (51) is made based on the setting of the character type setting mode. Further, the character cutting device of the present application is characterized in that the judgment by the character type judging means (51 ') is made based on the number of strokes or the shape of the character data. Further, in the character cutout device of the present application, the judgment in the character type judgment means is a recognition means for recognizing the cut out character.
(5, 10) is provided, and determination is performed based on the result of the recognition means (5, 10).

The character slicing device of the present application comprises an input means (1) for inputting character data, a character number estimating means (57) for estimating the number of characters from the character data input by the input means (1), and the number of characters. Change means (54) for changing the cutout parameter based on the number of characters obtained by the estimating means (57), and the changing means means
It is characterized by further comprising a cutting position detecting means (55) for detecting a cutting position based on the cutting parameter changed by (54).

The character slicing device of the present application comprises an input means (1) for inputting character data, a character number estimating means (57) for estimating the number of characters from the character data input by the input means (1), and the number of characters. Change means (54) for changing the cutout parameter based on the number of characters obtained by the estimating means (57), and the changing means means
(54) based on the number of characters of the result obtained by the cutout position detection means (55) and this cutout position detection means (55) for detecting the cutout position by the cutout parameter changed by
Re-processing instruction means (58) for performing the character cutting processing again is provided.

The character slicing device of the present application comprises an input means (1) for inputting character data, a character number estimating means (57) for estimating the number of characters from the character data input by the input means (1), and the character number. Change means (54) for changing the cutout parameter based on the number of characters obtained by the estimating means (57), and the changing means means
(54) cut-out position detecting means for detecting the cut-out position by the cut-out parameter changed by (54), the number of characters of the result obtained by this cut-out position detecting means (55), and the character number estimating means (57) When the number of characters in the former is smaller than the number of characters in the latter, the reprocessing instruction to perform the character cutting process again based on the number of characters in the result obtained by this cutting position detection means (55). Means (5
8) and are provided.

The character slicing device of the present application comprises an input means (1) for inputting character data, a character number estimating means (57) for estimating the number of characters from the character data input by the input means (1), and the number of characters. Change means (54) for changing the cutout parameter based on the number of characters obtained by the estimating means (57), and the changing means means
The detection of the cut-out position by the cut-out parameter changed by (54) is evaluated at least by dividing it into characters and between characters, and by this separately evaluated synthesis, the cut-out position detecting means (55) for detecting the cut-out position and this Based on the number of characters of the result obtained by the cutout position detection means (55), when performing the character cutout process again, the evaluation as a character in the cutout position detection means (55) is the character number estimation means (57)
Evaluate based on the number of characters obtained in, the evaluation as the inter-character evaluation is based on the number of characters of the result obtained by the cut-out position determining means (55) re-processing instruction means to perform the character cutting process again ( 58) and are provided.

The character slicing device according to the present application includes an input means (1) for inputting handwriting data, and a character slicing device for slicing characters from the handwriting data input by the input means (1). Character type determination means (51,51 ')
And a changing means (52, 54) for changing the cutout parameter based on the character type determined by the character type determining means (51, 51 ') and the cutout parameter changed by the changing means (52, 54). And a cutout position detecting means (53, 55) for detecting the cutout position.

According to the present invention, in the character cutout method of the character cutout device for judging the cutout position by determining the characterlikeness and the characterlikeness of the character data inputted from the input means (1), the characterlikeness It is characterized in that the reference data for determining the character likelihood is set individually. The present application is a character cutout for detecting a cutout position by determining character-likeness and character-likeness of character data based on a reference character width and a reference character pitch, even if there is at least character data input from the input means (1). In the character cutting method of the device, the values of the reference character width and the reference character pitch are individually set.

[0035]

In the character cutting device of the present invention, the character type of the writing data input by the input means (1) is determined by the character type determining means (51, 51 ').
And the cut-out parameter is changed based on the character type determined by the character type determination means (51, 51 ′). Further, in the character cutting device of the present invention, the number of characters of the writing data input by the input means (1) is estimated by the number-of-characters estimating means (57), and the changing means (54) cuts out the characters based on the obtained number of characters. And the cut-out position detecting means (55) based on the evaluation result.
Use to determine the character cutout position.

In the character cutting method according to the present invention, it is possible to individually judge the one-character likelihood and the inter-character likelihood.

[0037]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the character cutting device of the present application will be described with reference to FIGS. FIG. 5 is a diagram showing a configuration of the character cutout unit 7. In the figure, the same parts as in FIG.
The same reference numerals are given and the description is omitted.

Reference numeral 7 is a cutout portion. Reference numeral 51 is a character type determination unit for the input character string. The character type determination unit 51 determines what kind of character type the input character string is composed of. 7
2'is a determination unit. Reference numeral 52 is a determination threshold value changing unit for changing the cutout parameter according to the character type determined by the character type determining unit 51. Reference numeral 53 is a determination processing unit that performs determination between characters / non-characters.

That is, the determining unit 72 'is the character type determining unit 51.
The character-to-character / non-character-to-character determination is performed according to the character type determined in. 73 'is an optimum route detection unit. Reference numeral 54 is a cut-out feature evaluation unit whose parameters are changed according to the character type determined by the character type determination unit 51. Reference numeral 55 is a cutout position determination unit that determines the cutout position.

That is, the optimum route detecting section 73 'is a judgment threshold value changing section for changing the cutout parameter according to the character type judged by the character type judging section 51. 53 is the space between characters
It is a determination processing unit that performs determination between non-characters. FIG. 6 shows an example of a character type determination method in the character type determination unit 51. Figure 6
61 is an icon display area for input character type setting.

In this example, icons (61a to 61f) corresponding to the character types are provided on the tablet. Then, the input character type is registered by pointing the icon corresponding to the character type desired by the user with a pen before inputting. In FIG. 6, the input is limited to numbers and symbols. This method is particularly effective in the case of continuously inputting a specific character type such as numbers for slip input and alphabetic characters for English input.

Next, the judging unit 72 'and the optimum route detecting unit 7
The operation 3'will be described. The determination by the determination unit 72 ′ is the same as the determination by the OFF time, the FAR time, and the space width of the determination unit 72 in FIG. 1, and the determination processing is performed by the internal determination processing unit 53. However, each threshold is changed by the determination threshold changing unit 52 according to the result of the character type determining unit 51.

For example, when it is determined from the result of the character type determination unit 51 that the input character string is a numeral or an alphabetic character,
Set the space width threshold between characters from 3.0mm to 2.0mm
Change to This is because numbers and English characters are not extremely separated in the writing direction within a character like Kanji's biases and straw.

Also, even for characters that may be separated such as "5" and "7" in the numbers, the space width between one stroke and two strokes is extremely small, so even if the threshold value is changed to 1.0 mm. Good. The OFF time threshold may also be changed depending on the character type. The optimal path detection unit 73 ′ calculates the feature quantity when calculating the feature quantity of “character-likeness” and “character-likeness” based on the result of the character type determination unit 51.
Change the parameter according to the character type.

For example, when the character type determination unit 51 determines that a Chinese character is included, the standard character width is not set to be less than or equal to the height of the character string. In addition, by the character type determination unit 51,
If it is determined to be a character other than Kanji, the standard value of each feature amount can be set to a threshold value such as an aspect ratio (horizontal / vertical) of the standard character of 1.5 or less, so that accurate clipping can be performed.

Further, the method of calculating the characteristic amount can be changed depending on the character type. When a Chinese character is included, the character shape is greatly deformed, so the standard character width is obtained from the maximum width of the basic segment, and in other cases, it is the same as the height of the character string. By the way, in the first embodiment, when a general character string is input, it is necessary to frequently set the character type, so it is desirable that the character type can be automatically determined from the writing data. A second embodiment will be described in which the character type is automatically determined.

A second embodiment of the character cutting device of the present invention will be described with reference to FIG. FIG. 7 is a diagram showing a configuration of the character cutout unit 7. In the figure, the same parts as those in FIG. 1 are omitted, and the same parts as those in FIGS.
The same reference numerals are given to simplify the description. Reference numeral 7 is a cutout portion.

Reference numeral 72 is the same as the determination unit in FIG. 51 'is a character type determination unit for the input character string. The character type determination unit 51 'determines what type of character the input character string is composed of. 72 'is a determination unit. The determining unit 72 'changes the parameters for determining the character spacing / non-character spacing according to the character type determined by the character type determining unit 51' to determine the character spacing / non-character spacing. I do.

Reference numeral 73 'is an optimum route detecting section. The optimum route detection unit 73 'changes the parameters for determining the optimum route according to the character type determined by the character type determination unit 51, and determines the cutout position. That is, the optimum path detection unit 73 ′ is a determination threshold value changing unit that changes the cutout parameter according to the character type determined by the character type determination unit 51 ′. Reference numeral 53 is a determination processing unit that performs determination between characters / non-characters.

The character type determination method in this character type determination unit 51 'will be described. 1. Character type determination method 1 The determination unit 72 extracts a tentative basic segment with thresholds of OFF time, FAR time, and space width, as in the conventional case,
Calculate the number of strokes for each. When the number of basic segments is equal to or larger than the threshold value (e.g., 5) and the stroke number is within 2 strokes, the character type determination unit 51 'determines that the character string is a numeral or a lowercase letter. As a method for further increasing the certainty factor, a condition that one basic segment is continuous may be added.

Further, the determination of the number and the letter is made by examining the shape and determining the characteristic of each of the number and the letter. As a determination condition, it is particularly effective to use “0” or “1” in numbers easily in succession and the presence or absence of “i” or “j” in which strokes are separated in the vertical direction in lowercase letters. 2. Character type determination method 2 The determination unit 72 extracts a temporary basic segment with thresholds of OFF time, FAR time, and space width, as in the conventional case,
Calculate the number of strokes for each.

When the number of basic segments is equal to or greater than the threshold value (e.g., 5) and the number of strokes is 4 or more, the character type determination unit 51 'determines that the character string includes Chinese characters. 3. Character Type Determination Method 3 The determination unit 72 extracts a temporary basic segment with the thresholds of the OFF time, the FAR time, and the space width, as in the conventional case. Then, for each of them, the recognition unit 5 of FIG. 1 obtains a temporary character recognition result (character code). The character type determination unit 16 determines the character type based on the number and arrangement of the same character types.

As described above, numbers, English characters, Chinese characters, Katakana, etc. have their own writing characteristics. It goes without saying that the above-mentioned method is not the only method of using the writing feature for automatic character type determination. The determining unit 72 'and the optimum route detecting unit 73' operate in the same manner as in the first embodiment.

However, in the second embodiment, the character type determining unit 51 'automatically determines the character type from the writing data, and thus the determined character type may be incorrect. Therefore, the threshold value (various cutout parameters) in the determination unit 72 ′
It is better to set the change amount of 1 to be smaller than that in the first embodiment. Although the character type determination unit 51 'is specially provided in this embodiment, the result of the recognition unit 5 may be used. That is,
If the character types of the adjacent characters recognized by the recognition unit are the same, this character may be estimated to be the same character type, and various cutout parameters may be changed.

A third embodiment of the character cutting device of the present invention will be described with reference to FIGS. 8 and 9. FIG. 8 is a diagram showing a configuration of the character cutout unit 7. In FIG. 8, FIG.
1 are omitted, and the same parts as in FIG.
The same reference numerals are given to simplify the description. The optimum path detection unit 73 ″ detects a highly likely combination of stroke sets obtained based on the result of the determination unit 72, and operates in the same manner as the conventional one. This is a point including the portion 56.

The post-processing unit 56 uses the optimum route detection unit 7
With reference to the space width between the characters determined to be between the characters in 3 ", the character space is detected again. The optimum route detection unit 73" is "single character likeness" as described in FIG. Then, the optimum amount of route is calculated by calculating the feature amount representing "character-likeness" and calculating the total sum for each route.

As described in the above-mentioned Japanese Patent Laid-Open No. 6-124364, the feature amount of "character likeness" uses the aspect ratio of the character, the character width, the size of the character, the maximum space width in the character, etc. Deciding. Further, the feature amount of "character-likeness" is determined by using the character pitch, the space width between characters, etc., as shown in the above-mentioned Japanese Patent Laid-Open No. 6-124364.

The character format data is preliminarily set, or is estimated from the input character data. Here, the input of FIG.
3 "is determined in FIG. 9B. Here, the post-processing unit 56 detects the space width determined to be between characters by the optimum path detection unit 73".

That is, in the case of W1, W2. W3 is detected, and a blank space larger than the minimum value of this width is determined to be between characters, and FIG. 9 (d) is output. In this way, the optimum route detection unit 73 ″ determines,
By diverting the data between the characters and performing the character segmentation again, the character segmentation can be performed accurately even when a character having a different size is input.

However, in the case of the input as shown in FIG. 9 (e),
In some cases, it is cut out by mistake. Therefore, the post-processing unit 73 ″ determines that the space between the detected characters, W1 W2.W.
A space larger than the minimum value of 3 and larger than a predetermined width may be determined to be between characters. Figure 10
A fourth embodiment of the character cutting device of the present application will be described with reference to FIGS. FIG. 10 is a diagram showing a configuration of the character cutout unit 7. 10, the same parts as those in FIG. 1 are omitted, and the same parts as those in FIGS. 1 and 5 are designated by the same reference numerals to simplify the description.

The optimum route detection unit 73 "'detects a combination of stroke sets obtained on the basis of the result of the determination unit 72, which has a high possibility. The character number determination unit 57 estimates a standard character width and a standard character pitch from the input character data. An example of calculation for estimating the standard character width and standard character pitch will be shown.

Standard character width = height of character string × constant. Standard character pitch = height of character string x constant. Another example of calculation for estimating the standard character width and standard character pitch will be shown. Standard character width = maximum width of basic segment. Standard character pitch = maximum pitch of basic segment.

Then, the number of written characters is estimated from the above result. Estimated number of characters = (width of character string−standard character width) / standard character pitch From the estimated number of characters obtained by the character number determination unit 57, the number of existing characters between characters at this time can be estimated. The operation of the character number determination unit 57 in the fourth embodiment will be described by taking the case of input in FIG. 2 as an example.

As shown in FIG. 2, the space width E7 between the basic segments E5 is -1 mm, 4 mm, 2 m in order from the left.
m, 3.5 mm, 1 mm. If the space widths are sorted in descending order, 4mm, 3.5mm, 2m
m, 1 mm, -1 mm. It is assumed that the number of estimated characters is estimated to be “2 characters” by the above equation.

In the example of FIG. 2, as shown by E4, b
Between c and c and between d and e are definitive cut-out positions, the correct answer is at least 3 estimated characters. In the example of FIG. 2, the number of characters is estimated to be two when one character is written large as shown in FIG. Estimated number of characters is "2 characters"
This means that there is only one existing character. And the maximum space width is 4mm, so this 4m
m is the space between characters, and the second space width is 3.5 m
m is estimated to be the space present in the character.

That is, the standard of the space width is between 4 mm and 3.5 mm. The space width criterion is a threshold value that can be determined to be “likely between characters” if it is larger than that value and “likely to be non-character space” if smaller than that value. Here, the standard of this space width is set to 3.75 mm, which is between 4 mm and 3.5 mm.

In the stroke set E6, the smaller the maximum space width in the stroke set is, the higher the evaluation of "uniqueness" and the more difficult it is to separate. On the other hand, if the space width between stroke sets is larger than this value, the "character-likeness" is evaluated more and the stroke sets are more likely to be cut. That is,
The character number determination unit 57 estimates the number of characters, and according to the estimated number of characters, the reference of the space width of the cut-out feature evaluation unit 54 is
change.

The optimum route detection unit 73 "'is changed in this way, and in other respects, the cutout processing is performed in the same manner as in the prior art. Then," character-likeness "and" character-likeness "
The optimum route is obtained by obtaining the feature amount of and the total of the routes as in the conventional method. In this case, 2-5-8 was the highest score, and 2-4-6-8 was second.

From this result, character recognition and language processing may be performed as in the conventional case to obtain an optimum character string. However, the number of characters of the combination of 2-5-8 which obtained the highest score is "3 characters", and the result different from the estimated number of characters "2 characters" is obtained. The feature quantities used to calculate "result-likeness" and "character-likeness" used to calculate this result are
Since it is the evaluation value when the number of characters is “2 characters”, the evaluation of “character-likeness” and “character-likeness” may be redone based on the information “3 characters” obtained again.

An example of re-evaluation will be described as a fifth embodiment. A fifth embodiment of the character cutting device of the present application will be described with reference to FIG. FIG. 12 is a diagram showing a configuration of the character cutout unit 7. In FIG. 12, the same parts as in FIG. 1 are omitted, and the same parts as in FIGS.
The same reference numerals are given to simplify the description.

The optimum path detecting section 73 "" detects a highly probable combination of stroke sets obtained based on the result of the judging section 72. The number-of-characters change detection unit 58 of the optimum path detection unit 73 "" detects this when the number of characters in the cutout result is larger than the number of characters estimated by the number-of-characters determination unit 57. Then, this new number of characters is reset in the number-of-characters determination unit 57, and the optimum route detection unit 7
Cut out 3 "" again.

The number-of-characters determination unit 57 of the optimum route detection unit 73 "" changes the space width reference calculated and used in the case of "2 characters" in the example of FIG. As mentioned above, if the space widths are sorted in descending order, 4mm,
It becomes 3.5 mm, 2 mm, 1 mm, and -1 mm. The number of characters is "3 characters", which means that there is 2 between existing characters.
Individual.

The space width of 4 mm and 3.5 mm
Is between the characters, and the third space width is 2 mm,
It is presumed to be the space existing in the character. In other words, for the evaluation of the space width, the standard is 3.5 mm and 2 mm.
It is performed by resetting the distance to 3 mm. Further, the character pitch is recalculated by the following formula.

Standard character pitch = (width of character string−standard character width) / number of resulting characters. The standard character width, standard character aspect ratio, and standard character size are not changed during this re-evaluation. Based on the evaluation values of the features revised as described above, the "single character likeness" and the "character spacing likeness" are calculated by the optimum path detecting unit, and the cutout result is obtained.

The number of characters change detection unit 58 may detect that the "number of characters" as a result of this clipping has further increased, and this process may be performed again, but it is effective only once. In addition, at the time of re-evaluation using the number of characters of the cut out result,
The standard character width, the aspect ratio of the standard character, and the size of the standard character are not changed in order to handle large and small characters even if they are mixed.

[0076]

According to the character slicing device of claims 1 to 5 and 10 of the present application, the character slicing process is performed by determining the character type configuration of the input character string and performing the character slicing process according to the determination result. The accuracy can be improved. Claim 4 of the present application
According to the character cutting device of No. 5, it is possible to automatically judge the character type configuration of the input character string and perform the character cutting process according to the judgment result.

According to the character slicing device of claims 6 to 9 of the present application, it is possible to improve the character slicing accuracy even for irregular character data having different sizes and the like. According to the character cutout method of claims 11 and 12 of the present application, the criteria for determining the likelihood of one character and the likelihood of character spacing can be set individually. Therefore, by making one correspond to a value corresponding to a large character (for example, character width) and the other to a value set for a small character (for example, character pitch), the cutout accuracy is improved when irregular characters are input. can do.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a conventional character cutting device.

FIG. 2 is a diagram for explaining a basic segment / stroke set extraction method of the character segmentation device.

FIG. 3 is a diagram in which cutout evaluation values are expressed by a network.

FIG. 4 is a diagram showing an input example of a character cutting device.

FIG. 5 is a diagram showing an outline of a cutout portion 7 of a second embodiment of the present invention.

FIG. 6 is a diagram for explaining an example of a character type determination method according to the second embodiment.

FIG. 7 is a diagram showing an outline of a cutout portion 7 of a second embodiment of the present invention.

FIG. 8 is a diagram schematically showing a cutout portion 7 of a third embodiment of the present invention.

FIG. 9 is a diagram for explaining the operation of the third embodiment.

FIG. 10 is a diagram showing an outline of a cutout portion 7 of a fourth embodiment of the present invention.

FIG. 11 is a diagram for explaining the operation of the fourth embodiment.

FIG. 12 is a diagram showing an outline of a cutout portion 7 of a fifth embodiment of the present invention.

[Explanation of symbols]

1 input unit (input unit), 5 character recognition unit (recognition unit), 7 cutout unit, 71 stroke sequence extraction unit, 72 'determination unit, 73' optimal route detection unit, 73 "optimal route detection unit, 73"'optimal Route detection unit, 73 "" optimum route detection unit, 51 character type determination unit (character type determination unit), 51 'character type determination unit (character type determination unit), 52 determination threshold value changing unit (changing unit), 53 determination processing Section (cutout position detecting means), 54 cutout feature evaluation section (changing means), 55 cutout position determining section (cutout position detecting means), 56 post-processing section, 57 character number determination section (character number estimating means), 58 character number change detecting section (Reprocessing instruction means).

Claims (12)

[Claims] 1. An input means (1) for inputting character data,
In a character cutout device for cutting out characters from the character data input by the input means (1), a character type determination means (51, 51) for determining the character type of the input character set.
51 ') and changing means (52, 5) for changing the cut-out parameter based on the character type determined by the character type determining means (51, 51').
4) A character cutting device including and. 2. An input means (1) for inputting character data,
In a character cutout device for cutting out characters from the character data input by the input means (1), a character type determination means (51, 51) for determining the character type of the input character set.
51 ') and changing means for selectively using a plurality of cutting methods based on the character type determined by the character type determining means (51, 51')
A character cutting device including (52, 54). 3. The character type judging means (51) judges as follows.
3. The character cutting device according to claim 1, wherein the character cutting device is performed based on the setting of the character type setting mode. 4. The character cutting device according to claim 1, wherein the character type judging means (51 ′) makes the judgment based on the number of strokes or the shape of the character data. 5. The determination by the character type determination means is characterized in that a recognition means (5, 10) for recognizing the cut out character is provided and the determination is made based on a result of the recognition means (5, 10). The character cutting device according to claim 1 or 2. 6. The input means (1) for inputting character data, the character number estimating means (57) for estimating the number of characters from the character data input by the input means (1), and the character number estimating means (57). Changing means (54) for changing the cutout parameter based on the obtained number of characters, and cutout position detecting means for detecting the cutout position by the cutout parameter changed by the changing means (54)
(55) A character slicing device comprising: Character cutting device equipped with. 7. The input means (1) for inputting character data, the character number estimating means (57) for estimating the number of characters from the character data input by the input means (1), and the character number estimating means (57). Changing means (54) for changing the cutout parameter based on the obtained number of characters, and cutout position detecting means for detecting the cutout position by the cutout parameter changed by the changing means (54)
A character cutting device comprising (55) and a re-processing instruction means (58) for performing a character cutting process again based on the number of characters obtained as a result of the cutting position detecting means (55). 8. The input means (1) for inputting character data, the character number estimating means (57) for estimating the number of characters from the character data input by the input means (1), and the character number estimating means (57). Changing means (54) for changing the cutout parameter based on the obtained number of characters, and cutout position detecting means for detecting the cutout position by the cutout parameter changed by the changing means (54)
(55) and the number of characters of the result obtained by this cut-out position detection means (55) and the number of characters of the result obtained by the number-of-characters estimation means (57) are compared, and the number of characters of the former is smaller than the number of characters of the latter. Sometimes, the character cutting device further comprises reprocessing instruction means (58) for performing the character cutting processing again based on the number of characters obtained as a result of the cutting position detecting means (55). 9. An input means (1) for inputting character data, a character number estimating means (57) for estimating the number of characters from the character data input by the input means (1), and the character number estimating means (57). The changing means (54) for changing the cutout parameter based on the obtained number of characters, and the cutout position detected by the cutout parameter changed by the changing means means (54) are evaluated at least by dividing the character and the space between the characters. , By this separately evaluated synthesis,
Based on the number of characters of the result obtained by the cutout position detection means (55) and the cutout position detection means (55) to detect the cutout position, when performing the character cutout process again, in the cutout position detection means (55) The evaluation as a character is evaluated based on the number of characters obtained by the number-of-characters estimating means (57), and the evaluation as a character space is based on the number of characters of the result obtained by the cut-out position determining means (55). A character cutting device comprising reprocessing instruction means (58) for performing the character cutting process again. 10. Input means for inputting handwriting data (1)
And a character slicing device for slicing a character from the handwriting data input by the input means (1), a character type determining means (51,
51 '), a changing means (52, 54) for changing the cut-out parameter based on the character type determined by the character type determining means (51, 51'), and the changing means (52, 54). Position detecting means for detecting the cutting position by the cutting parameter
(53, 55) and a character cutting device. 11. A character cutout method of a character cutout device for detecting a cutout position by determining whether one character and the likeness of character data inputted from the input means (1) are judged, and A character cutting method characterized by individually setting reference data for determining the likelihood of characters. 12. The reference character width and the reference character pitch allow at least the character data input from the input means (1) to be
Judge the character-likeness and character-likeness of character data,
A character cutting method for a character cutting device for detecting a cutting position, wherein the values of the reference character width and the reference character pitch are individually set.