JPH0792817B2 - Online character recognizer - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to character recognition, and more particularly to an online character recognition device including Chinese characters.

[0002]

2. Description of the Related Art In the prior art, there is a method of character recognition as a combination of subset patterns obtained by subdividing Chinese characters as shown in Japanese Patent Laid-Open No. 58-82373. According to this method, as shown in Japanese Patent Laid-Open No. 58-82373, kanji are subdivided to recognize a sub-set of subdivided characters, and the kanji are identified as a set of the sub-sets which are recognized. This is a method in which the number of target characters can be reduced, the storage capacity required for the dictionary can be reduced, and as a result, the processing time required for recognition can be reduced.

[0003]

However, according to the above-mentioned method, the number of kanji to be recognized is up to about 2000 or less than the common kanji and the number of characters exceeds that, for example, the JIS 1st level kanji set 2965 or the JIS 1st level kanji set. There is a problem when you want to recognize Kanji used in place names, surnames, company names, etc. that are not included in the above as recognition targets.

For example, the letter "Akira" means

[Outer 1] According to the description of JP-A-58-82373,
Identified as the letter "Akira". However, if the target characters are up to JIS Level 1 Kanji set, for example, the characters "para" and "koro"

[Outside 2] However, it is difficult to identify.

Further, the character "Kan" has the same stroke as the character "Ki" that is judged from the combination of strokes, and if the stroke order is free, partial patterns of "Ta", "Half" and "Tree", and "Kichi" are formed. Therefore, it is still difficult to identify the characters.

There are about 1500 kinds of confusion shown in the above example only in the JIS first level kanji set. Not only two letters are distinguished in one kind here

[Outside 3] Most of the combinations have more than two characters such as, and it can be said that the discrimination processing of these is theoretically possible but practically impossible.

In order to solve this, there is a proposal that the stroke order is fixed to the content written in the guide for stroke order guidance issued by the Ministry of Education, but this is also not practical. That is, when the recognition target characters are expanded to the JIS first-level kanji set, it can be said that it is practically impossible to recognize the characters by the conventional recognition logic.

Some of the characters have the same or similar shapes and have different meanings (duplicate characters), such as D and D, R and L and M and C. It was difficult to recognize all the character types, because there were many such as and eight.

An object of the present invention is to solve the above problems and to identify a target character with a simple method even when the number of figurative characters exceeds 3000.

[0010]

The essential point of the present invention is to provide a partial pattern detail dictionary for verifying whether or not a subset pattern (hereinafter referred to as a partial pattern) itself is the partial pattern, After checking the necessary and sufficient conditions for the partial pattern by the detailed dictionary, it is recognized as a partial pattern and then the character recognition is performed. For a plurality of characters composed of exactly the same partial pattern, the character detailed dictionary Is provided, and a plurality of characters having the same or similar shapes are provided with an overlapping dictionary and are identified by some processing.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram showing an embodiment of the present invention, in which 1 is a tablet which shows the coordinates of a writing tool online when writing a character, and 2 is a stroke which limits the strokes written from the coordinates of the writing tool. A recognition unit 3, a stroke dictionary in which stroke information is stored in advance, a stroke sequence register for accumulating the output of the stroke recognition unit 2, a partial pattern recognition unit for identifying a partial pattern,
6 is a partial pattern dictionary in which partial patterns consisting of combinations of strokes are stored, 7 is a partial pattern detailed check dictionary in which various conditions for being a partial pattern are stored, and 8 is a plurality of characters having the same or similar shapes. The stored duplicate dictionary, 9 is a partial pattern string register, 10 is a character recognition unit that identifies a character from the identified partial pattern, 11 is a character dictionary that stores characters that are combinations of partial patterns, and 12 is a partial pattern. It is a character detail check dictionary in which the positional relationship of is stored.

Next, the operation will be described. The stroke written on the tablet 1 is output to the stroke recognition unit 2 and is recognized as one upper stroke compared with the reference data of the stroke dictionary 3 and input to the stroke sequence register 4. The stroke here means from the touch of the pen on the tablet to the release.

In the case of the embodiment of the present invention, the angle of the stroke,
The strokes are classified into 76 strokes using the direction, rotation, and loop. Therefore, the stroke dictionary 3 describes the characteristics of these strokes. FIG. 2 is an explanatory diagram showing a stroke to be recognized. The stroke recognizing method shown in FIG. 2 is not the object of the present invention, and therefore its explanation is omitted, but it can be easily recognized by the well-known stroke analysis method or KL expansion method. In the figure, SXX indicates a stroke code, and the right side of the stroke code indicates the object. In addition, each stroke has a width in the direction, length, etc.

[Outside 4] Is also defined as S04.

The following description will be given for the case of writing the character "member". FIG. 3 is an image diagram of the input. The numbers in parentheses in Fig. 3 indicate the stroke input order.
Indicates an input stroke code. Before describing the recognition method of the partial pattern recognition unit 5, a description method of each dictionary will be described for easy understanding. 4 and 5 are examples of the partial pattern dictionary 6, SC indicates a stroke code, which is shown in FIG.

The first line means that a stroke of stroke code S04 constitutes a partial pattern of 001-0. Similarly, the second line is a combination of stroke codes S01 and S03.

[Outside 5] Means that a partial pattern such as is constructed. On the right side of the stroke code and the partial pattern code, typical shapes shown by the respective codes are shown. In addition, a part pattern code is not displayed and only a figure is displayed, indicating that it is an independent character, not a partial pattern.

As is clear from FIGS. 4 and 5, one partial pattern is composed of a maximum of seven strokes.
4 and 5, only parts necessary for explaining the present invention are extracted in order to avoid redundancy, and actually, the number of items is about 650 and the number of partial patterns is about 370.
In order to simplify the explanation, the corresponding JIS 6226 code is added to the independent characters which only show the shape described above.

Next, as described above, the strokes written on the tablet 1 are recognized by the stroke recognizing unit 2 and input to the stroke sequence register 4. The stroke sequence register 4 is given the same number as the input order. The stroke code recognized by the stroke recognition unit 2 is stored in the register.

Using the stroke code stored in the stroke sequence register 4, the coordinate data of the start and end points of this stroke, and the information of the partial pattern dictionary 6, the partial pattern detail check dictionary 7, and the duplicate dictionary 8, the partial pattern recognition unit 5 is used. A character that does not become a partial pattern or a partial pattern of 7 strokes or less is recognized by. This situation will be described with reference to FIGS. 6, 7 and 8.

FIGS. 6, 7 and 8 are partial pattern recognition state diagrams showing the progress of processing each time a stroke code is input. In the figure, "input order" is the same as the input order shown in FIG. 3, "input stroke" shows its shape, and "stroke code" shows the stroke code as a result recognized by the stroke recognition unit 2. As described above, the stroke train register 4 has the same structure as shown in FIGS.
There is a register group corresponding to the “input order”, and the stroke code is stored in each register.

First, in input order (1)

[Outside 6] Is input, as is apparent from FIGS. 4 and 5, S02
There is no partial pattern candidate in the input order (1) because it does not become a partial pattern by itself.

In the input order (2)

[Outside 7] Is input, the strokes in the input order (1) and (2) are combined and the partial pattern dictionary 6
From (FIG. 4), 011-1, 011-2, 015-1, 0
Partial patterns 15-2, 015-3, and 015-4 and the characters "7" and "u" are candidates.

Here, it is possible that S02 of the input order (1) and S06 of the input order (2) are each a single partial pattern, but since S02 of the input order (1) is not registered as a single partial pattern, the input order is It is not necessary to consider the partial pattern starting with (2). The reason is that when the number of strokes from the first stroke (input order (1)) to the last stroke is 7 or less, all of them become single characters or partial pattern dictionary 6 in which partial patterns become characters. , Because the character dictionary 11 is configured. Therefore, the input up to the input order (2) is limited to the partial pattern composed of the strokes S02 and S06.

The partial pattern codes in FIGS. 6, 7 and 8 have 3 digits or 1 digit with a branch number in 3 digits. This means that a partial pattern expressed by a combination of a certain stroke group is 1 digit. If there is only one, it is displayed with three digits, and if there are multiple digits, it is displayed with a branch number.

Here, a further explanation will be given with reference to FIGS. 6, 7 and 8. 1, 5, 7, 16, 23, 27, and 40 cannot have a partial pattern or character because there is no combination of those strokes in the partial pattern dictionary 6. In addition, No. In the case of 8, the stroke S02 in the input order (1) cannot be one partial pattern with one stroke, so it is not necessary to consider the partial pattern starting from the input order (2).

Next, in the case of the input order (2) and thereafter, a combination of these strokes cannot become a character. This is because a single partial pattern that makes one character always includes the first stroke. Therefore, in the case of a combination of strokes in the input order (2) and thereafter, 1 is selected from the partial pattern dictionary 6
Even if there is a candidate for one character of the partial pattern, it may be ignored. Therefore, No. 2 and No. No. 6 is a combination of the same strokes, but No. 2 candidates for the characters "7" and "U" in 2 are N
o. 6 does not exist.

To explain further, the stroke S04 has a shape.

[Outside 8] Is output as a partial pattern candidate.

As described above, the stroke S04 is the most extreme example, but all strokes are defined with a certain range. Further, since the partial pattern is defined by 7 strokes or less, it is not necessary to consider the partial pattern exceeding 7 strokes.

The stroke order in the partial pattern and the stroke order of the partial pattern in the character are free, but the stroke order in one partial pattern must be continuous even if the stroke order is free. Skipping a plurality of strokes does not form one partial pattern. Therefore, the lines showing the partial patterns in FIGS. 6, 7, and 8 are continuous.

Next, the partial pattern detail check dictionary 7 will be described. FIG. 9 shows a partial pattern detailed check dictionary 7
It is explanatory drawing which shows the content of. The partial pattern detail check dictionary 7 is for checking which partial pattern is a combination of written strokes. For convenience of explanation, FIG. 9 shows Nos. 6 and 7 and 8. The items to be checked corresponding to No. 1 are picked up and rewritten. The columns correspond to those in FIG. 6, FIG. 7 and FIG. After the hyphen in the column, the check order is shown, the strokes A and B describe the stroke to be checked, and the check content column describes what to check. In this way, the coordinate data detected by the tablet 1 can be used to check the start point, end point, length, etc. of each stroke. The B code column describes the partial pattern code when the candidate is narrowed down by the check, and the JIS code column describes the JIS 6226 code for what can be a character when the candidate is narrowed down. The candidate column shows the shape of what remains as a candidate at the time of the check.

As will be easily understood by those skilled in the art, as a dictionary used for this processing, the check item column corresponding to the check content of FIG. There is a reference data column for the check result of No. If there is an input stroke, the above check item column is checked and the result and the reference data column match. The procedure is such that the candidates are determined when they match, but the description method as shown in FIG. 9 is adopted because it becomes complicated if the description is made according to the regular procedure.

The manner in which the partial pattern candidates are squeezed using the partial pattern detail check dictionary 7 of FIG. 9 described above will be described again with reference to FIGS. 3, 6, 7 and 8.

In FIG. 3, strokes are input in the order (1),
The case where (2) is written will be described. In this case, as described above, the partial patterns 011-1, 011-
2, 015-1, 015-2015-3, 0154, and the characters “7” and “U” are candidates and may be considered as 011-1 from the stroke shape.

That is, FIG. Reference numerals 2-1 and 2-2 show a process in which the partial pattern 011-1 is selected. N
o. From 2-1 there is a condition that "the start point of S02 and the start point of S06 are close", and the partial pattern selected when this condition is checked is

[Outside 9] And the character is "7". Furthermore, No. From 2-2, from the condition that “S02 and S06 have the same length in the Y direction”

[Outside 10] However, all other candidates are eliminated.

In this way, the No. As shown in 2-2, the partial pattern 011-1 has no remaining characters. Further, as described above, only the pattern satisfying the necessary and sufficient conditions for being recognized as a partial pattern remains as a partial pattern.

FIG. 10 shows remaining candidates as a result of checking the candidates appearing from the partial pattern dictionary 6 shown in FIGS. 4 and 5 by the partial pattern detail check dictionary 7 shown in FIG. 9 as described above. FIG. 10 is a diagram showing partial pattern candidates. “Input order”, “input stroke”, and “stroke code” in FIG. 10 are shown in FIGS. 6, 7 and 8.
Similar to the above, there is provided below the stroke code a figure column showing the figure from the first stroke to the stroke.

No. 9 and below show its shape for the partial pattern starting from the stroke of the input order in the middle in [] on the left side of each line. As described above, the candidate partial patterns are selected in FIG. These pieces of information are sent to the partial pattern string register 9 and further processed by the character recognition unit 10. It can be considered that the partial pattern string register 9 stores the state of FIG.

The character recognition unit 10 processes in the following procedure. That is, the partial pattern string register 9 has 10 strokes, and the number of strokes for forming one character in one partial pattern is 7
Since it exceeds, it cannot be a single character.

Further, since it becomes one character and is expressed by some combination of partial patterns, the lines showing the partial patterns in FIG. 10 must be connected from the input order (1) to (10) to form one character. There is. That is, in principle, the combination of the input order (1) (2), (3) to (6), (7) (8), and (9) (10) can also be a character. In other words, this combination becomes a character candidate.

Table 1 shows combinations of candidate partial patterns as described above.

[Table 1]

As described above, in this case, 10 combinations are rearranged in numerical order and it is checked whether or not they are registered in the character dictionary 11. An example of the character dictionary 11 is shown in FIG. FIG. 11 shows what character is obtained by combining the partial pattern number and its shape. The partial pattern column shows a combination of a partial pattern of three digits and a branch number of one digit and a total of four digits. The right side of the partial pattern number shows the shape of the partial pattern for reference.

The character field is the JIS code of the character formed by combining the partial patterns in the partial pattern field and the shape of the character. The shape is shown for reference. Character detail dictionary No. Indicates the item number of the character detail check dictionary 12,
Indicates that the character detail check dictionary 12 must be referenced. In the character dictionary 11, character candidates are identified by a 3-digit partial pattern excluding branch numbers. By the way, in the case of the example of FIG. 10 just described, the partial patterns are 046, 1
It is shown by FIG. 11 that the characters “member” or the character “song” are given as 97 and 018. When a plurality of character candidates appear in this way, the character recognition unit 10 narrows down the candidates using the character detail check dictionary 12.

FIG. 12 is an example of the character detail check dictionary 12, and “No.” indicates a registration number and is a number (FIG. 11) indicated by the character dictionary 11. The number of candidates is the number of character candidates. 2 is shown at m. This means that there are two candidates. "Partial pattern A,
“B” indicates which of a plurality of partial patterns is to be checked, and here, 046 (mouth) and 197 (eye) are checked.

The branch number column will be described later. The positional relationship indicates the positional relationship between the partial patterns, and No. A for m
Is on B, the partial pattern 046 (mouth) is a partial pattern 19 from the data based on the written characters.
If it is above 7 (eyes), it means the character "member" represented by 3077 shown in the JIS code column of the character column.

Similarly, when the partial pattern 046 (mouth) is to the left of the partial pattern 197 (eyes), it means that it is the character "song."

As is clear from FIG. 3, the partial pattern 04
Since 6 (mouth) is on the partial pattern 197 (eye), the written character is “member”. In this way, in order to check the positional relationship between each partial pattern,
The coordinate data detected by the tablet 1 can be used.

Next, the branch number column of FIG. 12 will be described. As described with reference to FIG. 11, in the character dictionary 11, character candidates are identified by 3-digit partial pattern numbers that do not use branch numbers. this is

[Outside 11] This is because the objects that are close to the same shape are grouped together in order to prevent them from becoming unrecognizable.

For example, the partial pattern 005 is shown in FIG. 1
As shown in 0

[Outside 12] It is a collection of. Therefore, for example, the characters composed of the partial pattern 005 and the partial pattern 046 include three characters of "fortune", "old", and "leaf" as is apparent from FIG. In the case of such characters, especially "fortune" and "old" are written carefully in consideration of the difference between those characters even when written by a person.

[Outside 13] There are few mistakes. Therefore, there is a branch number term to identify the character by examining this feature. 12 No.
In n, the branch number of the partial pattern A is 2, which means that if the condition of the described positional relationship is satisfied, it is the character “fortune”. In this way, FIG. The characters "fortune", "old", and "leaf" are identified from the partial pattern 005 and the partial pattern 046 by the three digits after n. It should be noted that in FIG. 12, the part with a minus sign indicates that the part is not related to the identification.

As is clear from the above description, the character recognition unit 10
The written character code is recognized, identified, and output.

Next, the overlapping dictionary (DJ) 8 which is a feature of the present invention regarding the recognition of the partial pattern will be described. Some characters have completely the same or similar shapes and cannot be discriminated by a person by the character alone. Although these are called duplicate characters, an example of the duplicate dictionary is shown in FIG. FIG.
The numbers, letters, kana, katakana, hiragana, kanji, and symbols listed in are examples of duplicate characters, and this range expands depending on the tolerance of the writing method. The examples are written in exactly the same way or with only one slight difference, and it is difficult to distinguish them by the character alone. As a general rule, these letters are in the order of FIG.
Alphabetical, Katakana, Hiragana, Kanji, and symbols are selected in this order, but if the immediately preceding character is an Alphabet, Katakana, Hiragana or Kanji, and if the symbol or number is alphabetic, Katakana, Hiragana or Kanji , The partial pattern recognition unit 5 functions to select the symbol as the first candidate.

Of course, if the first candidate is not the character intended by the writer, the character type of alphabetic characters, Katakana, Hiragana or Kanji, and symbols is selected by a switch not shown according to the intention of the writer and desired. You can get the output of

In the above description, a dictionary format is used to store various kinds of information, but these may take a processing format in which each generated information is sequentially processed instead of the dictionary format. it can. That is, since the stroke dictionary 3, the duplicate dictionary 8, the character detail check dictionary 12 and the like have little contents, they do not necessarily have to be in the dictionary format.

Further, in the present invention, the partial pattern recognition unit 5
Although the character recognizing unit 10 and the character recognizing unit 10 are described as separate units, the processing units may be integrated into one unit without being separated. It is considered that this can be easily understood by those skilled in the art together with the fact that this can be realized without taking the above-mentioned dictionary format, and therefore detailed description will be omitted.

As described in detail above, by adopting the configuration of FIG. 1 according to the present recognition method, it is possible to easily have a large number of characters to be recognized of 3000 characters or more, and the processing time required for the recognition processing when all character types are mixed. It is possible to provide a small online character recognition device.

Further, as can be easily understood by those skilled in the art, the present invention can be realized not only by a hardware configuration but also by an input unit, a microcomputer and a memory. As a powerful input device for man-machine interface, it can greatly contribute to the company.

[0055]

Since the present invention is configured as described above, it has the effects described below. First, in a method of hierarchically constructing characters such as strokes, partial patterns, and characters, a partial pattern detail dictionary is provided at the partial pattern stage, and necessary and sufficient conditions for the partial patterns are verified, and then character recognition is performed. By doing so, it is possible to dramatically reduce the number of verifications of partial pattern combinations in the character recognition unit. That is, when the method of the present invention is not used, it is necessary to verify the combinations for all the candidate partial patterns in FIG. 5, but the number of combinations that need to be verified by the partial pattern detailed dictionary is 10 The processing time required for is dramatically reduced.

Next, the character dictionary is verified with the three-digit representative code of the partial pattern, so that the character can be correctly identified even if it is a little complicated. Further, when different characters are included in a combination of the same partial pattern representative codes, it is possible to easily distinguish them by providing a character detail dictionary.

Furthermore, since a duplicate dictionary in which the same or similar characters are collected is provided and the characters are judged and identified based on the context of the characters, it is possible to recognize all characters mixedly. It is possible to provide a device that eliminates the troublesomeness of the conventional online character recognition device of writing characters and can be greatly removed from the restriction on the man-machine interface.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a stroke to be recognized.

FIG. 3 is an input image diagram when writing the character “member”.

FIG. 4 is an explanatory diagram of a partial pattern dictionary.

FIG. 5 is an explanatory diagram of a partial pattern dictionary.

FIG. 6 is a partial pattern recognition state diagram.

FIG. 7 is a partial pattern recognition state diagram.

FIG. 8 is a partial pattern recognition state diagram.

FIG. 9 is an explanatory diagram of a partial pattern detail check dictionary.

FIG. 10 is an explanatory diagram showing partial pattern candidates.

FIG. 11 is an explanatory diagram of a character dictionary.

FIG. 12 is an explanatory diagram of a character detail check dictionary.

FIG. 13 is an explanatory diagram of a duplicate dictionary.

[Explanation of symbols]

 1 tablet 2 stroke recognition unit 3 stroke dictionary 4 stroke string register 5 partial pattern recognition unit 6 partial pattern dictionary 7 partial pattern detail check dictionary 8 overlapping dictionary 9 partial pattern sequence register 10 character recognition unit 11 character dictionary 12 character detailed check dictionary

 ─────────────────────────────────────────────────── ─── Continuation of the front page (56) References Japanese Patent Laid-Open No. 59-27381 (JP, A) Material of the Institute of Electronics and Communication Engineers, PRL73-20, (48-5-23), PP. 1-9 IEICE Technical Committee Material PRL73-22, (Sho 48-5-23), PP. 1-9 IEICE Technical Report PRL82-37, (Sho 57-10-15), PP. 49-56

Claims (3)

[Claims] 1. A tablet for extracting strokes of handwritten characters and sequentially outputting them as stroke information, a dictionary for storing characteristics of characters, and a recognition unit for comparing the stroke information with the contents of the dictionary for recognition. An online character recognition device for identifying handwritten characters, comprising: a stroke dictionary that stores stroke information; and a stroke recognition unit that identifies the stroke information of the handwritten characters output from the tablet using the stroke dictionary. , A partial pattern dictionary storing a partial pattern consisting of a number of combinations of strokes that maximizes a predetermined natural number n, and a partial pattern detail dictionary storing a positional relationship between strokes for which a combination of strokes can become a partial pattern, A duplicate dictionary in which multiple partial patterns consisting of the same or similar figures are stored as characters A first combination for extracting a partial pattern candidate within a range that does not exceed the natural number n strokes every time the stroke information is input by comparing a combination of the output information of the stroke recognizing unit continuously output with the partial pattern dictionary. A second process of selecting a partial pattern by processing and comparing the positional relationship of the output information of the stroke recognizing unit with the partial pattern detailed dictionary, excluding the partial pattern candidates that do not meet the conditions, and further extracting as a character A partial pattern recognition unit having a third process of selecting a partial pattern by comparing the generated partial pattern candidate with an overlapping dictionary; and a character dictionary storing a character consisting of a combination of partial patterns arranged in a predetermined code order, The combination of the output information of the partial pattern recognition unit is rearranged in a predetermined code order, and the character dictionary is compared to identify the characters. An online character recognition device, comprising: a character recognition unit for identifying, and identifying characters written on the tablet. 2. A tablet that extracts strokes of handwritten characters and sequentially outputs them as stroke information, a dictionary that stores the characteristics of characters, and a recognition unit that compares the stroke information with the contents of the dictionary for recognition. An online character recognition device for identifying handwritten characters, comprising: a stroke dictionary that stores stroke information; and a stroke recognition unit that identifies the stroke information of the handwritten characters output from the tablet using the stroke dictionary. , A partial pattern dictionary storing a partial pattern consisting of a number of combinations of strokes that maximizes a predetermined natural number n, and a partial pattern detail dictionary storing a positional relationship between strokes for which a combination of strokes can become a partial pattern, A duplicate dictionary in which multiple partial patterns consisting of the same or similar figures are stored as characters A first combination for extracting a partial pattern candidate within a range that does not exceed the natural number n strokes every time the stroke information is input by comparing a combination of the output information of the stroke recognizing unit continuously output with the partial pattern dictionary. A second process of selecting a partial pattern by processing and comparing the positional relationship of the output information of the stroke recognizing unit with the partial pattern detailed dictionary, excluding the partial pattern candidates that do not meet the conditions, and further extracting as a character Same as the partial pattern recognition unit that has the third process of selecting the partial patterns by comparing the partial pattern candidates that have been generated with the overlapping dictionary, and the character dictionary that stores the characters that are combinations of partial patterns arranged in a predetermined code order. The identification condition for a plurality of characters consisting of a combination of partial patterns is set to the number of character candidates for each concatenation code with the character dictionary. A character detail dictionary additionally stored, and a character recognition section for rearranging combinations of output information of the partial pattern recognition section in a predetermined code order and for identifying characters by comparing the character dictionary and the character detail dictionary. An online character recognition device characterized by identifying characters written on the tablet. 3. A tablet that extracts strokes of handwritten characters and sequentially outputs them as stroke information, a dictionary that stores the characteristics of characters, and a recognition unit that compares the stroke information with the contents of the dictionary for recognition. An online character recognition device for identifying handwritten characters, comprising: a stroke dictionary that stores stroke information; and a stroke recognition unit that identifies the stroke information of the handwritten characters output from the tablet using the stroke dictionary. , A partial pattern dictionary that stores partial patterns consisting of a combination of strokes that maximizes a predetermined natural number n, and similar partial patterns are registered as a group having branch numbers, so that a combination of strokes can become a partial pattern. Partial pattern detail dictionary that stores the positional relationship between strokes and the same or similar shape Each time the stroke information is input, the natural number n is compared each time the stroke information is input, by comparing the partial pattern dictionary with the combination of the overlapping dictionary in which a plurality of partial patterns including A first process of extracting a partial pattern candidate within a range that does not exceed a stroke, and a positional relationship between output information of the stroke recognition unit and the partial pattern detailed dictionary are compared to exclude one of the partial pattern candidates that does not match the condition. And a partial pattern recognition unit having a third process of selecting a partial pattern by comparing the partial pattern candidates extracted as characters with the overlapping dictionary and selecting a partial pattern. It relates to a character dictionary that stores characters that consist of combinations arranged in order and a plurality of characters that consist of combinations of the same partial pattern. A character detail dictionary storing identification conditions with the number of character candidates added for each concatenated code to the character dictionary, and a combination of the code and the character by rearranging a combination of output information of the partial pattern recognition unit in a predetermined code order. And a character recognition unit that identifies a character by comparing the branch number information included in the output information of the partial pattern recognition unit and the position information for each partial pattern with the character detailed dictionary while comparing the dictionary. An online character recognition device characterized by identifying characters written on the tablet.