JPH0896082A - Character recognizing device and character recognizing method - Google Patents

Abstract

PURPOSE: To improve the character recognition rate by correcting input data by using plural correcting systems. CONSTITUTION: In the character recognizing device recognizing the coordinate information inputted within an input frame via coordinate input devices 1 and 2 as character information, a data correction part 33 has plural correcting systems and corrects input coordinate information by adding the importance information set for every correcting system. A character recognizing part 35 performs a character recognizing processing by using a plurality of correction coordinate information obtained at the time and establishes a candidate order from all the recognition candidates. According to the next candidate selection instruction, a correction system selection part 35 changes the importance information for every correcting system which is stored in a correcting system buffer 374.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognizing device and a character recognizing method used in a handwriting input type document creating device or a handwriting editor for recognizing characters handwritten in an input frame to create a document.

[0002]

2. Description of the Related Art A device in which a display device and a coordinate input device are laminated and integrated makes it possible to input characters and figures on paper as if they were being written, and has recently come to be used in various fields. Along with this, even in a document creating device or the like that conventionally uses a keyboard as an input device, a stylus pen is used to input information in the form of coordinate point sequence on a liquid crystal tablet, and the information is recognized as characters. Came.

However, the recognition accuracy of characters is not yet sufficient, and researches for improving the recognition accuracy are still ongoing. At the same time, it has been attempted to improve the character recognition accuracy by devising an input interface.

For example, an input frame for one character is provided on the screen,
If you enter characters by handwriting in the input box, 1
It is not necessary to cut out a character for each coordinate information of the character. Furthermore, if there is an input frame, by inputting small characters in the frame, it becomes easy to recognize as kana / alphabet lowercase letters, and it is possible to distinguish between a period and a middle point. Providing an input frame in this way is adopted by many systems as a means for improving recognition accuracy.

[0005]

As described above, conventionally, the input frame is provided to improve the accuracy of character recognition. However, on the other hand, when inputting characters at the boundary of the input box, compared to the conventional free format input method, there is too much consciousness that the character must be input inside the input box. The closer you get to the part, the more characters you can type, etc.
In many cases, the deformation of the character shape becomes severe, and the character recognition rate decreases.

The present invention has been made in consideration of such circumstances. When a character is input in an input box, it is checked how close it is to the vicinity of the boundary of the input box, and the coordinates of the vicinity of the boundary of the input box are examined. The larger the point, the larger the coordinate correction,
To provide a document creation apparatus and a character recognition method capable of improving a character recognition rate by correcting input data using a plurality of correction methods such as a method in which coordinate correction is increased toward the end of a stroke. To aim.

[0007]

A document creating apparatus according to the present invention has coordinate input means for inputting coordinate information in a predetermined area and a plurality of correction methods, and is input by the coordinate input means. A character recognizing process is performed by using a correction unit that corrects the coordinate information in consideration of importance information set for each of the correction methods, and each of the plurality of corrected coordinate information obtained by the correction unit. A character recognizing means, and a candidate rank determining means for deciding an output rank for a plurality of recognition candidate information obtained by the character recognizing means,
The candidate output means outputs the recognition candidate information in accordance with the output order determined by the candidate order determining means.

Further, according to the present invention, in response to a selection instruction of each of the recognition candidate information output by the candidate output means,
And a correction method selecting means for changing the importance information set for each of the correction methods.

[0009]

According to the above configuration, the input coordinate information is corrected by separately using a plurality of correction methods, the corrected coordinate information is character-recognized, and the recognition candidate rank is determined from all the results. Furthermore, by setting the importance for each correction method and providing the optimum correction method according to the input person,
The character recognition accuracy can be improved as compared with the conventional case.

[0010]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a diagram schematically showing an external configuration of a character recognition device according to an embodiment of the present invention. This device comprises a transparent tablet 1, a stylus pen 2, a control device 3, a display device 4, and an external storage device 5.

The transparent tablet 1 and the stylus pen 2 are
By pointing the transparent tablet 1 with the stylus pen 2, which is a component of the coordinate input device, coordinate information corresponding to the pointing position is input.

The control device 3 is composed of, for example, a microprocessor and controls the entire device. Here, the correction of the information of the two-dimensional coordinate point sequence input by the coordinate input device (1, 2) is performed. Performs processing and character recognition processing.

The display device 4 is composed of, for example, a liquid crystal display, and displays the position designated by the stylus pen 2 and the recognition result. As the display device 4, a plasma display or the like can be used in addition to the liquid crystal display. Here, a display device 4 including a liquid crystal display
The transparent tablet 1 is laminated and integrated. That is, the liquid crystal display and the transparent tablet 1 are formed to have the same coordinate plane with the same size, and the information displayed on the liquid crystal display can be visually recognized through the transparent tablet 1. By the transparent tablet 1 and the liquid crystal display which are laminated and integrated in this way, the coordinate pointing position on the transparent tablet 1 is displayed as information at the same position on the liquid crystal display, for example, to draw a character or figure on paper. You can enter information with a simple feeling.

The external storage device 5 is composed of, for example, a floppy disk device, a hard disk device, etc., and stores various information necessary for character recognition processing such as a character recognition dictionary 51.

FIG. 2 is a block diagram showing the internal structure of the character recognition device in the embodiment. The control device 3 includes an initial setting unit 31, an input unit 32, a data correction unit 33, a character recognition unit 34, a correction method selection unit 35, a display control unit 36, and a storage unit 3.
It consists of 7 etc.

The storage unit 37 has a coordinate data buffer 371 for temporarily storing information of the coordinate point sequence input from the coordinate input device (1, 2), coordinate information corrected for each of a plurality of correction methods ( Correction data) 372 for temporarily storing correction data), a character candidate buffer 373 for storing the result of character recognition of each correction coordinate information (correction data) together with an evaluation value, and each correction method Correction method buffer 3 for storing importance
74, a display data buffer 375 for temporarily storing data to be displayed on the display device 4, an image of an initial screen displayed at the time of system initialization, and the input frame 6 shown in FIG.
1, a screen image data section 376 storing coordinate information of the next candidate display area 62.

The initial setting section 31 carries out initialization processing of screen display, various buffers, counters and the like. The input unit 32 saves the coordinate information input from the coordinate input device (1, 2), determines whether to finish inputting one character / one character, and determines the next candidate.

The data correction unit 33 is a feature of the present invention, and uses a plurality of correction methods prepared in advance to separately correct the input coordinate information in consideration of the importance set for each correction method. Then, the coordinate information is converted into relative coordinates for character recognition and stored in the coordinate data buffer 372. The character recognition unit 34 performs character recognition processing on the corrected coordinate information for each correction method in the coordinate data buffer 372 by referring to the character recognition dictionary 51 stored in the external storage device 5.

The correction method selection unit 35 is also one of the features of the present invention, and sets the importance of each correction method in the data correction unit 33. Here, there are two methods of setting the importance, that is, a method of using the initial value as it is and a method of changing the importance according to the selection of the next candidate. The display control unit 36 performs control for displaying the recognition result of the character recognition unit 34 on the display device 4.

The external storage device 5 also stores a character recognition dictionary 51 used in the character recognition process. As shown in FIG. 3, the character recognition dictionary 51 stores information indicating characters, character codes, stroke numbers, and the start point, middle point, and end point of one image in relative coordinates with the start point of the first image as a reference. ing.

Next, the operation of the embodiment will be described. As shown in FIG. 4, an input frame 61 is provided on the transparent tablet 1, and the stylus pen 2 is used to input characters by handwriting. When the input of handwritten characters is completed, the recognition button 63 is touched. As a result, the handwriting data in the input frame 61 is recognized as a character, the first candidate of the recognition result is displayed in the display area 60 at the upper part of the screen, and the next and subsequent candidates are displayed in the next candidate display area 62. It should be noted that 64 is an end button for instructing the end.

When the next candidate is displayed in the next candidate display area 62 and the input person selects the next candidate, the characters in the display area 60 are replaced with the characters of the next candidate. This time,
Internally, the importance value of the correction method corresponding to the next candidate becomes high.

5 and 6 are flow charts showing the operation of this embodiment. At the beginning of the process, the initial setting section 31 clears various buffers, counters, etc., and displays the image data for the initial screen stored in the screen image data section 376 on the display device 4 (step 501).

Here, when the input person performs handwriting input of characters using the stylus pen 2 (step 502), first, the input section 32 checks the coordinate values of the screen data held in the screen image data section 376. Then, it is checked whether or not the current input operation is the next candidate selection operation (step 503).

If the input operation is a character input operation instead of the next candidate selection operation, the input unit 32 looks at the on / off state of the stylus pen 2 to check whether or not one stroke is being input (step 504). If one stroke is being input, the input unit 32 stores the input coordinate information at that time in the coordinate data buffer 371 (step 505). Coordinate data buffer 371
In FIG. 7, the x-coordinate and y-coordinate values and an image separator representing the end of input of one image are stored.

At this time, the processing shifts to the display control section 36,
The input coordinate information is output to the display device 4 via the display data buffer 375 (step 506). Further, the processing shifts to the data correction unit 33. The data correction unit 33 executes several kinds of correction methods one by one from the input coordinate information and the coordinate information of the input frame 61 stored in the screen image data unit 376 to obtain corrected coordinate information (correction data) for the input coordinate information. Is calculated (step 507).

Here, in the present embodiment, the data correction section 33 is assumed to perform the following five correction methods. First
In the correction method of No. 1, no correction is performed and the value of the first point of the first image is converted into relative coordinates with the origin.

The second correction method obtains a value obtained by multiplying the distance from the center point of the input frame 61 to the input point by a constant and the importance of the second correction method stored in the correction method buffer 374.

FIG. 8 shows a concrete example of the second correction method. It is checked how much the value of the x coordinate of the input point is from the center point of the input frame 61. In the example of FIG. 8, the input frame 6
Since the x coordinate of the center point of 1 is "40" and the input point is "25", the shift is "-15". The deviation is multiplied by a constant weight (0.1). Further, the value in the correction method buffer 374 is multiplied. Correction method buffer 37
As shown in FIG. 9, 4 stores the degree of importance for each correction method. Since the importance of the second correction method is "0.99", the correction value of the input coordinate is "-1.485" (-
15 × 0.1 × 0.99). After adding this value to the coordinates of the input point, it is converted into relative coordinates with the value of the first point of the first stroke as the origin.

The third correction method is a function in which the weight of the second correction method is not a constant but a function.
As shown in 0 (a), it is a quadratic function. As shown in FIG. 9A, the correction value of the third correction method is “−8.269”.
become. After adding this value to the coordinates of the input point, it is converted into relative coordinates with the value of the first point of the first stroke as the origin.

In the fourth correction method, weighting is applied according to the cumulative number of samplings from the starting point, and the value is multiplied by the degree of importance. In the example of FIG. 10B, the correction value is set to “0.883” with the value obtained by multiplying the number of sampling points from the starting point by the maximum cumulative sampling number as the weight. After adding this value to the coordinates of the input point, it is converted into relative coordinates with the value of the first point of the first stroke as the origin.

The final fifth correction method is the third and fourth correction methods described above.
In the example shown in FIG. 10C, the correction value is calculated as “−−7.379”. After adding this value to the coordinates of the input point,
Converts the point value to relative coordinates with the origin.

Each of the above-mentioned correction methods checks the position in the input frame 61 where the coordinate information is input, and determines the degree of correction of the coordinate information according to the input position. In particular, in the second and third correction methods, when determining the degree of correction of the coordinate information, the degree of correction is increased toward the boundary of the input frame 61. Further, in the fourth correction method, when determining the degree of correction of the coordinate information, the degree of correction is increased toward the end of the stroke. In the fifth correction method, when determining the degree of correction of coordinate information, the degree of correction is increased near the boundary of the input frame 61 and near the end point of the stroke.

Then, the correction data obtained in step 507 is stored in the correction data buffer 372 (step 508). Since steps 507 and 508 are performed for each of the correction methods one by one, the data correction unit 33 checks whether or not the five correction methods have been completed (step 509).
If all two correction methods have been completed, go to step 502.
If not completed, the process returns to step 507 to obtain the next correction method.

On the other hand, if one stroke is not being input in step 504, the input unit 32 continues to
It is checked whether or not the character input is completed (step 51).
0). The judgment of the end of the input of one character is made by the recognition button 64 in FIG.
It is determined by whether or not is touched. If the recognition button 64 is not touched, the input of one stroke is completed, so the input unit 32 sets 1 in the coordinate data buffer 371.
The image separator of the input of the fraction is stored (step 51
1) Further, after the cumulative sampling number from the starting point used in the fourth and fifth correction methods is set to "0" (step 51).
2) Return to step 502.

When the input unit 32 determines in step 510 that the input of one character is completed, the process proceeds to the character recognition unit 34. The character recognition unit 34 performs the following character recognition processing on the correction data obtained by each correction method (step 513).

The character recognition unit 34 first extracts several representative points (start point / middle point / end point in this embodiment) per data of one stroke, and processes the representative points. Further, the registered pattern of the character recognition dictionary 51 having the same number of strokes as the input data stored in the external storage device 5 is used.

The character recognition unit 34 calculates the distance between the relative coordinate information obtained in step 507 and each coordinate point of the start point, middle point and end point of one stroke. Next, the total value of the distances of the coordinate points is obtained, and the total value is divided by the score to calculate the evaluation value. Then, the two candidates with the smallest evaluation value are stored in the character candidate buffer 373.

In the character candidate buffer 373, as shown in FIG. 11, the character code of the recognition candidate, the evaluation value, and the information indicating which correction method was used are stored. When the character recognition process for one correction data is completed in step 513, the character recognition unit 34 checks whether or not the character recognition process is completed for all the correction data obtained by the five correction methods (step 514). If character recognition has not been completed for all the correction data, step 5
Return to 13.

When the character recognition processing is completed for all the correction data, the character recognition unit 34 extracts four candidates from the recognition candidate group of the character candidate buffer 373 as the final recognition candidates from the smallest evaluation value (step 515). In this embodiment, as shown in FIG. 11, “A”, “A”, “G”,
“O” is extracted.

Thereafter, the processing shifts to the display control section 36. The display control unit 36 erases the handwriting displayed in the input frame 61 (step 516), and displays the recognition result via the display data buffer 375 (step 517). In this case, the first candidate (“A”) is displayed in the display area 60 of FIG. 4, and the second and subsequent candidates (“AA”, “G”, “O”) are displayed in the next candidate display area 62 of FIG. 4. To be done. Then, the initial setting unit 31 clears the value in the coordinate data buffer 371 (step 518) and returns to step 502.

When the input person gives an instruction to select the next candidate in step 503, the input unit 32 checks which next candidate has been given (step 519). For example, assume that the third candidate “G” is selected. In that case, the process moves to the display control unit 36, and the display control unit 36 replaces the first candidate character “A” and the next candidate “G” displayed in the display area 60 of FIG. 4 (step 520).

After that, the processing shifts to the correction method selection section 35. The correction method selection unit 35 determines which correction method the instructed next candidate was obtained by using the character candidate buffer 37.
The value of 3 is checked (step 521), and the importance of the correction method is increased (step 522).

As a concrete example, when the importance of each correction method is set as shown in FIG.
When the recognition candidate character “G” in the correction method of No. 1 is selected as the next candidate, the importance of the fourth correction method becomes “0.01” points higher and the other importance becomes “0.01” points lower. The changed importance value is shown in FIG. After the value of the correction method buffer 374 is changed in this way, the process returns to step 502.

The present embodiment is not limited to the above embodiment. For example, in this embodiment, five correction methods are shown as the plurality of correction methods. However, other correction methods may be used, and in essence, a plurality of correction methods may be used.

Further, in this embodiment, the value of the correction method buffer 374 is changed at the timing of selecting the next candidate. However, not only the timing of selecting the next candidate, but the importance of the first candidate when the recognition candidate is output, for example. You may also increase the degree.

In the present embodiment, the character recognition processing is performed by matching the distance between the registered character of the character recognition dictionary 51 and the input coordinates, but the character is abstracted into a certain basic form and Other character recognition methods may be used as long as the evaluation value can be calculated, such as matching with the shape. In addition, the data structure of the character recognition dictionary 51 may be changed accordingly. In short, the present invention can be implemented with various modifications without departing from the scope of the invention.

[0048]

As described above, according to the present invention, a plurality of correction methods are prepared in the input character data correction processing performed as the preprocessing of character recognition, and the importance set for each correction method is set. The input data is corrected in consideration of each character, each corrected data is character-recognized, and the candidate rank is determined from all the character candidates, so that the character recognition accuracy is improved.

Furthermore, since the coordinate points are corrected by weighting the coordinate points input near the boundary of the input frame or weighting the stroke end, the coordinate points are corrected near the boundary of the input frame. It is possible to correct characters as if they were entered in a free format method without considering the input frame. As a result, the input person can create a document accurately and quickly, and a great effect is practically exerted.

[Brief description of drawings]

FIG. 1 is a diagram schematically showing an external structure of a character recognition device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an internal configuration of the character recognition device.

FIG. 3 is a diagram showing a data structure of a character recognition dictionary in the embodiment.

FIG. 4 is a diagram showing an initial screen in the embodiment.

FIG. 5 is a flowchart showing the operation of the embodiment.

FIG. 6 is a flowchart showing the operation of the embodiment.

FIG. 7 is a diagram showing a data structure of a coordinate data buffer in the embodiment.

FIG. 8 is a diagram for explaining a second correction method in the embodiment.

FIG. 9 is a diagram showing a data structure of a correction method buffer in the embodiment.

FIG. 10 is a diagram for explaining third to fifth correction methods in the same embodiment.

FIG. 11 is a view showing a data structure of a character candidate buffer and a candidate order in the embodiment.

FIG. 12 is a diagram showing the contents of a correction method buffer after selecting a next candidate in the embodiment.

[Explanation of symbols]

1 ... Transparent tablet, 2 ... Stylus pen, 3 ... Control device, 4 ... Liquid crystal display, 5 ... External storage device, 31 ...
Initial setting section, 32 ... Input section, 33 ... Data correction section, 34
... character recognition section, 35 ... correction method selection section, 36 ... display control section, 37 ... storage section, 371 ... coordinate data buffer, 37
2 ... Coordinate data buffer, 373 ... Character candidate buffer,
374 ... Correction method buffer, 375 ... Display data buffer, 376 ... Screen image data section, 51 ... Character recognition dictionary.

Claims (13)

[Claims] 1. A coordinate input means for inputting coordinate information in a predetermined area, and a plurality of correction methods, wherein the coordinate information input by the coordinate input means is provided for each of the correction methods. A correction unit that adds and corrects the set importance information, a character recognition unit that performs a character recognition process using each of the plurality of correction coordinate information obtained by the correction unit, and a character recognition unit that performs the character recognition process. And a candidate output means for outputting the above-mentioned recognition candidate information in accordance with the output rank determined by the candidate rank determining means. Character recognizer. 2. A coordinate input means for inputting coordinate information in a predetermined area, and a plurality of correction methods, wherein the coordinate information input by the coordinate input means is provided for each of the correction methods. A correction unit that adds and corrects the set importance information, a character recognition unit that performs a character recognition process using each of the plurality of correction coordinate information obtained by the correction unit, and a character recognition unit that performs the character recognition process. Candidate rank determining means for determining the output rank for the plurality of recognition candidate information, candidate output means for outputting each of the recognition candidate information according to the output rank determined by the candidate rank determining means, and this candidate output means for outputting. And a correction method selecting means for changing the importance information set for each of the correction methods in response to a selection instruction of the recognition candidate information. Character recognition device, characterized in that the. 3. Each of the correction methods is characterized in that a position in the area where the coordinate information is input is checked and a degree of correction of the coordinate information is determined according to the input position. The character recognition device according to claim 1 or 2. 4. The character recognition device according to claim 3, wherein when determining the degree of correction of the coordinate information, the degree of correction is increased toward the boundary of the area. 5. The character recognition device according to claim 3, wherein when determining the correction degree of the coordinate information, the correction degree is increased toward the end point of the stroke. 6. The character according to claim 3, wherein when the degree of correction of the coordinate information is determined, the degree of correction is increased in the vicinity of the boundary portion of the area and in the vicinity of the end point of the stroke. Recognition device. 7. A character recognition method of a character recognition device for recognizing coordinate information input in a predetermined area through a coordinate input device as character information, wherein the character recognition method has a plurality of correction methods, A character recognition method characterized by performing correction by adding importance information set for each correction method and using a plurality of correction coordinate information obtained at that time. 8. Each of the correction methods is characterized in that the position in the area where the coordinate information is input is checked and the degree of correction of the coordinate information is determined according to the input position. The character recognition method according to claim 7. 9. The character recognition method according to claim 8, wherein, when determining the degree of correction of the coordinate information, the degree of correction is increased toward the boundary of the area. 10. The character recognition method according to claim 8, wherein when determining the degree of correction of the coordinate information, the degree of correction is increased toward the end of the stroke. 11. The character according to claim 8, wherein when the degree of correction of the coordinate information is determined, the degree of correction is increased in the vicinity of the boundary portion of the area and in the vicinity of the end point of the stroke. Recognition method. 12. The coordinate information is input within a predetermined area, and a plurality of correction methods are provided, and the input coordinate information is corrected in consideration of importance information set for each of the correction methods, Character recognition processing is performed using the plurality of corrected coordinate information obtained for each of the correction methods, the output order for the plurality of recognition candidate information obtained at this time is determined, and the above-described recognition candidate information is output according to the output order. A character recognition method characterized by outputting. 13. The coordinate information is input within a predetermined area, and a plurality of correction methods are provided, and the input coordinate information is corrected in consideration of importance information set for each of the correction methods, Character recognition processing is performed using the plurality of corrected coordinate information obtained for each of the correction methods, the output order for the plurality of recognition candidate information obtained at this time is determined, and the output of each of the output recognition candidate information is performed. A character recognition method, characterized in that the importance information set for each of the correction methods is changed according to a selection instruction.