JPH09259225A - Character recognizing and converting method and character data processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform character recognition processing without confirming operation or correcting operation for a character recognition result by extracting character pattern data between stored character pattern data and the character pattern data and performing character recognition again. SOLUTION: When writing is done on a tablet 12 with a pen 10, its coordinates is detected by a coordinate detecting circuit 14 and a control part 16 displays the writing. The control part 16 once detecting no coordinate input being done for a specific time performs segmentation processing for a character string to divide the character string by characters, and performs character recognition processing. Consequently, the degree of correct answer conviction of of character code data is low, the coordinate data (character pattern data) is outputted, but when the degree of correct answer conviction is high, on the other hand, the character code data is substituted for and outputted. This output is stored in a RAM 24 and displayed at a display part 20 through a display memory 18. Namely, writing data is erased, the character code data is inserted and displayed as character font data at a cursor position, and the coordinate data is converted to size less than the spacing between lines and displayed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition method and a character data processing device.

[0002]

2. Description of the Related Art First, a character recognition device will be briefly described.

(1). The online character recognition device will be briefly described. Online character recognition technology is also called online handwritten character recognition technology, and is used in pen personal computers and electronic notebooks. The general operation of online character recognition technology will be described. For example, when writing with a stylus pen on an LCD integrated tablet, the handwriting coordinates are once displayed on the corresponding LCD. That is, the written characters are displayed.

Then, character recognition processing is performed based on the character string stroke coordinate data (character pattern data) which is the coordinate data at the time of writing, and the character code corresponding to this written character data is detected. Then, the handwritten character displayed on the LCD is erased, and instead, the character font corresponding to the detected character code is displayed on the LCD.

(2). The offline character recognition device will be briefly described. The off-line character recognition technology is mainly used in OCR devices. The general operation of online character recognition technology will be described. For example, an optical reading device takes in character image data (character pattern data) written on a sheet of paper, and performs character recognition processing of this character image data to detect a character code. Compared to the online character recognition technology, this offline character recognition technology cannot obtain the data of the stroke order of the characters, so the accuracy rate of recognition decreases.

This off-line character recognition technique is used for recognizing a relatively beautiful character or a printed character which is written on a paper having a predetermined format.

(3). The problems of the character recognition device will be briefly described. As described above, conventionally, character pattern data is character-recognized to form character code data, which is edited, displayed, printed, stored, and read. It has been pointed out that such a character recognition device has the following problems. (A). A uniform character changed to a character code gives a cold impression to others, and handwritten characters written on it may be preferable in some cases.

(B). Character recognition does not always output 100% correct characters as a recognition result. Therefore,
Confirmation work and correction work by humans are essential. (C). Character recognition processing is a heavy load. Therefore, C
In the case of processing by PU, if the capacity of this CPU is low, it takes a long time for recognition and the recognition waiting time becomes long. When processing by hardware, the circuit scale becomes large.

(D). If it's just a memo,
There is little need to perform character recognition processing and convert to character code data without showing it to others. (E). Writing other than letters cannot be recognized.

(4). As one of the solutions to the above problems, it has been proposed to store and display the character pattern data as it is. That is, it has been proposed to store and display the character pattern data itself without recognizing the character pattern data. However, by simply storing it, the handwritten character pattern data is treated just like the graphic information.

For this reason, various measures have been taken so that the handwritten character pattern data can be edited and displayed appropriately. For example, in "New Mobile Information Tool Liquid Crystal Pencom ZAURUS PI-6000" manufactured by Sharp Corporation, there is an "ink word processor" function. For example, FIG.
When written like IN1, the character pattern data (IN1) is changed to a predetermined size (normalized in size) and inserted at the cursor position (K). That is, FIG.
It is displayed like IN2.

This device measures the pen separation time, and when it exceeds a predetermined time, it judges that the input is completed.

As described above, since the editing unit is determined based on the pen separation time, in order to delete and edit in character units, the pen is not separated for a predetermined time each time one character is entered. must not.

By the way, a character data processing device in which the minimum editing unit is a character unit is disclosed in Japanese Patent Laid-Open No. 7-18249 (G06K9 / 3).
Proposed in 2).

In this Japanese Laid-Open Patent Publication No. 7-182449, lines and characters are cut out from the handwritten character data, and each handwritten character data is shaped based on the set of the cut out strokes (character unit, line unit).・ We arrange, insert, delete, copy and move character strings.

[0016] Further, cutting out in units of one character and performing editing processing is described in IPSJ Journal Aug. 1991 Vol.32 No.
8 Shinji Moriya, Toshihiro Morita, Koji Inai, Minoru Shimizu "Stroke editor and direct instruction / operation method" are also introduced.

[0017]

[Problems to be Solved by the Invention]

(1). The object of claim 1 of the present invention will be described. The present invention provides a character recognition conversion method capable of performing character recognition processing without checking and correcting character recognition results, and when the character recognition processing is insufficient, it can be corrected later. It is to propose.

(2). The objects of claims 2 to 4 of the present invention will be described. The printed character string and the carefully written character string can be cut out for each character with high accuracy. In general, a character string that is not polite may be cut out incorrectly. The present invention proposes a character data processing device that operates without any inconvenience even if the cutout is incomplete.

For example, if "to school" is input correctly and it can be cut out correctly, it is no problem to start a new line after "school" as shown in FIG. 2 (a). However, if the cut-out is incorrect and the “school” (41) is cut out as two characters, “tree” (42) and “cross” (43), as shown in FIG. It is possible to arrange "Chang" at the end of the line and the beginning of the next line.
Occur.

Similarly, when "Gaccohe" is input and is cut out correctly, it is possible to shape the character arrangement based on the line lines as shown in FIG. 3 (a). However, when the cutout is incorrect and the “ga” (44) is cut out as two characters (45, 46), it becomes difficult to read because it is shaped as shown in FIG. 3B. In this way, if one character is erroneously separated into two characters, the above-mentioned inconvenience occurs.

For this reason, the clipping parameter may be set so that one character is not mistakenly divided into two characters.
However, with this setting, the opposite occurs.
That is, the phenomenon that two characters are mistakenly regarded as one character increases. For example, when “to school” is input, the cutout is incorrect, and “to haha” is cut out as one character. In this case, in order to modify “to school” to “what is school”, “toha” must be input again after deleting “toha”.

As described above, it is necessary to correct two characters even though one character is corrected. (3). The objects of claims 5 to 8 of the present invention will be described. As mentioned above, without recognizing the character pattern data,
Displaying / storing character by character is considered.

Therefore, it is conceivable to carry or edit the stored character pattern data, and perform character recognition processing again with this data. The present invention proposes an effective character data processing device in such a case. (4). The purpose of claims 9 to 10 of the present invention will be described.

If the character recognition processing is presupposed, is the size of the character written on the writing surface a normal character size? Small font size? Need to judge. However, is it necessary to perform the same processing in the above-mentioned "ink word processor"? The present application proposes a suitable process in "ink word processor" or the like.

(5). The object of claim 11 of the present invention will be described. When writing a character in the frame displayed on the writing surface, is the size of the written character based on the size of the frame the normal size? It is possible to determine whether it is a small character size. Even if no frame is displayed on the writing surface, the size of a normal character can be detected by writing a character string consisting of multiple characters. Font size? It is possible to determine whether it is a small character size.

However, if only one character is written, is this written character a normal character size? Small font size? Can not be judged. Therefore, conventionally, when only one character is written, it is treated as a normal character. For example, FIG.
When "," is input in the display state of,
However, the result is as shown in FIG.

In the present invention, even if only one character is written, is the written character properly in the normal character size? Small font size? The present invention proposes a character data processing device for determining.

[0028]

SUMMARY OF THE INVENTION The present invention is a character recognition conversion method for performing character recognition processing on online handwritten character pattern data to convert it into character code data, in the case where the accuracy of the recognition result is high, the character code data is obtained. If the accuracy of the recognition result is low, it is stored as character pattern data, and the character pattern data is extracted from the stored character pattern data and character pattern data and character recognition is performed again. Characterized by performing conversion into character code data.

According to the present invention, the output arrangement position of the character pattern data is determined based on the character pattern data and the cutout data indicating the certainty of inter-character candidates for separating the character pattern data for each character. A character arranging means (S54) for setting is provided. The present invention relates to a character data processing device for detecting inter-character candidates for separating character pattern data for each character, and changes a detection level at which the inter-character candidates are considered to be between characters according to a processing target. It is characterized by doing.

The character pattern data includes one of character string stroke coordinate data and character image data. The present invention, in a character data processing device for inputting character pattern data and cutting out each character of this character pattern data, when corresponding cutting-out data is input to this character pattern data,
It is characterized in that cutting-out data (S64, S68) is provided with reference to the cutting-out data.

According to the present invention, by character string data editing processing, character spacing candidates designated to be between characters, character spacing candidates before and after a character string in which characters are inserted, or character strings deleted to delete characters. Character spacing candidates with columns packed are
It is characterized in that it is set to data indicating the space between the characters.
The present invention is a display means (20) for displaying character pattern data in units based on cut-out data, and edits the displayed character pattern data in the units, and based on this editing, the related cut-out data. It is characterized by comprising an editing means for modifying the data.

The unit is one character unit. The present invention is a character data processing device that divides character pattern data for each character and displays the character pattern data by normalizing it to a normal character size or a small character size and displaying the determination result of determining the size of each character of the character pattern data. If the degree is high, it is converted to a size according to this detection result, and if the certainty of this determination result is low, a conversion means (S84) for converting it to an intermediate size between this normal character size and small character size is used. It is characterized by being provided.

Further, the converting means (S84) is characterized in that when the certainty factor of the detection result by the size detecting means is low, it is converted into a predetermined intermediate character size and displayed.
According to the present invention, the character pattern data written on the writing surface is
In a character data processing device that divides each character and displays it by normalizing to a normal character size or a small character size,
When it is determined that the written character is a single character on the writing surface, referring to the size of the character written so far,
Size detecting means for detecting the size of this written character
It is characterized by including (S92, S90, S84).

[0034]

BEST MODE FOR CARRYING OUT THE INVENTION

[First Embodiment] A first embodiment according to claim 1 of the present invention will be described with reference to FIGS. FIG. 5 is a block diagram of the character data processing device.

Reference numeral 10 is a stylus pen. 12 is
It's a tablet. Reference numeral 14 is a coordinate detection circuit. 16
Is a control circuit. Reference numeral 18 is a display RAM.

Reference numeral 20 is a display unit composed of an LCD. Two
A ROM 2 stores schedule management software with a character string editing function and character recognition FEP. 24 is R
AM. The operation of this device, which is a drive device unit for writing / reading a floppy disk, will be described with reference to FIGS.

When writing on the tablet 12 with the pen 10, the coordinate detection circuit 14 detects the written coordinates and sends them to the control unit 16. The control unit 16 displays the handwriting with reference to the coordinates. That is, it is displayed as shown in FIG.
When the pen 10 is released for a predetermined time, character recognition processing is performed.

That is, when the control unit 16 detects that the coordinate input has not been made for the predetermined time, the control unit 16 starts the character recognition process using the coordinate input data so far. As in S12 of FIG. 6, first, a character string is cut out to divide it for each character.
Then, as in S14 of FIG. 6, character recognition processing is performed for each character.

As shown in S16 of FIG. 6, if the correctness of the character code data of the character recognition processing result is low, the coordinate data (character pattern data) is output. If the correct answer certainty of the character code data resulting from the character recognition processing is high, this character code data is replaced and output. This output is shown in FIG.
It is stored in the RAM 24 as in S18. Further, this output is displayed on the display unit 20 via the display memory 18.

That is, as shown in FIG. 8, the handwriting data is erased, the character code data is inserted and displayed at the cursor position as character font data, and the coordinate data (character pattern data) is converted and displayed in a size that fits between the lines. It
By the way, the mixed character data in the RAM 24 is
Since it can be transferred from 24 to the FD in the drive unit 26, it can be carried and ported to another character data processing device.

On the contrary, from another character data processing device to the FD
The mixed character data can be ported to the device shown in FIG. By the way, here, the user does not confirm the recognition result. That is, only the character pattern data having a very high correctness certainty is converted into the character code data.

Therefore, the character recognition may be performed again to confirm the recognition result by the user. When the start of this re-recognition operation is instructed, as shown in S32 of FIG.
Of the mixed character data of, only the character pattern data is extracted. Then, as in S34, character recognition processing of this character pattern data is performed. This character recognition process is
The user is requested to confirm the recognition result. Further, naturally, the unrecognizable character remains the character pattern data.

As a result, for example, the display as shown in FIG. [Second Embodiment] A second embodiment of the character data processing device according to claims 2 to 5 of the present invention will be described with reference to FIGS. FIG. 11 is a block diagram of the character data processing device. In FIG. 11, the same parts as in FIG.
The same reference numerals are given.

The operation of this device will be described with reference to FIGS. After writing, when the pen 10 is released for a predetermined time, character recognition processing is performed. That is, when the control unit 16 detects that the coordinate input has not been made for the predetermined time, the control unit 16 starts the character recognition process using the coordinate input data so far. S52 of FIG.
Now, the data for extracting the character string is detected.

In this case, the character peculiarity is not detected by separating the characters for separating the characters one by one, but the "plausibility" between the characters is detected. In other words, the “likelihood” (inter-character certainty data, cut-out data) of the detected inter-character candidates is 100.
If so, it is between characters. Further, if the “plausibility” (inter-character certainty degree data, cut-out data) of the detected inter-character candidates is 20, it is assumed that there is a low possibility that it is between characters.

Then, it is stored in the RAM 24. Also,
The coordinate data (character pattern data) is displayed after being converted to a size that fits between the lines. This display is S5 in FIG.
4 is performed. In S54, the arrangement position of the character pattern data is determined. At this time, the inter-character candidate having a low score of the inter-character certainty degree data is not a character interval,
This is done.

That is, the conventional FIG. 2 (b) and FIG. 3 (b)
The risk of being placed like is reduced. By the way, in this apparatus, when arranging the character pattern data, this arrangement is performed assuming that the inter-character candidate having a low score of the inter-character certainty data is not the inter-character space. However, in the other editing processing (deletion of one character, deletion of range designation, designation of character insertion position, designation of character recognition processing target, etc.) performed in S56, inter-character candidates having a low score of the inter-character confidence data are also processed. , And the process is performed assuming that there is a space between characters.

Therefore, the inconvenience caused by the character segmentation error is reduced. By the way, since the character pattern data and the inter-character certainty data in the RAM 24 can be transferred from the RAM 24 to the FD in the drive unit 26, it can be carried and transferred to another character data processing device.
Further, the character confidence data may not be transferred, and only the character pattern data may be transferred to the FD.

On the contrary, also in the apparatus of FIG. 11, it is possible to transfer the character pattern data and the like from another character data processing apparatus through the FD. Therefore, as shown in FIG. 13, character recognition processing may be performed. That is, cut-out / character recognition is performed by the character pattern data and the inter-character certainty data stored in the RAM 24.

Further, the cutout / character recognition is performed by the character pattern data and the inter-character certainty data stored in the FD. When the data stored in the FD is only the character pattern data and there is no inter-character certainty data, the character cutting process is performed only from the character pattern data.

When inter-character certainty data is present, not only character pattern data but also inter-character certainty data is used as a reference for character recognition. [Third Embodiment] A third embodiment of claims 6 to 8 of the present invention will be described with reference to FIGS.

FIG. 14 is a block diagram of the character data processing device. 14, the same parts as those in FIG. 11 are designated by the same reference numerals. The operation of this device is shown in FIGS.
This will be described with reference to FIG. It is assumed that the character pattern shown in FIG. 15A has already been displayed. The inter-character certainty factor of this character pattern data is shown in FIG.

That is, when this character pattern is subjected to character recognition processing, "12 chicken", "12 chicken 11", "chicken chicken"
"Niwato 11" and the like are considered as recognition candidates. FIG.
Character insertion processing is performed on the character pattern 5 (a) as shown in FIG. 15 (b). Then, as shown in FIG. 16B, the character confidence data on both sides of the inserted character is set to 100.
Change to

In other words, the inter-character certainty degree data changing process is performed based on the editing process. Here, an example in which the inter-character certainty data is changed in connection with the insertion process has been described, but the inter-character certainty data may be changed in the deletion process. [Fourth Embodiment] A third embodiment according to claims 9 to 11 of the present invention will be described with reference to FIGS.

FIG. 17 is a block diagram of the character data processing device. 17, the same parts as those in FIG. 11 are designated by the same reference numerals. The operation of this device is shown in FIGS.
This will be described with reference to FIG. After writing, when the pen 10 is released for a predetermined time, a character pattern is displayed at the cursor position.

First, as in S82 of FIG. 18, a character string is cut out to divide it for each character. Then, the character pattern data is displayed after being converted into a size that fits between the lines. It is assumed that the writing as shown in FIG. 19 is performed.
The second character "tsu" or "tsu" is the normal character "tsu" in FIG. 19 (a), and the small character "tsu" in FIG. 19 (b) (c). There are other characters, "ka"
It can be estimated from the size of "ko" and "u".

However, in the case of FIG. 19D, the determination is difficult. Therefore, as shown in S84 of FIG. 18, the intermediate character size is set. The above description is for the case where a character string of several characters is written. That is, when only one character is written as "tsu", it cannot be estimated from the size of other characters.

Therefore, as in S90 of FIG. 18, the character size input by the pen up to now is regarded as the normal character size, and S
The size of the character is estimated at 84. Naturally, there are three types of estimation results, that is, the normal character size, the small character size, and the intermediate character size. Note that the “character size that has been pen-entered so far” to be referred to in S90 may be the character size of a normal character that was previously input, or may be the average of a plurality of characters that have been pen-inputted so far. .

Further, in the above-described embodiment, the value of the inter-character certainty degree data has a variation range of 0 to 100,
As described in Japanese Patent Application No. 7-14312, there may be two types of positions, that is, inter-character positions that can be determined and ambiguous positions. Further, in the above-described embodiment, the character pattern data is character string stroke coordinate data input by the pen, but may be character image data input to the OCR device.

The character recognition processing in the above embodiment is
Although a simple example is shown, naturally, this may improve the character recognition accuracy by knowledge processing. In addition, although the intermediate character size in the above embodiment is one type,
Multiple types may be provided.

[0061]

According to the first aspect of the present invention, when the character recognition process can be performed without performing the confirmation work and the correction work of the character recognition result, and the character recognition process is insufficient. You can recognize it again later. Claims 2 to 4 of the present invention
According to this, it is possible to provide a character data processing device in which the operability is less deteriorated even if the detection accuracy between characters by the character string cutout process is not high.

According to claims 5 to 8 of the present invention, when character recognition is performed on character pattern data, if character confidence data corresponding to the character pattern data is attached,
Characters can be cut out with reference to this. According to claims 9 to 10 of the present invention, is the size of the character written on the writing surface a normal character size? Small font size? Even if the determination accuracy of is not high, it can be displayed without any trouble.

According to the eleventh aspect of the present invention, even when only one character is written, the size of this character can be accurately determined.

[Brief description of drawings]

FIG. 1 is a diagram for explaining an operation of a conventional character data processing device.

FIG. 2 is a diagram for explaining the operation of the character data processing device.

FIG. 3 is a diagram for explaining the operation of the character data processing device.

FIG. 4 is a diagram for explaining the operation of the character data processing device.

FIG. 5 is a block diagram of a character data processing device according to the first embodiment of the present invention.

FIG. 6 is a flow chart for explaining the operation of the first embodiment.

FIG. 7 is a diagram for explaining the operation of the first embodiment.

FIG. 8 is a diagram for explaining the operation of the first embodiment.

FIG. 9 is a flow chart for explaining the operation of the first embodiment.

FIG. 10 is a diagram for explaining the operation of the first embodiment.

FIG. 11 is a block diagram of a character data processing device according to a second embodiment of the present invention.

FIG. 12 is a flow chart for explaining the operation of the second embodiment.

FIG. 13 is a flow chart for explaining the operation of the second embodiment.

FIG. 14 is a block diagram of a character data processing device according to a third embodiment of the present invention.

FIG. 15 is a diagram for explaining the operation of the third embodiment.

FIG. 16 is a diagram for explaining the operation of the third embodiment.

FIG. 17 is a block diagram of a character data processing device according to a fourth embodiment of the present invention.

FIG. 18 is a flow chart for explaining the operation of the fourth embodiment.

FIG. 19 is a diagram for explaining the operation of the fourth embodiment.

[Explanation of symbols]

 12 ... Tablet, 20 ... LCD, 16 ... Control unit, 24 ... RAM.

Claims (11)

[Claims] 1. A character recognition conversion method in which online handwritten character pattern data is subjected to character recognition processing and converted into character code data. In the case where the correctness of the recognition result is high, the character recognition data is converted into character code data and stored. If the accuracy of the correct answer is low, store it as character pattern data, extract the character pattern data from the stored character pattern data and character pattern data, perform character recognition again, and convert it to character code data. Character recognition conversion method characterized by. 2. The output arrangement position of the character pattern data is set based on the character pattern data and the cutout data indicating the certainty of inter-character candidates for separating the character pattern data for each character. Character arrangement means (S54)
A character data processing device comprising: 3. A character data processing device for detecting inter-character candidates for separating character pattern data for each character, wherein a detection level at which the inter-character candidates are considered to be between characters is determined according to a processing target. A character data processing device characterized by being changed. 4. The character pattern data is any one of character string stroke coordinate data and character image data.
4. The character data processing device according to claim 2, wherein the character data processing device comprises: 5. In a character data processing device for inputting character pattern data and cutting out the character pattern data for each character, when the corresponding cutting-out data is input to the character pattern data, With reference to the cutting data,
A character data processing device comprising a cutting-out means (S64, S68) for cutting out. 6. An editing process of character pattern data,
A character spacing candidate designated as being between characters, a character spacing candidate before and after a character string in which a character has been inserted, or a character spacing candidate in which a character string has been deleted and packed into a character string is data that indicates this character spacing. A character data processing device characterized by being set to. 7. Display means (20) for displaying character pattern data in units based on cut-out data, and editing the displayed character pattern data in the units, and related cut-outs based on this editing. A character data processing device comprising an editing means for correcting data for use in character data processing. 8. The character data processing device according to claim 7, wherein the unit is one character unit. 9. A character data processing device that divides character pattern data for each character and normalizes and displays it in a normal character size or a small character size. If the certainty is high, it is converted to a size according to this detection result, and if the certainty of this determination result is low, it is converted to an intermediate size between this normal character size and small character size (S84). A character data processing device comprising: 10. The character data according to claim 9, wherein the conversion unit (S84) converts the display result into a predetermined intermediate character size and displays it when the certainty factor of the detection result by the size detection unit is low. Processing equipment. 11. A character data processing device for dividing character pattern data written on a writing surface for each character and normalizing and displaying the character pattern data to a normal character size or a small character size, wherein the characters written on the writing surface are If it is determined that it is a single character, refer to the size of the characters written so far,
Size detecting means for detecting the size of this written character
A character data processing device comprising (S92, S90, S84).