JPH08180137A - Character input device - Google Patents

Abstract

PURPOSE: To provide a character input device which outputs the most likelihood recognition result as Japanese and can improve a character recognition rate by utilizing general trends of Japanese. CONSTITUTION: A control means 103 executes a program stored in a storage means 102 to output character codes indicating >=1 character that is similar to writing data inputted by a coordinate input means 101 and further outputs one of the character codes as a recognition result. At this time, a character code indicating a character of the same kind as the kind of a character that a character code outputted as a last recognition result indicates is outputted preferentially as the recognition result by utilizing the general trend of Japanese that characters of the same kind appear successively.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character input device,
In particular, the present invention relates to a character input device that converts handwritten data of handwritten data into a character code and displays a character represented by the converted character code.

[0002]

2. Description of the Related Art In recent years, along with the miniaturization of computers, information processing apparatuses called "pen computers" which emphasize portability have become widespread. In general, pen computers
Since it is assumed to be carried around, the pen and tablet are used as input means, and the operator can perform input operations on the tablet while holding the main body with one hand and the pen with the other hand. it can. Since the pen computer does not require a fixed base unlike a keyboard and a mouse, it is suitable for operation while moving or standing.

As described above, when the operator inputs a character on the desktop computer, the operator inputs the character on the keyboard, but on the pen computer, the operator manually writes the character on the tablet.

The writing data input in this way is
By performing the character recognition process, the character code is converted into a character code, and the character represented by the converted character code is displayed.

The character recognition processing refers to a character recognition dictionary that stores, for each character, a standard writing pattern that is a standard writing pattern of a character and a character code that represents the character. This is a process that calculates the degree of similarity indicating the degree of similarity to the standard writing pattern and outputs the character code corresponding to the standard writing pattern for which the calculated degree of matching is greater than a predetermined value as the recognition result. The character code that represents the specified character is not always output as the recognition result.

For example, the characters "ka" and "power" are very similar, and the characters "1" (kanji) and "-(long sound)" are very similar. Are very similar, it is difficult to clearly distinguish which character's writing data was input, and it may not be possible to output the character code representing the character intended by the operator as the recognition result. is there. At this time,
The operator has to perform a recognition result correction operation for correcting the recognition result, which is an operational burden.

In order to reduce the burden on the operator, it is important to improve the character recognition rate of character recognition processing (the rate at which a correct recognition result can be output).
A character recognition rate of 95% can be obtained.

However, in the conventional character recognition processing, since the degree of coincidence of the writing data of each character is obtained for each character, any character that is similar to the above-mentioned example will be recognized. It was still difficult to clearly distinguish whether or not the writing data of characters was input.

Therefore, in addition to the character recognition processing, word processing is performed to automatically correct the recognition result by referring to the previous and next recognition results on the assumption that the recognition result is a Japanese sentence. can do.

The word processing is, for example, "Ka (Katakana)".
"Writing power" (Kanji) has been entered or not
In order to distinguish whether or not the writing data of the character was input, the previous recognition result was referenced, and if the previous recognition result was the character code representing the character "heavy", the word "gravity" was used. It presumes that the writing data of the constituent characters seems to have been input, and preferentially outputs the character code indicating the character "Kanji" rather than the character code indicating the character "Ka (katakana)" as the recognition result. Is. In addition, in order to perform word processing, it is necessary to provide a word dictionary that stores, for each word existing in Japanese, a character code that represents a character forming the word.

Furthermore, by adding information indicating the connection relation between words to the word dictionary, it is possible to perform Japanese processing so as to check the possibility that the words are connected.

For details of word processing and Japanese language processing, see Japanese Patent Application Laid-Open Nos. 1-231185 and 2-124.
See the '81 publication.

[0013]

By the way, in general, in character recognition processing, first, about 10 character codes are first output as recognition candidates for writing data of one character, and one of them is output as a recognition candidate. One is output as a recognition result. Therefore, in order to determine which recognition candidate is to be output as the recognition result for the writing data of one character, word processing and Japanese processing must be performed on all combinations of the recognition candidates. For example, 10 × 10 combinations of recognition candidates occur for writing data of two characters, and 10 × 10 × 10 combinations of recognition candidates occur for writing data of three characters. In order to perform word processing and Japanese processing for the combination of recognition candidates of
It requires a great deal of processing time and is practically difficult at the processing speed of a small information processing apparatus such as a pen computer.

Therefore, an object of the present invention is to output a recognition result that is likely to be Japanese by utilizing the general tendency of Japanese without performing word processing and Japanese processing.
It is to provide a character input device capable of improving the character recognition rate.

[0015]

As a general tendency of Japanese used in the present invention, firstly, in Japanese,
The tendency is that characters of the same character type often appear consecutively for several characters. For example, katakana and English letters are often used as nouns (foreign words, proper nouns, etc.) in sentences, and the number is 2
Many last more than one digit.

As a general tendency of Japanese used in the present invention, secondly, there are hundreds of thousands of words in Japanese, but the words used in daily life are at most It is about several hundred words, and there is a tendency that a specific word often appears repeatedly. In particular, the second tendency is remarkable in one sentence with a narrowed theme.

Therefore, the present invention utilizes the first tendency described above, and of the character codes output by the character recognition processing, the same character type of the character represented by the character code output as the previous recognition result. Character codes that are character types are preferentially output as recognition results.

Further, according to the present invention, by utilizing the second tendency, a plurality of consecutive character codes that have been output as a recognition result represent a plurality of characters which are elements constituting one word. It is regarded as a character code and stored, and thereafter, the recognition result is output based on the stored content. However, it does not store all of the two consecutive character codes that have been output as recognition results. For example, a character code that represents two consecutive kanji characters, two consecutive katakana characters, and two consecutive alphabetic characters may be stored. I try to remember.

Specifically, the present invention provides a character input device utilizing the above first tendency, (1) handwriting input means for inputting handwritten data of handwritten characters, and (2) for each character, A standard writing pattern that is a standard writing pattern of characters, a character code that represents the character, and a character recognition dictionary storage unit that stores the character type of the character, (3) Writing data input by the handwriting input unit. The degree of similarity indicating the degree of similarity to the standard writing pattern is obtained, and the character code and the character type corresponding to the standard writing pattern for which the obtained degree of matching is larger than a predetermined value, together with the obtained degree of matching, are recognized. Character recognition means for outputting as information, (4)
Character recognition information storage means for storing the character recognition information output by the character recognition means, (5) For each character type, the probability that the character that appears next to the character of the character type will be Indicating that the character recognition information storage means stores a predetermined probability value, and (6) the character recognition information storage means stores in accordance with the probability value stored in the character connection information storage means. For each character recognition information being calculated, an addition point to be added to the degree of coincidence indicated by the character recognition information is calculated, and the character recognition information having the largest total value of the degree of coincidence and the calculated addition point is used as the recognition result. (7) Display means for displaying the character represented by the character code indicated by the character recognition information output as the recognition result by the handwriting input means, and the added point calculation means for outputting (7) the handwriting input means. Providing a character input device equipped with It is.

The probability value stored in the character type concatenation information storage means is a probability value in which the character type of the character that appears next is the same as the character type of the previous character that appears next. The character type of the character is set to be larger than the probability value that the character type of the previous character is different from the character type of the previous character, and the addition point calculation means uses the probability value stored in the character type connection information storage means. Accordingly, when calculating the addition point for the character recognition information, the probability value corresponding to the character recognition information is determined from the character type indicated by the character recognition information and the character type indicated by the character recognition information output as the previous recognition result. It is preferable that the larger the calculated probability value, the larger the addition point for the character recognition information.

Further, how many character recognition information including the same character type as the character type indicated by the character recognition information output as the recognition result by the addition point calculation means have been continuously output as the recognition result until then. It is possible to further include a continuous character type number calculating means for calculating the number of continuous character types shown, and in this case, the addition point calculating means further calculates the continuous character type number calculating means. The larger the number of continuous character types, the smaller the addition point for the character recognition information including the same character type as the character type indicated by the character recognition information output as the previous recognition result.

On the other hand, the present invention is, as a character input device utilizing the above second tendency, (1) handwriting input means for inputting handwritten data of handwritten characters, (2) for each character,
A standard writing pattern that is a standard writing pattern of the character, a character code that represents the character, and a character recognition dictionary storage unit that stores the character type of the character, (3) Writing input by the handwriting input unit The degree of coincidence indicating the degree to which the data is similar to the standard writing pattern is obtained, and the degree of coincidence obtained is a character code and character type corresponding to the standard writing pattern larger than a predetermined value, together with the obtained degree of coincidence.
Character recognition means for outputting as character recognition information, (4) Character recognition information storage means for storing the character recognition information output by the character recognition means, and (5) Characters representing a plurality of characters constituting an element of one word. A concatenated character information storage means for storing concatenated character information in the form of a code, (6) based on the concatenated character information stored in the concatenated character information storage means,
For each character recognition information stored in the character recognition information storage means, an addition point to be added to the degree of coincidence indicated by the character recognition information is calculated, and the sum of the degree of coincidence and the calculated addition point is the largest. The character recognition information is output as a recognition result by an addition point calculation means, (7) written by the handwriting input means, and a character code indicated by the character recognition information output by the addition point calculation means as a recognition result. There is provided a character input device provided with a display unit for displaying the displayed characters.

If a plurality of character recognition information indicating the same predetermined character type is continuously output as the recognition result, the concatenated character information storage means stores the characters indicated by the plurality of character recognition information. It is assumed that the character represented by the code is a plurality of characters constituting one word, and a set of character codes indicated by the plurality of character recognition information is stored as the concatenated character information. I have to.

In addition to the above configuration, character recognition information including a character code representing a character displayed by the display means,
The display means may further include a reception means for receiving an instruction to replace the character recognition information other than the character recognition information. In this case, the display means indicates that the reception means replaces the character recognition information. When the instruction is accepted, the displayed character is changed to the character represented by the character code indicated by the character recognition information instructed to be replaced, and displayed, and the concatenated character information storage means is also provided. If a plurality of character recognition information indicating the same predetermined character type is continuously output as a recognition result when the above-mentioned receiving means receives an instruction to replace the plurality of character recognition information, Considering that the character represented by the character code indicated by is a plurality of characters that form one word, and combining the character codes indicated by the plurality of character recognition information, It can be configured to store a contact character information.

The addition point calculating means calculates the addition point for the character recognition information based on the concatenated character information stored in the concatenated character information storage means, and the character code indicated by the character recognition information and the last time. The addition point for the character recognition information in which the concatenated character information including a combination with the character code indicated by the character recognition information output as the recognition result is stored in the concatenated character information storage means is a predetermined value. You can do so.

In the present specification, the term "character" includes a symbol.

[0027]

As described above, the degree of coincidence is a value obtained as a result of performing character recognition processing on the writing data on a character-by-character basis, and when the character recognition information having the highest degree of coincidence is output as a recognition result, the context is not significant. Since the recognition result that is not referred to is output, the recognition result as intended by the operator may not be output for the writing data of the characters that are difficult to distinguish. Therefore, in the present invention, the character recognition information having the largest total value of the addition point and the degree of coincidence calculated by referring to the content of the character recognition information output as the previous recognition result is output as the recognition result. .

Of the contents of the character recognition information output as the previous recognition result, the contents referred to when calculating the addition point for the character recognition information is the character type when the above first tendency is used. The character code is used when the second tendency described above is used.

Specifically, in the character input device using the first tendency, the probability value stored in the character type connection information storage means appears next to the character of the character type for each character type. It is a predetermined value based on the result of statistically investigating what kind of character the character type is, in particular, the character type of the next character will be the same as the character type of the previous character. The probability value is set to be larger than the probability value that the character type of the character that appears next is a character type that is different from the character type of the previous character.

Therefore, the addition point for the character recognition information calculated by the addition point calculation means according to the probability value is the character recognition including the same character type as the character type indicated by the character recognition information output as the previous recognition result. The addition point for information is larger than the addition point for character recognition information including a character type different from the character type indicated by the character recognition information output as the previous recognition result.

Further, when the addition point calculation means calculates the addition point of the character recognition information, the character recognition information having the largest sum of the degree of coincidence indicated by the character recognition information and the calculated addition point is output as the recognition result. Therefore, the character recognition information including the same character type as the character type indicated by the character recognition information output as the previous recognition result is preferentially output as the recognition result.

Therefore, it is possible to output a recognition result that is likely to be Japanese and improve the character recognition rate.

Further, in detail, in the character input device utilizing the second tendency, the concatenated character information stored in the concatenated character information storage means includes a plurality of characters which are elements constituting one word. It is a set of character codes to represent.

Therefore, the addition point for the character recognition information calculated by the addition point calculation means based on the concatenated character information is indicated by the character code indicated by the character recognition information and the character recognition information output as the previous recognition result. Only the addition points of the character recognition information stored in the concatenated character information storage unit of the concatenated character information including a combination of the character code and the character code have a predetermined value.

Further, when the addition point calculation means calculates the addition point for the character recognition information, the character recognition information having the largest total value of the matching degree indicated by the character recognition information and the calculated addition point is set as the recognition result. If a plurality of character recognition information indicating the same predetermined character type (eg, Katakana, Kanji, English character) is continuously output as a recognition result, the concatenated character information storage means outputs the plurality of character recognition information. Considering that the character represented by the character code indicated by the character recognition information of is a plurality of characters that are elements that form one word,
Since a set of character codes indicated by the plurality of character recognition information is stored as the concatenated character information, a plurality of consecutive character recognitions including the same character type that have been output as a recognition result even once. After that, the information is preferentially output as the recognition result.

Therefore, since the same effect as learning as a word can be obtained, it is possible to output a recognition result that is likely to be Japanese and improve the character recognition rate without performing word processing or Japanese processing. become able to. Furthermore, since word processing and Japanese language processing are not performed, the recognition result can be output at high speed.

[0037]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a functional block diagram of the character input device of this embodiment.

In FIG. 1, 101 is coordinate input means, 1
Reference numeral 02 is a storage unit, 103 is a control unit, 104 is a display unit, 105 is an external storage unit, 106 is a power supply, and 107 is a power switch.

The coordinate input means 101 is specifically a tablet for inputting handwritten writing data online or a scanner for inputting handwritten writing data offline. In the present embodiment, the coordinate input means 101 is used. 10
A case where 1 is a tablet will be described.

The storage means 102 is specifically a main memory, and stores various programs, a character recognition dictionary, and the like. Further, the control means 103 is specifically a CPU
And executes various programs stored in the storage unit 102.

The display unit 104 is specifically a liquid crystal display device or a CRT, and displays the writing data input by the coordinate input unit 101 and the execution results of various programs by the control unit 103. In this embodiment, a case where the display unit 104 is a liquid crystal display device will be described.

The external storage means 105 is specifically a floppy disk or an IC card. In addition, the power source 106
Supplies current to these means, and the power switch 107
Is for the operator to instruct turning on / off of current supply by the power source 106.

FIG. 2 is an external view of the character input device of this embodiment.

As shown in FIG. 2, the character input device of this embodiment is roughly composed of a main body 201 and an attached pen 202. The main body 201 has a built-in storage means 102, control means 103, and power supply 106, a power switch 107, a display-integrated tablet 203 in which a liquid crystal display device is arranged under a transparent tablet, and an external storage means 1.
05 is provided with an insertion port 204.

When the operator touches the display-integrated tablet 203 with the pen 202, the coordinate data of the position touched by the pen 202 is detected by the tablet, and the detected coordinate data is output to the control means 103. It
This allows the operator to use the pen 2 without using the keyboard.
With 02, handwriting data can be handwritten and various instructions can be given.

Next, the flow of operations by the operator in the character input device of this embodiment will be described with reference to FIGS.

FIG. 3 shows a screen display example of the display-integrated tablet 203 immediately after the power switch 107 is turned on.

As shown in FIG. 3, the display-integrated tablet 203 has a body area 301 for displaying characters (characters represented by character codes), a cursor 302 for indicating the display position of the next character, and an operator. Is a character frame 303 for handwriting writing data of characters, a recognition button 304 for instructing to perform character recognition processing on the handwritten writing data, and all recognition candidate characters (in character codes) output by the character recognition processing. Recognition candidate frame 305 displaying (characters represented)
And a confirmation button 306 for instructing to display the recognition result character (character represented by the character code) in the body area 301.
And are displayed.

When inputting characters, the operator firstly
As shown in FIG. 4, the writing data of a character is handwritten in the character frame 303 with the pen 202. Accordingly, the tablet detects the coordinate data of the handwriting data, the pixel of the liquid crystal display plate corresponding to the detected coordinate data is reversely displayed, and the handwriting of the handwriting data is displayed in the character frame 303. Note that, here, the operator is handwriting the writing data intended for the character “day”.

Then, as shown in FIG. 5, the operator touches the recognition button 304 with the pen 202 to instruct to perform the character recognition processing on the handwritten data. As a result, character recognition processing is performed on the writing data, one or more recognition candidates are output, and the character of the first recognition candidate among these recognition candidates is the character as the recognition result. It is displayed in the character frame 303 instead of the data. Here, the character (“day”) as intended by the operator is displayed in the character frame 303. Which of the one or more recognition candidates is to be determined as the first recognition candidate will be described later.

Then, similarly to the above-mentioned operation, the operator handwrites the writing data of the character with the pen 202 in the character frame 303 and presses the recognition button 304 with the pen 2 as shown in FIG.
By touching 02, it is instructed to perform character recognition processing on the data. Here, the operator is
I am handwriting handwritten data intended for the character "standing".

As a result, character recognition processing is performed on the writing data, and one or more recognition candidates are output.
The character of the first recognition candidate among these recognition candidates is the character frame 3 instead of the writing data as the character of the recognition result.
03, but here, for example, as shown in FIG. 7, a different character (“main”) different from the character intended by the operator (“day”) is displayed in the character frame 303. In this case, the operator performs a recognition result correction operation for correcting a desired recognition candidate among the recognition candidates other than the first recognition candidate (recognition result) to the first recognition candidate.

That is, the operator, as shown in FIG.
By touching the character (“main”) of the recognition result displayed in the character frame 303 with the pen 202, all recognition candidate characters having the character recognition result as the first recognition candidate are displayed in the recognition candidate frame 305. Instruct to do so. As a result, all recognition candidate characters including the first recognition candidate (here, “main”, “standing”, “cheap”, “wide”, “plate”) are displayed in the recognition candidate frame 305. The operator touches the intended character (“upright”) of these characters with the pen 202 to instruct that the recognition candidate of the character should be corrected to the first recognition candidate. As a result, the recognition candidate of the character (“standing”) that the operator has instructed to correct is corrected to the first recognition candidate, and as shown in FIG. 9, the character of the first recognition candidate after correction (“ "" Is displayed in the character frame 303 in place of the character ("main") of the first recognition candidate before correction.

When the operator finishes the recognition result correction operation, the operator pushes the confirm button 306 with the pen 20 as shown in FIG.
By touching with 2, it is instructed to display the character of the recognition result displayed in the character frame 303 in the body area 301. As a result, the characters (“day”, “standing”) as the recognition result displayed in the character frame 303 are displayed in the body area 301.

Now, the character input device of this embodiment is different from the conventional one when determining the first recognition candidate (recognition result) from the one or more recognition candidates output by the character recognition processing. By utilizing the general tendency of Japanese language without performing word processing or Japanese language processing, it is possible to output a recognition result that is likely to be Japanese and improve the character recognition rate.

As a general tendency of Japanese used in this embodiment, firstly, in Japanese, katakana and English characters are used as nouns (foreign words, proper nouns, etc.) in sentences. Therefore, there is a tendency that a large number of characters appear consecutively in many cases.
Further, in Japanese, there are many numbers in which two or more digits continue, and therefore, there is a tendency that the numbers often appear continuously for several characters.

When this first tendency is used, for example, depending on how the handwriting data is handwritten, the operator can select "shi (katakana)", "su (katakana)", "te (katakana)", "mu ( "Katakana)" was intended, but "shi", "mata (kanji)", "thousand (kanji)",
The character of the recognition result "Mu" is displayed,
The operator intends the characters "P (alphabet)", "O (alphabet)", "T (alphabet)", "A (alphabet)", "T (alphabet)", "O (alphabet)" However, "P", "0
It is possible to prevent the characters of the recognition result such as "(number)", "Ding (kanji)", "A", "Ding (kanji)", and "0 (number)" from being displayed.

Therefore, as a function for utilizing this first tendency, in the present embodiment, a recognition candidate that inherits the character type of the character of the previous recognition result is determined as the first recognition candidate (recognition result). I have a function to do. Hereinafter, this function is referred to as a character type inheritance function.

As a general tendency of Japanese used in this embodiment, secondly, there are hundreds of thousands of words in Japanese, but only a few words are used in daily life. It is about a hundred words, and there is a tendency that a specific word often appears repeatedly.

Therefore, as a function for utilizing this second tendency, in the present embodiment, the word input once is learned, and thereafter, the recognition result (first
(Recognition candidate of) is provided.
However, in the present embodiment, two consecutive Kanji, two consecutive Katakana,
Two consecutive letters are regarded as the elements that make up one word, and learning is performed. Hereinafter, this function is referred to as a connection learning function.

In the present embodiment, two functions, that is, the character type inheritance function and the connection learning function are provided, but it is easy to provide only one of them. Nor.

In the character input device of the present embodiment, the character recognition processing and the processing for realizing the character type inheritance function and the connection learning function are specifically performed by the control means 103.
It can be realized by executing various programs stored in the storage unit 102. The storage unit 102 also stores various data (including a character recognition dictionary) used by various programs.

FIG. 11 is a diagram schematically showing the contents of the storage means 102.

As shown in FIG. 11, the storage means 102 is
Various program areas 1101 and recognition candidate data area 1
102, a concatenation frequency data area 1103, an attenuation parameter data area 1104, and a character type inherited character number area 110.
5, a connected storage data area 1106, a coordinate data area 1107, and a body data area 1108.

The various program areas 1101 are areas for storing various programs including a character recognition processing program for performing character recognition processing, a main processing program described later, a connection learning processing program described later, and an addition point calculation processing program described later. .

The recognition candidate data area 1102 is an area in which the recognition candidates output by the character recognition processing are stored, and details thereof will be described later.

Concatenation frequency data area 1103, attenuation parameter data area 1104, character type inherited character number area 1105
Is an area for storing data necessary for realizing the character type inheritance function, and details of these areas will be described later.

The concatenation storage data area 1106 is an area for storing data necessary for realizing the concatenation learning function, and details thereof will be described later.

The coordinate data area 1107 is an area in which the coordinate data of handwritten writing data is stored, and the body data area 1108 is an area in which the character code of the character displayed in the body area 301 is stored. .

Incidentally, these areas 1101 to 1108
Can be realized by individual memories, which are partially or wholly independent, as required.

FIG. 12 is a diagram schematically showing the contents of the recognition candidate data area 1102, and shows the contents in the screen display shown in FIG.

In FIG. 12, 1201 is a recognition candidate storage area in which recognition candidates for handwritten data handwritten in the first character frame 303 are stored, and 1202 is handwritten data handwritten in the second character frame 303. Is a recognition candidate storage area in which the recognition candidates are stored, and the recognition candidate data area 1104 is provided with at least as many recognition candidate storage areas as the number of character frames 303.

The recognition candidate storage area includes, for each of all the recognition candidates, a character code field 1203 in which the character code of the recognition candidate is stored, and a character type field 1204 in which the character type of the character of the recognition candidate is stored. , A matching degree field 1205 in which the matching degree of the recognition candidate is stored,
The addition point field 1206 stores addition points which are output results of the character type inheritance function or the connection learning function.

For example, in the recognition candidate storage area 1201, in the character code field 1203, "day",
The character codes of the recognition candidates for the five characters "month", "yen", "eyes", and "white" are stored in the character type field 1
In “204”, “Kan” indicating that each character is a Kanji is stored.

In the character recognition process, a recognition candidate is output for each character by referring to a character recognition dictionary that stores standard writing patterns that are standard writing data and corresponding character codes. It is supposed to do. That is, by collating the handwritten writing data with the standard writing pattern in the character recognition dictionary, a value indicating the degree of similarity between the two shapes (here, the larger the value, the higher the degree of similarity. ) As the degree of coincidence, and when the degree of coincidence obtained is larger than a predetermined threshold value, the character code corresponding to the standard writing pattern is output as a recognition candidate together with the degree of coincidence obtained. There is. Therefore, the character code of the output recognition candidate is stored in the character code field 1203, and the matching degree of the recognition candidate is stored in the matching degree field 1205.
Will be stored in.

In this embodiment, the character recognition dictionary stores the character types of the characters (Hiragana, Katakana, Kanji, numbers, letters, symbols). The character type is also output together with the degree of coincidence. Therefore, the output recognition character type is
It is stored in the character type field 1204.

By the way, in the conventional character recognition processing, the recognition candidate having the highest degree of coincidence is determined as the first recognition candidate and is preferentially output as the recognition result. However, this degree of matching is a value obtained based on the result of collating the writing data with the standard writing pattern on a character-by-character basis, so that a recognition result in which the context is not referenced is output. It was Therefore, as described above, "shi", "su", "te", "mu"
"P", "O", "T",
In some cases, the recognition result of the characters "A", "T", "O" cannot be obtained as intended.

Therefore, in the present embodiment, the addition point for each recognition candidate is calculated by referring to the recognition result of the previous character, and the calculated addition point is stored in the addition point field 1206. A value obtained by summing the matching score stored in the matching score field 1205 and the addition points stored in the addition score field 1206 when determining the first recognition candidate from the one or more recognition candidates. The recognition candidate having the highest (hereinafter referred to as priority) is determined as the first recognition candidate. The details of the addition point calculation processing for calculating the addition points will be described later.

FIG. 13 is a diagram schematically showing the contents of the concatenation frequency data area 1103.

In the concatenation frequency data area 1103, for each character type, a value determined in advance based on the result of statistically investigating what kind of character the character that appears next to the character type is It is stored.

In FIG. 13, i indicates the character type of the previous character, and j indicates the character type of the next character. The i-th row and j-th column shows the probability Pij that the character type j appears next to the character type i.
For example, when the previous character is katakana, the probability that the next character is katakana is 70%.

FIG. 14 shows the attenuation parameter data area 110.
It is a figure which shows the content of 4 typically.

Katakana, letters, and numbers tend to appear consecutively as described above, but the number of consecutive characters is finite. Therefore, in the present embodiment, the character type inheritance function is designed to gradually attenuate the function each time a character of the same character type continues, and the attenuation parameter data area 1104 includes the attenuation for that purpose. The rate is stored.

Since the average number of consecutive characters of the same character type differs for each character type, the attenuation rate is based on the result of statistical investigation of the number of consecutive characters of the character type for each character type. It is a predetermined value.

FIG. 15 is a diagram schematically showing the contents of the inherited character number area 1105.

In the inherited character number area 1105, the number of inherited characters used for attenuating the character type inheritance function is stored together with the attenuation rate stored in the attenuation parameter data area 1104. The inherited character number is a value indicating how many characters the character type of the character of the recognition candidate determined as the first recognition candidate has been the same character type up to that time. For example, as shown in FIG. 9, “2” is stored in the inherited character number area 1105 in a state where the screen is displayed after the operator performs the recognition result correction operation.

FIG. 16 is a diagram schematically showing the contents of the concatenated storage data area 1106.

In the concatenated storage data area 1106, when the operator performs a recognition result correction operation, the character code of the character for which correction is instructed and the character code of the preceding character are paired. (Hereinafter referred to as a concatenated character code set) is stored. For example, as shown in FIG. 9, in the state where the screen display is displayed after the operator performs the recognition result correction operation, the character code of the character “day” and the “standing” are displayed in the connected storage data area 1106. A concatenated character code set that is a combination of the character code of the character "" is stored.

However, in the concatenated storage data area 1106, when the character type of the character instructed by the operator to be corrected is the same as the character type of the preceding character (the character displayed in the character frame 303). And the type is Kanji, Katakana,
The content is stored only if it is one of the alphabetic characters.

Since Kanji, Katakana, and English characters are often repeatedly input as words, two consecutive Kanji characters, two consecutive Katakana, and two consecutive alphabetical characters that have been input once constitute one word. The concatenated character code set, which is a combination of those character codes, is stored in the concatenated storage data area 1106, and thereafter, the character code stored as the concatenated character code set in the concatenated storage data area 1106. By preferentially deciding the recognition candidate of No. 1 as the first recognition candidate, it is possible to output a recognition result that is likely to be Japanese and improve the character recognition rate without performing word processing or Japanese processing. Also, since word processing and Japanese processing are not performed, the recognition result can be output at high speed.

Next, in the character input device of this embodiment,
Processing for realizing the character type inheritance function and the connection learning function will be described with reference to FIGS. 17 to 22.

FIG. 17 is a flow chart of the main processing program.

The main processing program is the power switch 1
It is activated when 07 is input, and first, a hardware-like initialization process is performed (step 1701). Subsequently, it is determined whether or not the tablet has detected the coordinate data (step 1702). When it is determined that the coordinate data has been detected, whether the coordinate data is within the character frame 303 (step 1703), recognition is performed. Whether it is in the button 304 (step 1704) or in the confirm button 306 (step 1)
705), is it within the recognition candidate frame 305 (step 17)
06).

If the coordinate data detected by the tablet is within the character frame 303 (step 1703), the character frame 3
When handwriting data is written in 03 (the screen display shown in FIG. 4) and when the characters displayed in the character frame 303 are touched (screen display shown in FIG. 8). Since there are two possible cases, it is checked whether or not characters are displayed in the character frame 303 (step 1707).

If no character is displayed in the character frame 303, it means that the writing data is handwritten in the character frame 303. Therefore, the liquid crystal display device corresponding to the coordinate data detected by the tablet is displayed. The pixels are displayed in reverse, and the handwriting of handwritten writing data is displayed as in the screen display shown in FIG. 4 (step 1708),
The coordinate data is stored in the coordinate data area 1107 (step 1709). Further, if a character is displayed in the character frame 303, it means that the character displayed in the character frame 303 is touched. Therefore, the recognition candidate of the character is the first recognition candidate. All recognition candidates
The character of the extracted recognition candidate is extracted from the recognition candidate data area 1102 and displayed in the recognition candidate frame 305 as in the screen display shown in FIG. 8 (step 1710).

If the coordinate data detected by the tablet is in the recognition button 304 (step 170).
4) This means that the recognition button 304 has been touched. Therefore, it is checked whether or not the coordinate data is stored in the coordinate data area 1107 (step 1711). A character recognition processing program is activated so as to perform character recognition processing on coordinate data (writing data) (step 1712). The character recognition processing performed by the character recognition processing program is as described above. When the character recognition processing program ends the character recognition processing, the recognition candidates (character code, character type, degree of coincidence) output by the character recognition processing are matched. The data are stored in the recognition candidate data area 1102 in descending order (step 1713),
It is possible to determine how many characters the characters of the same character type (characters displayed in the character frame 303) as the character type of the first recognition candidate character at the time of storage are consecutive by the first character. It calculates backward from the recognition candidate character and stores the calculation result in the inherited character number area 1105 (step 1714).
Specifically, this calculation is performed on the recognition candidate data area 1102.
, The first recognition candidate storage area in which the first recognition candidate is stored can be referred to by referring to the character type field 1205 of the first recognition candidate stored in all the recognition candidate storage areas. .

Subsequently, after the coordinate data area 1107 is initialized (step 1715), addition is performed so that addition point calculation processing for calculating the addition points of the recognition candidates stored in the recognition candidate data area 1102 in step 1714 is performed. The point calculation processing program is started (step 1716). In addition,
Details of the addition point calculation processing will be described later.

When the addition point calculation processing program ends the addition point calculation processing, in the recognition candidate data area 1102, the calculated addition points are stored in the addition point field 1206 and the matching degree stored in the matching degree field 1205. Since the recognition candidates are rearranged in descending order of priority, which is the sum of the addition points stored in the addition point field 1206, the recognition candidate with the highest priority (1
The second stored recognition candidate) is determined as the first recognition candidate, and the determined first recognition candidate character is displayed in the character frame 303 as in the screen display shown in FIGS. 5 and 7. (Step 1717).

If the coordinate data detected by the tablet is in the confirm button 306 (step 170)
5) Since it means that the confirm button 306 has been touched, it is checked whether or not the recognition candidate is stored in the recognition candidate data area 1102 (step 1718). If the recognition candidate is stored, The character code of the first recognition candidate among them is added to the end of the data in the body data area 1108 and stored (step 1719), and the character of the character code is displayed at the display position indicated by the cursor 302 in the body area 301 ( Step 1720).

Subsequently, the recognition candidate data area 1102 is initialized (step 1721) and the inherited character number area 1105 is set.
Is initialized (step 1722), and the inside of the character frame 303 is initialized (step 1723).

If the coordinate data detected by the tablet is within the recognition candidate frame 305 (step 170)
6) It is checked whether or not the character is displayed in the recognition candidate frame 305 (step 1724). If the character is displayed, the recognition candidate frame 30 is displayed as in the screen display shown in FIG.
Since it means that the character displayed in No. 5 is touched, in the recognition candidate data area 1102, the recognition candidate of the touched character, that is, the recognition candidate of the character instructed to be corrected and its time point. The first recognition candidate of the character is replaced by the first recognition candidate of the character
(Step 1725).

Subsequently, the connection learning processing program is activated so as to perform the connection learning processing for learning the recognition candidates of the character instructed to be corrected (step 1726). The details of the connection learning process will be described later.

When the connection learning processing program ends the connection learning processing, the character of the character code of the first recognition candidate after the replacement in step 1725 is displayed in place of the character displayed in the character frame 303 ( Step 172
7), the inside of the recognition candidate frame 305 is initialized (step 17).
28). In addition, the number of consecutive characters by the same character type (characters displayed in the character frame 303) as the character type of the first recognition candidate character after being replaced in step 1725 is determined. , Is calculated backward from the character, and the calculation result is stored in the inherited character number area 1105 (step 1729). Specifically, this calculation is stored in all the recognition candidate storage areas before the recognition candidate storage area in which the recognition candidates of the character of the first recognition candidate are stored in the recognition candidate data area 1102. This can be done by referring to the character type field 1205 of the first recognition candidate.

Subsequently, the characters after the first recognition candidate character after the replacement in step 1725 (character frame 303
The addition point calculation processing program is started so as to perform the addition point calculation processing for calculating the addition point of the recognition candidate of the character displayed in (step 1730).

Here, the reason why the addition point calculation processing is performed in step 1730 will be described.

After the operator handwrites the writing data for a plurality of characters in the character frame 303, the operator presses the recognition button 304 with the pen 20.
When touched with 2, the characters of the recognition result of the writing data for the plurality of characters are displayed in the character frame 303. At this time, if the operator performs the recognition result correction operation for the character in the middle. , Inherited character number area 110 in step 1729
The number of inherited characters stored in 5 may be different from the value previously stored, and accordingly, the addition points of the recognition candidates of the characters after the character may change. Therefore,
In such a case, the addition points of the recognition candidates of all the characters after the character for which correction is instructed are recalculated, and the recognition candidates are sorted in descending order of priority, which is the sum of the degree of coincidence and the recalculated addition points. To be replaced. The addition point calculation process is performed in step 1730 in order to recalculate this addition point.

When the addition point calculation processing program ends the addition point calculation processing, in the recognition candidate data area 1102, the recalculated addition points are stored in the addition point field 1206 and the matching degree stored in the matching degree field 1205. And the addition points stored in the addition point field 1206 are sorted in descending order of priority, the recognition candidates are rearranged. Therefore, the recognition candidate having the highest priority (the recognition candidate stored first) ) Is determined as the first recognition candidate, and the character of the determined first recognition candidate is displayed in the character frame 303 (step 1731).

In this embodiment, the connection learning process is performed in step 1726 only when the operator performs the recognition result correcting operation. However, if the capacity of the storage means 102 is large to some extent, the operation is performed. It is also possible to increase the chances of learning by performing the connection learning process every time the person handwrites the writing data of characters.

Further, in the present embodiment, it is premised that one or more recognition candidates are always output by the character recognition processing performed in step 1712. However, in the character recognition processing, the handwritten data written by the operator is written. If no standard writing data in which the degree of coincidence between and is larger than a predetermined threshold is stored in the character recognition dictionary, no recognition candidate is output. In such a case, lower the threshold. By performing the character recognition process again, one or more recognition candidates are always output. However, since recognition candidates having a low degree of coincidence, that is, recognition candidates having a shape that is not very similar to each other, are output, in such a case, the concatenated storage data area 1106 is referred to and the past If there is a recognition candidate of a character that has been handwritten even once, it is possible to prevent the recognition candidate other than the recognition candidate from being output, so that an extra recognition candidate is not output.

FIG. 18 is a flow chart of a connection learning processing program for performing the connection learning processing in step 1726.

In the connection learning process, when the operator performs the recognition result correction operation, the character code of the character instructed to be corrected and the character code immediately before (the character displayed in the character frame 303) A concatenated character code set that is a combination of character code and
This is a process of storing in the concatenated storage data area 1106. However, if the character type of the character instructed to be modified is the same as the character type of the previous character, and the character type is one of Kanji, Katakana, or English. Only
Stores the concatenated character code set.

Specifically, it is checked whether the character type of the character instructed to be corrected by the operator is kanji, katakana, or alphabet (step 1801), and kanji, katakana, or alphabet. If any of the above, it is determined whether or not the character type of the character designated to be modified is the same as the character type of the previous character. A code set is acquired (step 1802). To obtain the concatenated character code set, specifically, the recognition candidate data area 11
In 02, it can be performed by referring to the recognition candidate storage area immediately preceding the recognition candidate storage area in which the recognition candidates of the character instructed to be corrected are stored.

If the concatenated character code set can be acquired in step 1802, the acquired concatenated character code set is stored in the concatenated storage data area 1106 (step 1803).

When storing the concatenated character code set,
It is possible that the same concatenated character code set has already been stored. In such a case, the already stored concatenated character code set may be overwritten and stored. May be allowed to be stored.

It is also possible to store the concatenated character code sets in the newest order. In this case, when the same concatenated character code set is already stored, the already stored concatenated character code set is deleted. Then, the newly acquired concatenated character code set is stored.

FIG. 19 is a flow chart of an addition point calculation processing program for executing the addition point calculation processing in step 1716 and step 1730.

As described above, after the character recognition processing is performed, the recognition candidates (character code, character type, matching degree) output by the character recognition processing are stored in the recognition candidate data area 1102 in descending order of matching degree. Although stored, the degree of coincidence is the result of performing character recognition processing on the writing data on a character-by-character basis, and does not refer to any context. Therefore,
In step 1716, by performing the addition point calculation process shown in the flowchart of FIG.
The addition points of the recognition candidates stored in 102 are calculated, and the recognition candidates are replaced in the descending order of priority, which is the sum of the coincidence and the calculated addition points.

Further, as described above, when the operator performs a recognition result correction operation for a character in the middle, the number of inherited characters may be changed. Along with this, the addition points of the recognition candidates of the characters after the character may be changed. May change. Therefore, in step 1730, by performing the addition point calculation process shown in the flowchart of FIG. 19 by the number of characters after the character that the operator has instructed to correct, all the characters after the character instructed to be corrected by the operator are performed. With respect to the character, the addition points of the recognition candidates of the character are recalculated, and the recognition candidates are replaced in descending order of the sum of the matching degree and the recalculated addition point.

As will be described below, the addition points include addition points by the concatenation learning function and addition points by the character type inheritance function. In the present embodiment, the addition points by the concatenation learning function are prioritized. There is. That is, normally, the addition point by the character type inheritance function is used, and especially for the character for which the operator has performed the recognition result correction operation in the past, the addition point by the connection learning function is preferentially used. By using the general tendency of Japanese, the character recognition rate can be improved according to the operation of the operator.

Specifically, all the recognition candidates stored in the former are referenced by referring to the recognition candidate storage area in which the recognition candidate for which the addition point is to be calculated is stored and the previous recognition candidate storage area. Of the character code of the first recognition candidate stored in the latter and the character code of the first recognition candidate stored in the latter, and whether or not a concatenated character code set matching the group is stored in the concatenated storage data area 1106. Check (step 1901). If a concatenated concatenated character code set is stored (step 1902), an addition point by the concatenation learning function is calculated for the character code recognition candidates in the set, and the calculated addition point is added to the addition point field 1206.
(Step 1903). Specifically, the addition point by the connection learning function is a predetermined positive value.

In the connection learning process, when it is allowed to store overlapping connection character code sets in the connection storage data area 1106, two or more matching connection character code sets must be stored. Therefore, at that time, the number of learnings can be reflected in the addition point by setting the value obtained by multiplying the number of stored concatenated character code sets matching the predetermined positive value as the addition point. it can. Further, in the concatenation learning process, when the concatenation character code sets are stored in the concatenation storage data area 1106 in the newest order, the newer the time when the matching concatenation character code set is stored, the larger the value becomes. By doing so, new / old learning can be reflected in the addition points.

Then, when the calculation of the addition points is completed for all the recognition candidates for which the addition points are to be calculated, recognition is performed in the recognition candidate data area 1102 in descending order of the sum of the degree of coincidence and the addition points. The candidates are rearranged (step 1904).

FIG. 20 is a diagram showing a specific example of the addition points by the concatenation learning function. As shown in the screen display of FIG. 9, after the operator performs the recognition result correction operation, the operator does the second time. The contents of the recognition candidate data area 1102 in the case of handwriting the writing data intended for the characters "day" and "standing" are shown in FIG.

When the operator handwrites the writing data intended for the characters “day” and “standing”, the first time, the recognition candidates of the characters “day” and “main” different from the operator's intention are obtained. Although it is determined as the first recognition candidate and is erroneously recognized, the character code of the character "Sun" and the character "Stand" are recognized due to the recognition result correction operation for the character "Stand". Since a concatenated character code set that is a set of the character code of the character is stored in the concatenated storage data area 1106, the addition point by the concatenation learning function is used for the second time, as shown in FIG. The addition point of the recognition candidate of the character "standing" is "+50". Therefore, the recognition candidate of the character "standing" having the highest priority, which is the sum of the degree of coincidence and the addition point according to the learning content, is replaced with the recognition candidate of the character "main", and the characters "day" and "standing" are replaced. Is determined as the first recognition candidate, and the recognition result as intended by the operator can be obtained.

As described above, among the combinations of the character codes of all the recognition candidates whose addition points are to be calculated and the character code of the character immediately before (the character code of the first recognition candidate), the concatenation is performed. A set that matches the concatenated character code set stored in the storage data area 1106 means that the set is an element that constitutes one word.
The addition points of the character code recognition candidates in the set (addition points by the concatenation learning function) are calculated, and the degree of determination as the first recognition candidate increases. Therefore, even if you do not perform word processing or Japanese processing, by the connection learning function,
It is possible to improve the character recognition rate by outputting the recognition result of "day" and "standing" as intended by the operator.

On the other hand, if the matching concatenated character code set is not stored (step 1902), the addition point by the character type inheritance function is calculated (step 1905).

Here, the calculation processing of the addition point by the character type inheritance function of step 1905 will be described with reference to the flowchart of FIG.

As described above, when character recognition processing is performed character by character as in the conventional case, "PO", "T",
In the case of difficult-to-identify characters such as "A", "T", and "O", erroneous recognition occurs. Therefore, in this embodiment,
For each character type, a value (probability) predetermined based on the result of statistically investigating the character type of the character that appears next to the character of the character type is stored in the concatenation frequency data area 1103. In addition, the character type inheritance function can be realized. In addition, katakana, letters, and numbers tend to appear consecutively as words in many cases, but since the number of consecutive characters is finite, in order to gradually attenuate the character type inheritance function, For each character type, a value (attenuation rate) determined in advance based on the result of statistically examining the number of consecutive characters of the character type is stored in the attenuation parameter data area 1104. A value (the number of inherited characters) indicating how many characters the character type of the character of the recognition candidate determined as the recognition candidate has been the same character type until then is stored in the inherited character number area 1105. .

Specifically, the recognition candidate data area 1102
In, the addition point of the recognition candidate is determined by checking whether or not the recognition candidate is stored in the recognition candidate storage area immediately preceding the recognition candidate storage area in which the recognition candidate whose addition point is to be calculated is stored. It is checked whether or not there is a character (character displayed in the character frame 303) immediately before the character to be calculated (step 2101). If there is a character, the variable i is set to the character before the character. The character type (the character type of the character of the first recognition candidate of the recognition candidates of the previous character) is set (step 2102), and the variable j is the recognition candidate currently being processed (the kth recognition candidate). The character type of the character) is set (step 2103).

Subsequently, from the variable i (character type of the character immediately before) and the variable j (character type of the character of the kth recognition candidate),
The addition point Dij of the kth recognition candidate is calculated using (Equation 1) (step 2104).

Dij = (Pij + N × αi × Pij) × A (Equation 1) Here, Pij is the probability stored in the concatenation frequency data area 1103, and N is the inherited character number area 1105.
Is the number of inherited characters stored in, and αi is the attenuation rate stored in the attenuation parameter data area 1104. Also, A is a parameter for adjusting the weighting of the addition points by the character type inheritance function, and has a predetermined value.

In the concatenation frequency data area 1103, as shown in FIG. 13, the probability Pij is determined in advance so that the tendency that hiragana, katakana, letters and numbers are continuous becomes strong. The addition point Dij of the recognition candidates of the character having the same character type as that of
The character type inheritance function can be realized.

Further, as the number N of inherited characters increases, the addition point Dij gradually decreases. For example, in the case of katakana, when N = 0, Dij = (70 + 0) × 0.5 = 35
Therefore, when N = 1, Dij = (70 + 1 × (−0.
2) × 70) × 0.5 = 28, and when N = 2, Di
j = 21, N = 3, Dij = 14, N
= 4, Dij = 7. Here, A =
Although calculated as 0.5, the value of A should be changed according to the degree of coincidence, for example, the value of A is increased as the average value of the degree of coincidence of recognition candidates is increased. You can

As described above, as the number of inherited characters N increases, the addition point Dij turns to a negative number at a certain point, so it is checked whether or not the addition point Dij is positive (step 2
105) If it is positive, the addition point Dij is stored in the addition point field 1206 of the recognition candidate data area 1102 (step 2106). Note that if the addition point Dij is not determined to be positive or negative, the recognition candidate of a character having the same character type as the previous character has a high priority while it continues for a certain number of characters. When the number of characters exceeds the number of consecutive characters, the priority can be reduced.

Subsequently, after calculating the addition points Dij for all the recognition candidates for which the addition points are to be calculated and storing the calculated addition points Dij in the addition point field 1206 (step 2107), the recognition candidate data area is calculated. 1102
In step S2108, the recognition candidates are rearranged in the descending order of priority, which is the sum of the coincidence and the addition points. Then, as a result of the rearrangement, the number of consecutive characters of the same character type as the character type of the recognition candidate character that has become the first recognition candidate is determined by the first recognition candidate. It calculates backward from the character and calculates the calculation result.
It is stored in 105 (step 2109). This calculation is
Specifically, in the recognition candidate data area 1102, the character type field of the first recognition candidate stored in all the recognition candidate storage areas before the recognition candidate storage area in which the first recognition candidate is stored. This can be done by referring to 1205.

FIG. 22 is a diagram showing a specific example of the addition points by the character type inheritance function.
The contents of the recognition candidate data area 1102 in the case where the handwriting data intended for the characters "te" and "mu" are handwritten are shown.

The operator selects "shi", "su", "te", "mu".
When handwriting the writing data intended for the character
22 (a), unless the addition point by the character type inheritance function is used.
As shown in, the highest degree of coincidence is "shi", "mata",
Although the recognition candidates of the characters "thousand" and "mu" are determined as the first recognition candidate, they are erroneously recognized.
When the addition point by the character type inheritance function is used, as shown in FIG. 22B, the recognition candidates of the characters “su” and “te” having the highest priority, which is the sum of the coincidence degree and the addition point Dij, are “ Since it is replaced with the recognition candidates of the characters “m” and “thousand”, the recognition candidates of the characters “shi”, “su”, “te”, and “mu” are determined as the first recognition candidates, and the intention of the operator is determined. You will get the same recognition results.

As described above, with respect to the recognition candidates of the characters after the character "shi", since the character type of the preceding character is katakana, the addition of the recognition candidates whose character type is katakana is added. The degree to which the points are calculated and determined as the first recognition candidate increases. Therefore, even if word processing or Japanese processing is not performed, the character type inheritance function enables
It is possible to improve the character recognition rate by outputting Japanese-like recognition results such as "su", "te", and "mu".

[0140]

As described above, according to the present invention,
By using the general tendency of Japanese language without using word processing or Japanese language processing, a character input device that can output a recognition result that seems to be Japanese and can improve the character recognition rate is provided. It becomes possible to provide.

Further, since word processing and Japanese language processing are not performed, the recognition result can be output at high speed.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a character input device according to an embodiment.

FIG. 2 is an external view of the character input device according to the present embodiment.

FIG. 3 is an explanatory diagram showing a screen display example of the present embodiment.

FIG. 4 is an explanatory diagram showing a screen display example of the present embodiment.

FIG. 5 is an explanatory diagram showing a screen display example of the present embodiment.

FIG. 6 is an explanatory diagram showing a screen display example of the present embodiment.

FIG. 7 is an explanatory diagram showing a screen display example of the present embodiment.

FIG. 8 is an explanatory diagram showing a screen display example of the present embodiment.

FIG. 9 is an explanatory diagram showing a screen display example of the present embodiment.

FIG. 10 is an explanatory diagram showing a screen display example of the present embodiment.

FIG. 11 is an explanatory diagram schematically showing the contents of storage means.

FIG. 12 is an explanatory diagram schematically showing the contents of a recognition candidate data area.

FIG. 13 is an explanatory diagram schematically showing the contents of a concatenation frequency data area.

FIG. 14 is an explanatory diagram schematically showing the contents of an attenuation parameter data area.

FIG. 15 is an explanatory diagram schematically showing the content of the inherited character number area.

FIG. 16 is an explanatory diagram schematically showing the contents of a concatenated storage data area.

FIG. 17 is a flowchart of a main processing program.

FIG. 18 is a flowchart of a connection learning processing program.

FIG. 19 is a flowchart of an addition point calculation processing program.

FIG. 20 is an explanatory diagram showing a specific example of an addition point by the connection learning function.

FIG. 21 is a flowchart of addition point calculation processing by the character type inheritance function.

FIG. 22 is an explanatory diagram showing a specific example of addition points by the character type inheritance function.

[Explanation of symbols]

101 ... Coordinate input means, 102 ... Storage means, 103 ... Control means, 104 ... Display means, 105 ... External storage means, 1
06 ... Power source, 107 ... Power switch, 201 ... Main body, 2
02 ... Pen, 203 ... Tablet with integrated display, 204 ...
Insertion slot, 301 ... Body area, 302 ... Cursor, 303
... character frame, 304 ... recognition button, 305 ... recognition candidate frame,
306 ... Confirm button, 1101 ... Various program areas,
1102 ... Recognition candidate data area, 1103 ... Concatenation frequency data area, 1104 ... Attenuation parameter data area, 11
05 ... Character type inherited character number area, 1106 ... Concatenated storage data area, 1107 ... Coordinate data area, 1108 ... Body data area, 1201, 1202 ... Recognition candidate storage area, 12
03 ... Character code field, 1204 ... Character type field, 1205 ... Matching degree field, 1206 ... Addition point field.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Hiroyuki Kumai, Inventor Hiroyuki Kumai, 292 Yoshida-cho, Totsuka-ku, Yokohama, Kanagawa, Ltd.Inside the Hitachi Media Visual Media Research Institute (72) Yasumasa Matsuda, 292 Yoshida-cho, Totsuka-ku, Yokohama, Kanagawa Hitachi, Ltd. Visual Media Research Center

Claims (11)

[Claims] 1. A handwriting input means for inputting handwritten data of handwritten characters, a standard writing pattern which is a standard writing pattern of the character, a character code representing the character, and a A character recognition dictionary storage unit that stores a character type and a matching degree that indicates the degree to which the writing data input by the handwriting input unit resembles the standard writing pattern are obtained.
A character recognition means for outputting the character code and the character type corresponding to the standard writing pattern for which the obtained coincidence is larger than a predetermined value, together with the obtained coincidence, as character recognition information, and the character output by the character recognizing means. A character recognition information storage unit that stores recognition information, and a predetermined probability value that indicates, for each character type, the probability that the character that appears next to the character of the character type will be. According to the stored character type connection information storage means and the probability value stored in the character type connection information storage means, the character recognition is performed for each character recognition information stored in the character recognition information storage means. An addition point calculating means for calculating an addition point to be added to the coincidence degree indicated by the information, and outputting character recognition information having the largest total value of the coincidence degree and the calculated addition point as a recognition result; Written by And a display unit for displaying a character represented by a character code indicated by the character recognition information output as the recognition result by the addition point calculation unit. 2. The character input device according to claim 1, wherein character recognition information including a character code representing a character displayed by said display means is replaced with character recognition information other than said character recognition information. The display means is further provided with a receiving means for receiving an instruction, and when the receiving means receives the instruction for replacing, the displayed character is represented by a character code indicated by the character recognition information instructed to replace the character. When the display means changes and displays the character, the addition point calculation means displays the character as the recognition result after the character recognition information including the character code representing the character. A character input device, which calculates the addition points again for all the character recognition information outputting the character recognition information, and outputs the recognition result again. 3. A handwriting input means for inputting handwriting data of a handwritten character, a standard writing pattern which is a standard writing pattern of the character, a character code representing the character, and a character code of the character. A character recognition dictionary storage unit that stores a character type and a matching degree that indicates the degree to which the writing data input by the handwriting input unit resembles the standard writing pattern are obtained.
A character recognition means for outputting the character code and the character type corresponding to the standard writing pattern for which the obtained coincidence is larger than a predetermined value, together with the obtained coincidence, as character recognition information, and the character output by the character recognizing means. Character recognition information storage means for storing recognition information, concatenated character information storage means for storing concatenated character information in which character codes representing a plurality of characters that are elements forming one word are stored, and the concatenated character information storage Based on the concatenated character information stored in the means, the addition point to be added to the matching degree indicated by the character recognition information is calculated for each character recognition information stored in the character recognition information storage means, and the matching is calculated. The character recognition information having the largest sum of the degree and the calculated addition point, as the recognition result, the addition point calculation means, the handwriting data input by the handwriting input means, and the above The addition point calculation means includes a display means for displaying a character represented by a character code indicated by the character recognition information output as a recognition result, and the concatenated character information storage means is a plurality of predetermined character types indicating the same character type. If the character recognition information is continuously output as the recognition result, it is considered that the character represented by the character code indicated by the plurality of character recognition information is a plurality of characters that are elements forming one word. A character input device characterized in that a set of character codes indicated by the plurality of character recognition information is stored as the concatenated character information. 4. A handwriting input means for inputting handwritten data of handwritten characters, a standard writing pattern that is a standard writing pattern of the character, a character code representing the character, and a character code of the character, for each character. A character recognition dictionary storage unit that stores a character type and a matching degree that indicates the degree to which the writing data input by the handwriting input unit resembles the standard writing pattern are obtained.
A character recognition means for outputting the character code and the character type corresponding to the standard writing pattern for which the obtained coincidence is larger than a predetermined value, together with the obtained coincidence, as character recognition information, and the character output by the character recognizing means. Character recognition information storage means for storing recognition information, concatenated character information storage means for storing concatenated character information in which character codes representing a plurality of characters that are elements forming one word are stored, and the concatenated character information storage Based on the concatenated character information stored in the means, the addition point to be added to the matching degree indicated by the character recognition information is calculated for each character recognition information stored in the character recognition information storage means, and the matching is calculated. Degree calculation means for outputting the character recognition information having the largest sum of the degree and the calculated addition point as a recognition result, and the character including the character code representing the character displayed by the display means. Receiving means for receiving an instruction to replace the recognition information and other character recognition information other than the character recognition information, handwriting data input by the handwriting input means, and the addition point calculation means output as a recognition result. Display means for displaying the character represented by the character code indicated by the character recognition information, and the display means replaces the displayed character when the receiving means receives an instruction to replace the character. The display is changed to a character represented by the character code indicated by the character recognition information instructed to the effect, and the concatenated character information storage means is predetermined when the acceptance means accepts the instruction to replace. If a plurality of character recognition information indicating the same character type is continuously output as the recognition result, the character represented by the character code indicated by the plurality of character recognition information indicates one word. Character input characterized in that it is regarded as a plurality of characters that are elements to be formed, and a combination of character codes indicated by the plurality of character recognition information is stored as the concatenated character information. apparatus. 5. The character input device according to claim 4, wherein the addition point calculation means, when the display means changes and displays a character, after the character recognition information including a character code representing the character, A character input device, which calculates the addition point again for all the character recognition information that has already output the character recognition information as the recognition result, and outputs the recognition result again. 6. The character input device according to claim 3, 4 or 5, wherein the addition point calculation means adds points for character recognition information based on the concatenated character information stored in the concatenated character information storage means. When calculating, the concatenated character information including a combination of the character code indicated by the character recognition information and the character code indicated by the character recognition information output as the previous recognition result is stored in the concatenated character information storage means. A character input device, characterized in that the addition point for the character recognition information is set to a predetermined value. 7. The character input device according to claim 3, 4, 5 or 6, wherein, for each character type, a probability of what character type the character that appears next to the character type is The character point connection information storage means for storing a predetermined probability value is further provided, and the addition point calculation means is any one of the character recognition information stored by the character recognition information storage means. When the concatenated character information including a combination of the character code indicated by the character recognition information and the character code indicated by the character recognition information output as the previous recognition result is stored in the concatenated character information storage means, The addition point is calculated based on the concatenated character information stored in the concatenated character information storage means, and if not, the addition point is further calculated according to the probability value stored in the character type concatenation information storage means. Sentences characterized by calculations Input device. 8. The character input device according to claim 1, 2 or 7, wherein the probability value stored in said character type connection information storage means is such that the character type of the character that appears next is the character type of the preceding character. The probability value to be the same as the character type is larger than the probability value that the character type of the character that appears next is different from the character type of the previous character. When calculating the addition point for the character recognition information according to the probability value stored in the character type connection information storage means, the character type indicated by the character recognition information and the character recognition information output as the previous recognition result indicate A character input device, wherein a probability value corresponding to the character recognition information is obtained from a character type, and the larger the obtained probability value is, the larger the addition point for the character recognition information is. 9. The character input device according to claim 8, wherein the character recognition information including the same character type as the character type indicated by the character recognition information output as the recognition result by the addition point calculation means is continuous until then. And further includes a continuous character type number calculating means for calculating the number of continuous character types indicating whether or not it is output as a recognition result, wherein the addition point calculating means further includes the continuous character type calculated by the continuous character type number calculating means. A character input device characterized in that the larger the number, the smaller the addition point for the character recognition information including the same character type as the character type indicated by the character recognition information output as the previous recognition result. 10. A first step of outputting a character code representing one or more characters having a high degree of similarity to the handwritten character's writing data, and a character code currently output. Or 1
And a second step of outputting the result as a recognition result. In the second step, the character code output as the previous recognition result is displayed among the one or more character codes currently output. A character recognition method characterized by preferentially outputting a character code that represents a character having the same character type as the character type of the specified character as a recognition result. 11. One of a first step of outputting a character code representing one or more characters having a high degree of similarity to the handwritten character's writing data and a character code currently output. Or 1
And a second step of outputting one as a recognition result, and in the second step, the character type of the character represented by the character code among the one or more character codes currently output. And the character type of the character represented by the character code output as the previous recognition result is the same predetermined character type, and the character code and the character code output as the previous recognition result are paired. The character code that has been output in the past as two consecutive recognition results,
A character recognition method characterized by preferentially outputting as a recognition result.