JPH0749924A - Handwritten character recognizing device - Google Patents

Abstract

PURPOSE:To reduce an operator's burden at the time of recognizing a character by recording the information related to correction as correction history for every writer of a handwritten character, and recognizing the character by referring to not only a character dictionary but also the correction history by a character recognizing means. CONSTITUTION:In an updating processing of correction history 4, with regard to a handwritten character 1 of a prescribed writer, in a primary recognition processing, when the code data 13 of A is outputted, in a secondary character recognition processing, a description for instructing to correct it to the code data 13 of B and to output it is stored in a field of the written code data A in the correction history 4. When details are also stored together with additional information 15, it is useful for deciding the case when a machine executes a secondary character recognition processing. For instance, in a character recognition process of reliability as a result of character recognition and a corrected character, this data is code data, etc., of other character being a candidate of the result of character recognition, and they are displayed on a display device 71 in accordance with an operator's request, by which the operator's load can be reduced.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a handwritten character recognition device, and more particularly to a handwritten character recognition device for efficiently recognizing handwritten characters by the same writer.

[0002]

2. Description of the Related Art A character dictionary used in a conventional handwritten character recognition device stores information describing characteristics of image data of each character when the character is normally handwritten. Therefore, regarding image data of characters obtained by photoelectrically converting handwritten characters, the coordinate position of feature points such as end points and inflection points in the image data, distances between feature points, structural features such as protrusions and depressions, etc. Are extracted, and these characteristics are compared with the characteristics of each character described in the character dictionary to determine several kinds of characters as candidates for the character represented by the image data.

Next, detailed data such as the inclination angle of the straight line portion of the image data and the degree of curvature of the curved line portion,
One of the most probable characters is determined from a plurality of candidate characters by collating with the data described in the character dictionary, and the image data of the character is converted into the code data of the character.

When this conversion is completed, the image data before conversion and the image data read out (from the font memory) by using the code data after conversion are compared and displayed on the display device, and this is observed. If the operator determines that the correction is necessary, the operator inputs the necessary correction, and correct code data corresponding to the image data before conversion is stored.

[0005]

As described above, in the conventional handwritten character recognition device, the operator inputs a correction each time an erroneous recognition occurs in the character recognition by the machine, but the handwritten character is the writer (operator). If there is a characteristic of the individual, that is, a characteristic peculiar to the individual (so-called writing habit), and the machine misrecognizes because of the characteristic peculiar to this individual, the machine misrecognizes each time the same character appears and the operator recognizes it. Since the same correction is repeatedly input, there is a problem that the operator's burden is heavy.

An object of the present invention is to solve the above-mentioned problems in the conventional apparatus, and to provide a handwritten character recognition apparatus in which an operator only needs to make a correction input once for the same character handwritten by the same person. The purpose is to do.

The invention entitled "Handwritten character recognition device" (hereinafter referred to as the prior invention) disclosed in Japanese Patent Application Laid-Open No. 3-166677 describes that "the character cannot be recognized due to individual differences in character style, stroke order, time variation, and the like. Or, when it becomes difficult, a user character recognition dictionary is generated from input handwriting data, and by making this a database, learning of character recognition for input handwriting is disclosed. However, in this prior invention, it is necessary to input handwriting data having handwritten character information, that is, data in stroke order is required, and like the present invention, only handwritten character image data corresponds to code data corresponding to the image data. The problem remains that it cannot be applied to a character recognition device that converts to.

[0008]

According to the present invention, a correction history referred to by a character recognition means for determining code data of a character from image data of a handwritten character together with a character dictionary is provided, and the correction history is provided for each writer of a handwritten character. Edit to
Characters peculiar to the author were stored for each character. Further, when the operator corrects the character code determined by the machine, the content of the correction is registered in the correction history so that the same correction for the same character is not necessary thereafter.

[0009]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention. Image information 11 represented by handwritten characters 1 is converted into image data 12 by an image input device 2 and input to a character recognition device 5. The character recognition device 5 refers to the character dictionary 3 and the correction history 4 and converts the image data of the character into code data of the character and outputs the code data (the output destination is not shown). In this conversion, whether or not the conversion was performed correctly,
It is displayed on the display device 71 of the interactive input device 7. When the operator determines that the correction is necessary by looking at this display, the necessary correction is input from the input device 7. When the correction is input, the correction program 6 is activated according to the input, and the contents of the correction history 4 are updated.

The character recognition device 5 reads out the contents of the image memory 51 which temporarily stores the image data 12 before conversion and refers to the character dictionary 3 and the correction history 4 to convert the image data into the code data 13. Means 52,
Code memory 5 for temporarily storing this code data 13
3. An image generating means 54 for reading the code data stored in the code memory 53 to generate the image data 14 corresponding to the character code, and an image memory 55 for temporarily storing the image data 14. The image generating means 54 for generating the image data of the character from the code data of the character is composed of, for example, a font memory.

Next, the operation of the apparatus shown in FIG. 1 will be described.
FIG. 2 is a flow chart showing the operation of the apparatus of FIG. 1, and 21 to 26 show respective steps. If the content of the correction history 4 is blank in the initial state, the operation of the character recognition device 5 is the same as that of the conventional device, and the process proceeds from step 23 to step 25. If there is a correction history regarding the character of the writer, the code data determined in step 22 is corrected in step 24 or it is determined that the correction is unnecessary. Code data 13 determined in this way
Is stored in the code memory 53, and the code data output process of step 25 is performed.

In the code data output process 25, the image data stored in the image memory 51 and the image data stored in the image memory 55 are displayed on the display device 71 in comparison with each other. When the operator determines that the displayed image data are essentially the same and inputs the determination result, the code data stored in the code memory 53 is output as correct code data in the output memory (see FIG. Although not shown), when the operator determines that the correction is necessary, the correction program 6 is activated.

FIG. 3 shows the contents of the character recognition additional information output processing in step 26 of FIG. In FIG. 3, 31 to 35
Is each step. Step 31 is judged by the input from the operator, and if there is a correction point, step 32
Thus, the code data correction process is performed. In this case, the operator inputs the correct code data from the keyboard of the input device 7. When inputting code data such as kanji, the input device must have a kana-kanji conversion function. After completing step 32, steps 33 to 3
5, the correction history 4 is updated.

In the correction process of updating the correction history 4, when the code data of A is output in the primary recognition process of step 22 regarding the handwritten character of the predetermined writer, this is detected in the secondary character recognition process 24. It suffices to store in the correction history 4 the description of correcting and outputting B as code data in the code data A column of the writer. However, if more detailed additional information is also stored, it is useful for the determination when the machine executes the secondary character recognition processing 24. For example, the reliability as the character recognition result and the code data of other characters which are candidates for the character recognition result in the character recognition process of the corrected character are displayed on the display device 71 according to the operator's request. This reduces the operator's load.

[0015]

As described above, according to the present invention, every time the correction is performed by the operator, the information related to the correction is recorded as the correction history for each writer of the handwritten character,
Since the character recognition means performs character recognition by referring to not only the character dictionary but also the correction history, the operator's burden in character recognition can be significantly reduced.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a flowchart showing the operation of the present invention.

FIG. 3 is a flowchart showing details of character recognition additional information output processing of FIG.

[Explanation of symbols]

 1 handwritten character 2 image input device 3 character dictionary 4 correction history 5 character recognition device 6 correction program 7 interactive input device 12 image data before conversion to code data 13 code data 14 image data read from code data 15 additional information 71 display apparatus

Claims (4)

[Claims] 1. A conversion result obtained by converting image data of a handwritten character into code data of the character corresponding to the image data with reference to a character dictionary that stores information describing characteristics of image data of each character. In the handwriting character recognition device in which the operator inputs a correction for the conversion result when the operator who has observed the conversion result is judged to be unsuitable, the image data of each character handwritten by the writer edited for each writer When the image data of the handwritten character is converted into the code data of the character corresponding to the image data, the correction history storing the characteristic peculiar to the writer is provided.
With reference to the character dictionary and the correction history, each time the operator performs the correction, a correction program is activated, and a necessary correction is recorded in the matter related to the character of the writer in the correction history. A handwritten character recognition device. 2. The features of the image data of each character stored in the character dictionary include the feature point positions of feature points such as end points and inflection points, the distance between feature points, structural features such as convex portions and concave portions, angles, and the like.
The handwritten character recognition device according to claim 1, further comprising detailed data such as a degree of curvature. 3. The conversion result observed by the operator is
The handwritten character recognition device according to claim 1, wherein the image data before being converted into the character code data and the image data read by the converted code data are displayed on the display device. . 4. The correction history is registered including the reliability as a character recognition result and information on other characters which are candidates for the character recognition result in the character recognition process of the corrected character. The handwritten character recognition device according to claim 1, wherein the handwritten character recognition device is a character recognition device.