JP2639314B2 - Character recognition method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition system, and more particularly, to a method for determining different character types having the same or similar form (hereinafter referred to as similar characters) by weighting rules for weighting the character types. Related to character recognition method.

[0002]

2. Description of the Related Art In a conventional character recognition system, similar characters, for example, mouth (kanji) and b (katakana) or 1 (number) and l
It is difficult to distinguish similar characters such as (English letters)
Various methods are used to determine similar characters.

For example, a character type determination table is prepared, and it is checked whether or not the input character is registered as a similar character. If the input character is registered as a similar character, the character is identified using the character type determination table. (See JP-A-61-39175).

[0004]

In the above-described conventional character recognition method, a character type determination table is prepared for determining similar characters. Therefore, if similar character data must be registered in advance. In addition, there is a problem that similar characters other than the registered data cannot be determined.

SUMMARY OF THE INVENTION In view of the above, it is an object of the present invention to provide a character recognition system which can distinguish similar characters without preparing a character type determination table and registering similar character data. is there.

[0006]

SUMMARY OF THE INVENTION A character recognition system of the present invention is a character recognition system for recognizing an input character and converting it into a sentence mixed with kanji and kana, and an input means for inputting a character and an input by the input means. A character recognition unit that converts the converted character into a conversion character candidate code with reference to a character recognition dictionary, and converts the converted character candidate code converted by the character recognition unit into katakana, hiragana,
Similar characters that assign character type weighting attributes of Kanji type, number type and English character type, assign a combination weighting attribute for each three characters, and determine similar characters based on the sum of the weighting attribute that is the sum of the character type weighting attribute and the combination weighting attribute A determination unit.

[0007]

Next, the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of a character recognition system according to one embodiment of the present invention. The character recognition method according to the present embodiment includes a tablet 1, a character recognition dictionary 2, a character recognition unit 3, a weight development table 4, a weight rule file 5, and a similar character determination unit 6.

The tablet 1 is input means for inputting characters by handwriting.

The character recognition dictionary 2 is a dictionary in which characteristic information of individual characters is stored.

The character recognizing section 3 recognizes the character by comparing the character form information from the tablet 1 with the character feature information of the character recognition dictionary 2 and converts the character into one or more converted character candidate codes. This is a means for storing in the weighting development table 4.

As shown in FIG. 2, the weighting rule file 5 is a file for storing weighting rules for weighting character types from converted character candidate codes. In FIG. 2, an area a is an area for registering a rule number, and an area b is an area for describing the contents of a weighting rule.

As shown in FIG. 3, the weighting development table 4 is a table on a memory for developing weighting attributes. In FIG. 3, an area A is an area for registering the input character order of the characters input by handwriting on the tablet 1, an area B is an area for registering the processing character order when the similar character determination unit 6 processes a combination for every three characters, The area C is an area for registering the conversion candidate order of the converted character candidate code converted by the character recognizing unit 3, the area D is an area for registering the converted character candidate code, and the area F is a similar character discriminating unit 5 for converting the processing character order. The area for registering the combination weighting attribute generated when the weighting attribute of the character candidate code is processed by a combination of three characters, the area is the sum of the weighting attribute that is the sum of the character type weighting attribute and the combination weighting attribute of each converted character candidate code. The area to be registered is an area in which the order of each converted character candidate code is registered as a weighting order of 1, 2, 3,... That.

The similar character discriminating unit 6 converts the converted character candidate codes converted by the character recognizing unit 3 into katakana and hiragana based on the weighting rules of the weighting rule file 5.
A means for assigning character type weighting attributes of kanji type, number type and alphabetic type, assigning a combination weighting attribute for each of the three characters, and determining similar characters based on the sum of the weighting attribute which is the sum of the character type weighting attribute and the combination weighting attribute. is there.

Referring to FIG. 6, the processing performed by the similar character discriminating unit 6 includes a character type weighting attribute assigning step 61, an input character order n less than 3 determination step 62, a processing character order assigning step 63, and a character type weighting attribute comparison. Step S
64 and the same character type weighting attribute determining step S65
, A combination weighting attribute assigning step S66, a weight appending total assigning step 67, a weighting order assigning step S68, and a processing character order erasing step 69.

Next, the operation of the character recognition system according to the present embodiment configured as described above will be described.

Each time a handwritten character is input from the tablet 1, the character recognition unit 3 assigns an input character sequence n (n is an integer of 1 or more) in ascending order from 1, and assigns the input character to the character recognition dictionary 2 Is converted to one or more converted character candidate codes using a known character recognition algorithm such as a feature matching method, and the converted character candidate codes are assigned the conversion candidate order and stored in the weighting expansion table 4.

Each time the character recognizing unit 3 converts a character into a converted character candidate code and stores it in the weighting expansion table 4, the similar character discriminating unit 6 is activated with the input character order n as an argument.

First, the similar character discriminating unit 6 assigns a character type weighting attribute to each converted character candidate code in the input character order n of the weighting development table 4 based on the weighting rule of the weighting rule file 5 (step 61).

Next, the similar character discriminating section 6 judges whether or not the input character order n is less than 3 (step 62).

If the input character sequence n is 3 or more, the similar character discriminating unit 6 processes the input character sequence (n−2), (n−1) and n in the weighted development table 4 into the processing character sequence 1, 2 and 3 are given (step 63).

Next, the similar character discriminating unit 6 compares the character type weighting attribute of the converted character candidate code of the processing character order 3 with the character type weighting attribute of the converted character candidate code of the processing character order 1 and 2 (step S1). 64) It is determined whether or not there is a combination having the same character type weighting attribute of the converted character candidate codes in the processing character order 1, 2, and 3 (step 65). If there is no combination of the same character type weighting attributes, the similar character determination unit 6 shifts the control to step 67.

If there is the same combination of the character type weighting attributes of the converted character candidate codes in the processing character order 1, 2, and 3, the similar character discriminating unit 6 determines the combination of the converted character candidate codes having the same combination of the character type weighting attributes. One is assigned to each weighting attribute (step 66).

Next, the similar character discriminating unit 6 gives the sum of the character type weighting attribute and the combination weighting attribute of each converted character candidate code to the total weighting attribute (step 6).
7).

Subsequently, the similar character discriminating unit 6 assigns weighting orders 1, 2, 3,... To the converted character candidate codes in descending order of the value of the total weighting attribute (step 68).

Finally, the similar character discriminating unit 6 deletes the processing character order 1, 2, 3 (step 69), and ends the processing.

Next, the operation of the character recognition system of this embodiment will be described.
Referring to FIG. 4 and FIG.
A specific description will be given using an example of inputting “ri”, “f”, and “de”.

First, when the character "A" is input, the character recognizing unit 3 generates a conversion character candidate code of "A" and assigns 1 to the input character sequence in the weighting expansion table 4 and 1 to the conversion candidate sequence. The code of “A” is registered as the conversion character candidate code, and 3 is registered as the katakana type of the character type weighting attribute of the character type weighting attribute of the weighting expansion table 4 by the similar character identification unit 6 (see the column of input character order 1 in FIG. 4). ).

Next, when the character "me" is input, the character recognizing section 3 generates converted character candidate codes of "me (katakana)", "x (alphabet)" and "x (symbol)". In the weighting expansion table 4, the input character order is 2, the conversion candidate order is 1, 2 and 3, and the conversion character candidate codes are "me (katakana)", "x (alphabet)" and "x (symbol)". Each is registered, and the similar character discriminating unit 6 registers 3 for the katakana type, 3 for the alphabet type, and 0 for all others in the character type weighting attribute of the weight development table 4 (see the column of input character order 2 in FIG. 4). ).

Subsequently, when the character “R” is input, the character recognition unit 3 generates converted character candidate codes of “Ri (Hiragana)” and “Ri (Katakana)” and inputs the converted character candidate code to the weighting expansion table 4. The character order is 3, the conversion candidate order is 1 and 2, and the conversion character candidate code is “Ri (Hiragana)” and “Ri (Katakana)” are registered. As the character type weighting attribute, 3 is registered as the hiragana type, and 3 is registered as the katakana type (see the column of input character order 3 in FIG. 4).

Next, the similar character discriminating unit 6 converts the character type weighting attributes of the conversion character candidate codes “ri (katakana)” and “ri (hiragana)” in the processing character order 3 and the conversion character candidate code in the processing character order 1 Character type weighting attribute of "A (Katakana)" and converted character candidate codes "Me (Katakana)", "x (English character)" and "X"
(Symbol) ”is compared with the character type weighting attribute, and the combination of three characters having the same character type weighting attribute is the katakana type, so the conversion character candidate codes“ a (katakana) ”,
“1” is assigned to each of the combination weighting attributes “me (Katakana)” and “ri (Katakana)” (see the first row of the combination weighting attribute in FIG. 5).

Next, when the character "f" is input, similar processing is performed by the character recognizing unit 3 and the similar character discriminating unit 6, and the converted character candidate codes "me (katakana)" and "ri (katakana)""1" is assigned to each of the combination weighting attributes "" and "ka" (see the second row of the combination weighting attribute in FIG. 5). Thereby, the conversion character candidate codes “me (Katakana)” and “x (English character)” in the input character sequence 2
And the value of the sum of the weighting attributes of “× (symbol)” increases in this order, and the weighting order is 1, 2, and 3, respectively.
(See the column of input character sequence 2 in FIG. 5).

Subsequently, when the character "de" is input, similar processing is performed by the character recognizing unit 3 and the similar character discriminating unit 6, and the conversion character candidate codes "ri (hiragana)" and "3" in the input character order 3 are input. The value of the total weighting attribute of “ri (Katakana)” decreases in this order, and the weighting order 2
And 1 are given (see the column of input character order 3 in FIG. 5).

Further, the total value of the weighting attributes of “ka (katakana)” and “power (kanji)” of the conversion character candidate code in the input character order 4 becomes larger in this order, and weighting orders 1 and 2 are given, respectively. (See the column of input character sequence 4 in FIG. 5).

As a result, the weighting order of the katakana type for the characters "me", "ri" and "ka" in which a plurality of converted character candidate codes have been generated becomes 1, and "a", "a",
Similar characters such as "me", "ri", "ka" and "de" are correctly determined.

In the above-described embodiment, the case of handwriting input using a tablet as the input means 1 has been described as an example. However, the similar character determination process of the similar character determination unit 6 shown in FIG. As an input means, OCR (Optical Character
It will be apparent that the present invention can be similarly applied to the case where (r Reader) is used.

[0037]

As described above, the present invention assigns character type weighting attributes of katakana type, hiragana type, kanji type, numeric type and alphabet type to the converted character candidate code based on the weighting rule, and provides By assigning the combination weighting attribute and determining the similar character by the total weighting attribute, the similar character is determined without registering the similar character data in the character type determination table in order to determine the similar character. There is an effect that can be.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a character recognition system according to one embodiment of the present invention.

FIG. 2 is a diagram illustrating the contents of a weighting rule file in FIG. 1;

FIG. 3 is a diagram showing the contents of a weighting development table in FIG. 1;

FIG. 4 is a state transition diagram of a weight development table of FIG. 3;

FIG. 5 is a state transition diagram of the weighting development table of FIG. 4;

FIG. 6 is a flowchart showing processing of a similar character discriminating unit in FIG. 1;

[Explanation of symbols]

 1 Tablet 2 Character Recognition Dictionary 3 Character Recognition Unit 4 Weight Expansion Table 5 Weighting Rule File 6 Similar Character Discrimination Unit

Claims (4)

(57) [Claims] 1. A character recognition system for recognizing an input character and converting the character into a sentence mixed with kanji and kana, comprising: input means for inputting a character; and converting the character input by the input means with reference to a character recognition dictionary. A character recognition unit for converting to a character candidate code, and character type weighting attributes of katakana type, hiragana type, kanji type, numeric type and alphabet type are assigned to the converted character candidate code converted by the character recognition unit based on a weighting rule. And a similar character discriminating unit for assigning a combination weighting attribute for each of the three characters and discriminating similar characters based on a weighting attribute sum that is a sum of the character type weighting attribute and the combination weighting attribute. 2. An area for registering an input character order, an area for registering a processing character order, an area for registering a conversion candidate order, an area for registering a conversion character candidate code, and a combination weighting table. 2. The character recognition method according to claim 1, comprising an area for registering an attribute, an area for registering a total of weighted attributes, and an area for registering a weighting order. 3. The character recognition system according to claim 1, wherein said input means is a tablet. 4. The input means comprises an OCR.
The character recognition method described.