JP2011123788A - Online character recognition device - Google Patents

Abstract

An on-line character recognition apparatus for distinguishing homomorphic and different characters is provided.
A kanji character group representing a kanji character is stored in the kanji dictionary, and a character group other than the kanji character group is stored in the kana character dictionary. The handwritten character input unit 5 inputs characters (input characters) written by the user, and the writing speed detection unit 21 detects the writing speed V _n-0 at that time. The writing speed storage unit 13 stores writing speeds V _n−m ,..., V _n−2 , V _{n−1 up} to m characters before. The calculation unit 22 weights the writing speeds V _nm ,..., V _n-2 , V _n-1 and the writing speed V _n-0 respectively, and calculates a calculated value _Thn that is an average value thereof. The weighting for the writing speed Vn _-0 is higher than the weighting for the writing speeds Vn _-m , ..., Vn _-2 , Vn _-1 . Dictionary selection unit 23, when the calculated value T _hn is lower than the threshold T _h, selects the Chinese character dictionary 3 as the selected dictionary. The character recognition unit 24 performs recognition processing with reference to the selected dictionary.
[Selection] Figure 1

Description

  The present invention relates to an on-line character recognition device that refers to a dictionary and recognizes a character that matches a character written by a user.

  In general, input devices such as a keyboard and a pointing device (for example, a mouse) have been widely used for inputting information to an information processing device using characters. On the other hand, in recent years, an on-line character recognition apparatus has been developed that captures similar information using a device (handwritten character input unit) that can input characters by handwriting. In order for the online character recognition apparatus to recognize handwritten characters, a dictionary storing a plurality of characters is prepared.

  For example, in an online character recognition device (conventional online character recognition device) described in Japanese Patent Application Laid-Open No. 9-106442, a plurality of dictionaries are prepared for each writing speed in order to improve the character recognition rate.

  Specifically, the conventional online character recognition apparatus includes a handwritten character input unit, a control unit, and a storage unit. The control unit includes a writing speed detection unit, a dictionary selection unit, and a character recognition unit. The storage unit includes a plurality of dictionaries. A plurality of dictionaries store a plurality of characters for each writing speed.

  The user writes a character by moving the pen or his / her finger with respect to the handwritten character input unit. At this time, the handwritten character input unit inputs characters written by the user as input characters. The writing speed detection unit detects the writing speed when the input character is input to the handwritten character input unit as the input writing speed. The dictionary selection unit selects a dictionary corresponding to the input writing speed as a selection dictionary from a plurality of dictionaries. The character recognition unit refers to the selected dictionary and recognizes a character that matches the input character.

JP-A-9-106442

  The conventional on-line character recognition device aims to prepare a plurality of dictionaries for each writing speed.

  However, there are important problems in recognizing characters that match the input characters.

  There are several characters, such as kanji “power”, katakana “ka”, kanji “Yu”, and katakana “ta”. That is, there are characters that are similar in shape to katakana (hereinafter referred to as isomorphic different characters).

  In this case, the conventional on-line character recognition device does not distinguish between isomorphic and different characters, so when the user writes the isomorphic characters in the handwritten character input unit, the homomorphic and different characters are erroneously recognized. there is a possibility.

  In the following, means for solving the problems will be described using the reference numerals used in the embodiments for carrying out the invention in parentheses. This symbol is added to clarify the correspondence between the description of the claims and the description of the mode for carrying out the invention, and the technical scope of the invention described in the claims. Must not be used to interpret

The on-line character recognition device of the present invention includes a kanji dictionary (3), a kana character dictionary (4), a handwritten character input unit (5), a writing speed detection unit (21), and a writing speed storage unit (13). , A calculation unit (22), a dictionary selection unit (23), and a character recognition unit (24). The kanji dictionary (3) stores a kanji character group representing kanji among a plurality of characters. The kana character dictionary (4) stores character groups other than the kanji character group among a plurality of characters. The handwritten character input unit (5) inputs characters written by the user as input characters. The writing speed detection unit (21) detects a writing speed (V _n-0 ) when an input character is input to the handwritten character input unit (5). The writing speed storage unit (13) has m writing speeds (V _nm ,..., V _n-2 , V _n-1 ) that are writing speeds up to m characters before (m is an integer of 1 or more). Is stored. The calculation unit (22) weights the _m writing speeds (Vn _−m ,..., Vn ₋₂ , Vn ₋₁ ) and the writing speeds (Vn ₋₀ ), and averages them. A certain calculated value (T _hn ) is calculated. Here, the weighting for the writing speed (V _n-0 ) is higher than the weighting for the m writing speeds (V _n−m ,..., V _n−2 , V _n−1 ). When the calculated value (T _hn ) is lower than the threshold value (T _h ), the dictionary selection unit (23) selects the Chinese character dictionary (3) as the selection dictionary, and the calculated value (T _hn ) is equal to or greater than the threshold value (T _h ). If so, the kana character dictionary (4) is selected as the selection dictionary. The character recognition unit (24) performs a recognition process for recognizing a character that matches the input character with reference to the selected dictionary.

Among the multiple characters, there are homomorphic characters whose shapes are similar between kanji and katakana. For this reason, a kanji character group representing a kanji character among a plurality of characters is stored in the kanji character dictionary (3), and character groups other than the kanji character group are stored in the kana character dictionary (4). However, when the user writes the above-mentioned isomorphic characters, the writing speed is almost the same. Therefore, in the on-line character recognition device of the present invention, the dictionary selection unit (23) determines whether or not the input character is a Chinese character, or determines whether or not the Chinese character is written continuously. Therefore, the weight for the writing speed (V _n-0 ) is set to be higher than the weight for the m writing speeds (V _n−m ,..., V _n−2 , V _n−1 ). In this case, m writing speeds (V _nm ,..., V _n-2 , V _n-1 ) and writing speeds (V _n-0 ) are weighted from m characters before the input character, respectively. It is preferable to set so as to increase in this order. Thereby, when the calculated value (T _hn ) is lower than the threshold value (T _h ), the input character may be a Chinese character. Alternatively, there is a possibility that the input characters are kanji and kanji are written continuously. Thus, in the on-line character recognition device of the present invention, the kanji character dictionary (3), the kana character dictionary (4), the writing speed, and the weight are used to distinguish homomorphic characters from kanji and katakana. When an isomorphic character is written in the character input section (5), the isomorphic character can be correctly recognized.

FIG. 1 shows a configuration of an on-line character recognition apparatus according to an embodiment of the present invention. FIG. 2 shows the Chinese character dictionary 3 of FIG. FIG. 3 shows the kana character dictionary 4 of FIG. FIG. 4 is a flowchart illustrating the operation of the online character recognition apparatus according to the embodiment of the present invention. FIG. 5 is a diagram for explaining a method of calculating the writing speed V _n-0 . FIG. 6 is a diagram for explaining a method of calculating the writing speed V _n-0 . FIG. 7 shows the writing speed storage unit 13 of FIG. 1 and is a diagram for explaining a calculation method of the calculated value _Thn .

  Hereinafter, an on-line character recognition apparatus according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

  FIG. 1 shows a configuration of an on-line character recognition apparatus according to an embodiment of the present invention. The online character recognition device according to the embodiment of the present invention includes a computer 1, a storage device 2, a handwritten character input unit 5, and an output device 6.

  The storage device 2 is connected to the computer 1 and includes a kanji dictionary 3 and a kana character dictionary 4.

  FIG. 2 shows the Chinese character dictionary 3. The kanji dictionary 3 stores a kanji character group representing kanji among a plurality of characters. As shown in Fig. 2, the kanji character group has a shape similar to that of the kanji character “power”, which is similar to katakana “ka”, and katakana “ta”. The Chinese character “Yu” is included. Although not shown in the drawing, the kanji character group further includes “2 (2)” which is similar in shape to katakana “ni” and “10 ( 10) "and the like.

  FIG. 3 shows the kana character dictionary 4. The kana character dictionary 4 stores character groups (hiragana, katakana, numbers, alphabets, etc.) other than the kanji character group among a plurality of characters. As shown in Fig. 3, the character group is shaped like a katakana "ka", which is similar in shape to the kanji "power", or a kanji "yu". Similar katakana “ta” is included. Although not shown, the character group further includes a katakana “ni” shape similar to the Chinese character “2 (2)” and a katakana “10” shape similar to the Chinese character “10 (10)”. Na "etc. are also included.

  The handwritten character input unit 5 is connected to the computer 1 and inputs characters written by the user as input characters. The handwritten character input unit 5 takes in the coordinates (X, Y) of the point pressed with the tip of the pen by operating the pen on the tablet. Thereby, a coordinate point sequence is input as data for each stroke. The handwritten character input unit 5 is exemplified by a touch panel that draws characters with a finger, a mouse that moves a cursor on a display device, and draws characters, in addition to those using a pen.

  The computer 1 refers to the kanji dictionary 3 or the kana character dictionary 4 and recognizes a character that matches the input character input to the handwritten character input unit 5 and outputs the recognition result to the output device 6.

  The output device 6 includes a display device and an information processing device connected to the computer 1. For example, the display device displays a character that matches the input character as the output of the recognition result, and the information processing device executes a predetermined process using the character that matches the input character.

  The computer 1 includes a CPU (Central Processing Unit) 10 that is an execution unit and a storage unit 11 that is a recording medium in which a computer program 12 to be executed by the CPU 10 is stored. The CPU 10 reads the computer program 12 from the storage unit 11 and executes it when the computer is started up.

  The computer program 12 includes a writing speed detection unit 21, a calculation unit 22, a dictionary selection unit 23, and a character recognition unit 24. These operations will be described later.

  The storage unit 11 includes a writing speed storage unit 13. Here, the kanji dictionary 3 and the kana character dictionary 4 may be built in the computer 1 instead of the storage device 2. That is, the storage unit 11 in the computer 1 may be provided. Further, the information processing apparatus may be built in the computer 1. That is, a computer program that is provided in the computer 1 and executes the operation thereof may be stored in the storage unit 11.

  FIG. 4 is a flowchart illustrating the operation of the online character recognition apparatus according to the embodiment of the present invention.

  A user writes a character by moving a pen or his / her finger with respect to the handwritten character input unit 5. At this time, the handwritten character input unit 5 inputs a character written by the user as an input character (step S1).

The writing speed detection unit 21 detects the writing speed V _n-0 when an input character is input to the handwritten character input unit 5 and stores it in the writing speed storage unit 13 (step S2).

A method for calculating the writing speed V _n-0 will be specifically described.

により表される。 As shown in FIG. 5, the number of coordinate points of the L stroke (L is an integer equal to or greater than 1) when an input character is input to the handwritten character input unit 5 is P (L), and the kth ( k is the coordinate point of an integer of 1 or more) _(X L _(k), and Y L (k)), when the writing speed of the L-stroke th was _{V L,} the writing speed _{V L} of the L stroke th

Is represented by

により表される。 As shown in FIG. 6, when the number of strokes when an input character is input to the handwritten character input unit 5 is SM and the writing speed is V _n-0 , the writing speed V _n-0 is

Is represented by

  At this time, it is assumed that the writing speed storage unit 13 does not store m writing speeds up to m characters before (m is an integer of 1 or more) by the writing speed detection unit 21. That is, when the user has started writing the first character, this number of characters has not been reached (step S3-NO).

  In this case, the dictionary selection unit 23 selects the Chinese character dictionary 3 (step S6). The character recognition unit 24 refers to the Chinese character dictionary 3. At this time, the character recognition unit 24 performs a recognition process of recognizing a character that matches the input character in the kanji character group stored in the kanji dictionary 3. In the recognition process, when the character recognition unit 24 recognizes a character that matches the input character (step S7—YES), the recognition result is output to the output device 6 (step S10).

  At this time, even if the character recognition unit 24 refers to the Chinese character dictionary 3, there may be a case where there is no character that matches the input character (step S7-NO). In this case, the character recognition unit 24 refers to the kana character dictionary 4 (step S8). At this time, the character recognition unit 24 performs a recognition process for recognizing a character that matches the input character in the character group stored in the kana character dictionary 4. In the recognition process, when the character recognition unit 24 recognizes a character that matches the input character (step S9), the recognition result is output to the output device 6 (step S10). If the character recognition unit 24 cannot recognize a character that matches the input character, the character recognition unit 24 recognizes a character similar to the input character (step S9), and outputs the recognition result to the output device 6 (step S10).

  When the user moves the pen or his / her finger with respect to the handwritten character input unit 5, steps S1 and S2 are executed again.

Now, as shown in FIG. 7, the writing speed storage unit 13 includes m writing speeds V _n−m ,..., V _n−2 , V _n−1 up to m characters before the writing speed detection unit 21. Is stored. That is, the number of writing speeds stored in the writing speed storage unit 13 has reached m (step S3-YES).

The calculation unit 22 weights the _m writing speeds V _nm ,..., V _n-2 , V _n-1 and the writing speed V _n-0 , respectively, and calculates a calculated value T _hn that is an average value thereof. Calculate (step S4).

  Here, the weighting will be specifically described.

  Some of the characters include kanji and katakana, such as the kanji “power”, the katakana “ka”, the kanji “yu”, and the katakana “ta”. There are characters that are similar in shape. Therefore, a kanji character group representing a kanji character among a plurality of characters is stored in the kanji character dictionary 3, and character groups other than the kanji character group (such as hiragana, katakana, numbers, and alphabets) are stored in the kana character dictionary 4. Yes. However, when the user writes the above-mentioned isomorphic characters, the writing speed is almost the same.

Therefore, the writing speed V _n-0 is determined by the dictionary selection unit 23 described later in order to determine whether or not the input character is a Chinese character, or to determine whether or not the Chinese character is written continuously. Is set so as to be higher than the weights for _m writing speeds V _nm ,..., V _n−2 , V _n−1 . In this case, the weights for the _m writing speeds V _nm ,..., V _n-2 , V _n-1 and the writing speed V _n-0 increase in this order from the m character preceding character to the input character. It is preferable to set in this way.

A method for calculating the calculated value _Thn will be specifically described.

により表される。 The m writing speeds stored in the writing speed storage unit 13 are input to the handwritten character input unit 5 as V _nm ,..., V _n-2 , V _{n-1 in} this order from m characters to 1 character before. the writing speed when the character is inputted as _{V n-0, (m +} 1) pieces of writing speed _{V n-m, ..., V} n-2, V n-1, V n-0 weighting for each _{a n -m, ...,} and _{a n-2, a n-} 1, a n-0, when the calculated value was _{T hn,} calculated value _{T hn} is

Is represented by

When the calculated value T _hn is less than the threshold value T _h (step S5-YES), the dictionary selection unit 23, whether the input character is a Chinese character, as being written Chinese characters are continuously, select the Chinese character dictionary 3 A dictionary is selected (step S6). The character recognition unit 24 refers to the kanji dictionary 3. At this time, the character recognition unit 24 performs a recognition process of recognizing a character that matches the input character in the kanji character group stored in the kanji dictionary 3. In the recognition process, when the character recognition unit 24 recognizes a character that matches the input character (step S7—YES), the result is output to the output device 6 (step S10).

On the other hand, when the calculated value _{T hn} is equal to or more than the threshold _{T h} (step S5-NO), dictionary selection unit 23 selects the Kana character dictionary 4 as selected dictionary (step S8). The character recognition unit 24 refers to the kana character dictionary 4. At this time, the character recognition unit 24 performs a recognition process for recognizing a character that matches the input character in the character group stored in the kana character dictionary 4. In the recognition process, when the character recognition unit 24 recognizes a character that matches the input character (step S9), the recognition result is output to the output device 6 (step S10). If the character recognition unit 24 cannot recognize a character that matches the input character, the character recognition unit 24 recognizes a character similar to the input character (step S9), and outputs the recognition result to the output device 6 (step S10).

  Moreover, even if the character recognition unit 24 refers to the Chinese character dictionary 3, there may be a case where there is no character that matches the input character (step S7-NO). In this case, the character recognition unit 24 refers to the kana character dictionary 4 (step S8), performs recognition processing (step S9), and outputs the recognition result to the output device 6 (step S10).

Specifically described when the calculated value T _hn is lower than the threshold T _h.

  First, a case where the input character is a Chinese character will be described.

For example, when m is 2, the characters corresponding to the writing speeds V _n-2 , V _n-1 , V _n-0 are “between”, and the user writes “between” kanji as input characters. If it is, the writing speed Vn _-0 is lower than the writing speeds Vn _-2 and Vn _-1 . That is, the value of the writing speed V _n-0 is smaller than the values of the writing speeds V _n−2 and V _n−1 . In this case, since the weights a _n−2 , a _n−1 and a _n−0 are higher in this order from the character two characters before the input character, the value of a _{n− 0} × V _n−0 is a _n−1. There is no big difference compared with the value of * Vn _-1 and an _-2 * Vn _-2 . Accordingly, the calculated value _{T hn} is so lower than the threshold _{T h,} weighting _{a n-2, a n-} 1, a n-0 and threshold _{T h} is set.

  Next, a case where the input characters are kanji and the kanji are written continuously will be described.

For example, when m is 2, the characters corresponding to the writing speeds V _n-2 , V _n-1 , V _n-0 are “instantaneous power”, and the user inputs “kanji” In the case of writing "Ryo)", the writing speed Vn _-0 is higher than the writing speeds Vn _-2 and Vn _-1 . That is, the value of the writing speed Vn _-0 is larger than the values of the writing speeds Vn _-2 and Vn _-1 . In this case, since the weights a _n−2 , a _n−1 and a _n−0 are higher in this order from the character two characters before the input character, the value of a _{n− 0} × V _n−0 is a _n−1. XV _n-1 , an _{n-2 It} becomes larger than the value of V _n-2 , but the value of an _n-1 × V _n-1 , an _n-2 × V _n-2 itself is an _n−. Very small compared to the value of ₀ × V _n−0 . Accordingly, the calculated value _{T hn} is so lower than the threshold _{T h,} weighting _{a n-2, a n-} 1, a n-0 and threshold _{T h} is set.

  Next, a case where the input character is katakana and katakana is continuously written will be described.

For example, m is 2, and the letters corresponding to the writing speeds V _n-2 , V _n-1 , V _n-0 of “Writing in Japanese with Katakana” are “Nahaka” and used. When a person writes “Ka” of katakana as an input character, the writing speed V _n-0 does not change significantly compared to the writing speeds V _n−2 and V _n−1 . That is, the value of the writing speed V _n-0 does not change significantly compared to the values of the writing speeds V _n−2 and V _n−1 . In this case, since the weights a _n−2 , a _n−1 and a _n−0 are higher in this order from the character two characters before the input character, the value of a _{n− 0} × V _n−0 is a _n−1. × V _n-1 , a _n-2 × V _n-2 is larger than the value. Accordingly, the calculated value _{T hn} is such that more than the threshold _{T h,} weighting _{a n-2, a n-} 1, a n-0 and threshold _{T h} is set.

Thus, there are many 2 or 3 kanji characters, and m is preferably 1 to 3 in consideration of four-character idioms. Therefore, it is preferable that the weights a _n−m ,..., A _n−2 , a _n−1 , a _n−0 are set so as to increase in this order from the m character before the input character. As described above, when m is 2, the weights are a _n−2 , a _n−1 , and a _n−0 . For example, when a value of 10 is assigned to three values, the weights a _n−2 , a _n−1 and a _n−0 are set higher in this order from the character two characters before the input character, It is preferable that values such as 5, 1, 1, 3, 6 are set.

  As described above, in the on-line character recognition apparatus of the present invention, homomorphic characters are distinguished from kanji and katakana by the kanji dictionary 3, the kana character dictionary 4, the writing speed, and the weighting. For this reason, when a user writes an isomorphic character in the handwritten character input unit 5, the isomorphic character can be recognized correctly.

The reason for this is that, as described above, among the plurality of characters, there are isomorphic characters whose shapes are similar between kanji and katakana. For this reason, a kanji character group representing a kanji character among a plurality of characters is stored in the kanji character dictionary 3, and a character group other than the kanji character group is stored in the kana character dictionary 4. However, when the user writes the above-mentioned isomorphic characters, the writing speed is almost the same. Therefore, in the online character recognition device of the present invention, the dictionary selection unit 23 determines whether or not the input character is a Chinese character, or determines whether or not the Chinese character is written continuously. The weighting for the writing speed Vn _-0 is set to be higher than the weighting for the m writing speeds Vn _-m , ..., Vn _-2 , Vn _-1 . In this case, the weights for the _m writing speeds V _nm ,..., V _n-2 , V _n-1 and the writing speed V _n-0 increase in this order from the m character preceding character to the input character. It is set as follows. Thus, when the calculated value T _hn is lower than the threshold T _h, there is a possibility input character is Kanji. Alternatively, there is a possibility that the input characters are kanji and kanji are written continuously. Thus, in the on-line character recognition apparatus of the present invention, the kanji character dictionary 3, the kana character dictionary 4, the writing speed, and the weighting are used to distinguish homomorphic different characters from kanji and katakana. When an isomorphic character is written on the same character, it can be recognized correctly.

  Note that the on-line character recognition device according to the embodiment of the present invention may further include a configuration of a conventional on-line character recognition device as long as the effect of the present invention is realized.

1 computer,
2 storage devices,
3 Kanji dictionary,
4 Kana character dictionary,
5 Handwritten input part,
6 output device,
10 CPU,
11 storage unit,
12 computer programs,
13 Writing speed storage,
21 Writing speed detector,
22 calculation unit,
23 Dictionary selection part,
24 character recognition part

Claims (9)

A kanji dictionary in which a kanji character group representing a kanji among a plurality of characters is stored;
A character dictionary in which a character group other than the Kanji character group among the plurality of characters is stored,
A handwritten character input unit for inputting characters written by the user as input characters;
A writing speed detection unit that detects a writing speed when the input character is input to the handwritten character input unit;
a writing speed storage unit for storing m writing speeds that are writing speeds up to m characters before (m is an integer of 1 or more);
Weighting each of the m writing speeds and the writing speed and calculating a calculation value that is an average value thereof, wherein the weighting for the writing speed is based on the weighting for the m writing speeds Higher
A dictionary selection unit that selects the kanji dictionary as a selection dictionary when the calculated value is lower than a threshold; and a dictionary selection unit that selects the kana character dictionary as the selection dictionary when the calculated value is equal to or greater than the threshold;
An online character recognition device comprising: a character recognition unit that performs a recognition process for recognizing a character that matches the input character with reference to the selection dictionary. 2. The online character recognition device according to claim 1, wherein the m writing speeds and the writing speeds are weighted in this order from a character preceding m characters to the input character. The m writing speeds stored in the writing speed storage unit are set to V _nm ,..., V _n-2 , V _{n-1 in} this order from m characters before to 1 character, and the handwritten character input unit The writing speed when the input character is input is set to V _n-0, and weighting is applied to (m + 1) writing speeds V _nm , ..., V _n-2 , V _n-1 , V _n-0 . When each of a _nm ,..., A _n-2 , a _n-1 , a _n-0 and the calculated value is _Thn , the calculated value _Thn is

The on-line character recognition device according to claim 2 represented by: The number of coordinate points of the Lth stroke (L is an integer of 1 or more) when the input character is input to the handwritten character input unit is P (L), and the kth of the L stroke (k is an integer of 1 or more). the coordinate point and _{(X L (k), Y} L (k)), when the writing speed of the L-stroke th was _{V L,} the writing speed _{V L} of the L stroke th

Represented by
When the stroke number when the input character is input to the handwritten character input unit is SM and the writing speed is Vn _-0 , the writing speed Vn _-0 is:

The on-line character recognition device according to claim 1, represented by: When the m writing speeds up to m characters before are not stored in the writing speed storage unit, the dictionary selection unit selects the Kanji dictionary,
The character recognition unit performs the recognition process with reference to the Chinese character dictionary,
The on-line character recognition device according to any one of claims 1 to 4, wherein if the character that matches the input character does not exist even if the kanji dictionary is referenced, the recognition process is performed with reference to the kana character dictionary. A computer connected to a handwritten character input unit for inputting a character written by a user as an input character, and having a storage unit;
The m writing speeds up to m characters before (m is an integer of 1 or more) are stored in the storage unit, and the writing speed when the input characters are input to the handwritten character input unit is detected. A writing speed detection unit,
Weighting each of the m writing speeds and the writing speed and calculating a calculation value that is an average value thereof, wherein the weighting for the writing speed is based on the weighting for the m writing speeds Higher
When the calculated value is lower than the threshold, a kanji dictionary storing a kanji character group representing a kanji is selected as a selection dictionary, and when the calculated value is equal to or greater than the threshold, A dictionary selection unit that selects, as the selection dictionary, a character dictionary in which characters other than the Kanji character group are stored,
A computer comprising: a character recognition unit that performs a recognition process for recognizing a character that matches the input character with reference to the selection dictionary. The computer according to claim 6, wherein the kanji dictionary and the kana character dictionary are built in the storage unit. A method executed by a computer connected to a handwritten character input unit for inputting characters written by a user as input characters,
storing m writing speeds, which are writing speeds up to m characters before (m is an integer of 1 or more), in the storage unit;
Detecting a writing speed when the input character is input to the handwritten character input unit;
Weighting each of the m writing speeds and the writing speed and calculating a calculated value that is an average value thereof, wherein the weighting for the writing speed is more than the weighting for the m writing speeds high,
If the calculated value is lower than the threshold, a step of selecting, as a selection dictionary, a kanji dictionary storing a kanji character group representing a kanji among a plurality of characters;
When the calculated value is equal to or greater than the threshold, a character dictionary in which a character group other than the kanji character group among the plurality of characters is stored is selected as the selection dictionary;
An online character recognition method comprising: recognizing a character that matches the input character with reference to the selected dictionary.   The computer program which makes the said computer perform each step of the online character recognition method of Claim 8.

Images