JPH08339415A - Character input device - Google Patents

Abstract

PURPOSE: To eliminate the need to input writing data and lighten burden on an operator by enabling the operator to correct an error in the segmentation of a segmentation process. CONSTITUTION: When writing data of plural characters are inputted through an input means 101, the segmentation process is performed and the writing data are segmented into the characters. Then feature points are extracted from writing data of one character and displayed by a feature point display process. Here, characters are displayed with character spacing at each segmentation position as the result of the segmentation process. Through a feature editing process, the error in the segmentation of the segmentation process is corrected by using the displayed feature points. After the correction, a dictionary collating process is performed and the recognition result is displayed.

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 having an online handwritten character recognition function for converting a handwritten character string handwritten into a character code.

[0002]

2. Description of the Related Art A computer is equipped with a tablet,
There is a device that converts a character handwritten on an tablet by an operator into a character code by a character recognition process and inputs the character code. According to this device, it is possible to input characters by handwriting, and even an operator unfamiliar with the keyboard can easily
You can enter characters. Further, the tablet can be miniaturized, which is effective in saving space.

The handwritten character recognition processing generally has a recognition dictionary, and the recognition dictionary has a standard handwritten pattern for one character (hereinafter referred to as "standard writing pattern"). It is data stored in combination with a character code. In the handwritten character recognition process, when one character of writing data is input, the recognition dictionary is referenced, the writing data of the operator is compared with the standard pattern of the recognition dictionary, and the character of the most similar standard pattern is recognized. Output the code as a recognition result.

Further, the writing data input by the operator may not be one character, but a plurality of characters may be continuously input.
In this case, as a pre-process of the handwritten character recognition process, a process of dividing the writing data for a plurality of characters into individual characters (hereinafter, referred to as “segmentation process”) is performed.
The handwritten character recognition process was performed after dividing into writing data for characters.

[0005]

However, in the segmentation processing, there are cases in which character breaks are erroneously determined. In the handwritten character recognition processing, the processing is performed based on the handwritten data that is erroneously separated, so that a recognition result that is completely different from the operator's intention is generated. In this case, the operator has to input the writing data again. Further, if an error occurs in the segmentation due to the segmentation process, the next character will also have an incorrect recognition result, and therefore the operator often re-inputs the writing data for a plurality of characters.

As described above, the segmentation error in the segmentation process has been a heavy burden on the operator.

[0007] The present invention has been made in view of the above points,
The purpose of this is to enable the operator to correct the segmentation error in the segmentation process, thereby avoiding the need to re-input the writing data and reducing the burden on the operator.

[0008]

In order to achieve the above object, a character input device according to the present invention has an input means for inputting writing data of a plurality of characters and a writing data of a plurality of characters.
Segmentation processing that divides into writing data for each character, feature point extraction processing that extracts characteristic points from the writing data for each character, feature point display processing that connects and displays feature points with straight lines, and radicals of characters A character structure dictionary that stores the structure, a feature point editing process that corrects segmentation errors in the segmentation process using the displayed feature points, and for each character,
A recognition dictionary that stores a pair of standard patterns and character codes, a dictionary matching process that matches feature points with the recognition dictionary and determines whether the shapes are similar, and a display unit that displays the recognition result. Set up.

When the writing data of a plurality of characters is input by the input means, a segmentation process is performed to divide the writing data into each character. Next, the feature point extraction process is performed from the writing data for one character, and the feature point is displayed by the feature point display process. Here, a character interval is displayed for each delimiter position that is the result of the segmentation processing. The feature point editing process corrects the segmentation error in the segmentation process using the displayed feature points. After correction, the dictionary matching process is performed and the recognition result is displayed.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a basic block diagram of a character input device according to an embodiment of the present invention. 101
Is an input means such as a tablet for inputting the writing data of the operator. Reference numeral 102 denotes a display unit that displays the recognition result. A storage unit 103 stores a recognition result and various programs. Reference numeral 104 denotes a control unit that controls each of the above units. Reference numeral 105 is a power supply for supplying power to each of the above means.

FIG. 2 is a diagram schematically showing a program and data stored in the storage means 103.

Reference numeral 201 is a segmentation processing program that divides the writing data for a plurality of characters into individual characters.

Reference numeral 202 is a feature point extraction processing program. The characteristic point extraction processing is processing for extracting characteristic points from the writing data input by the operator. FIG. 3 is a diagram showing an example of the feature point extraction processing. When the handwriting data as shown in FIG. 3A is input, a specific number of coordinate groups are spread and output for each stroke. FIG. 3B shows an example in which the characteristic coordinates of three points are determined per stroke and the three points are connected by a straight line and displayed.

Reference numeral 203 is a feature point edit processing program. This program is a process in which the operator edits the displayed feature points and corrects the delimiter position in the segmentation process.

Reference numeral 204 is a dictionary collation processing program, which collates the characteristic points as the result of the characteristic point extraction processing with the standard pattern of the recognition dictionary 205 and judges whether or not the shapes are similar. . Compare the feature points with the recognition dictionary,
The character code of the most similar character is output as the recognition result.

A schematic diagram of the recognition dictionary is shown in FIG. A standard pattern 401 and a character code 402 are stored as a set for each character. The standard pattern 401 is the result of extracting characteristic points from typical writing data.

The character structure dictionary 206 is dictionary data in which characters are divided and stored for each radical and the characteristic points of each radical are also retained. FIG. 9 is a diagram schematically showing a character structure dictionary. 901 is a character code. 902 decomposes the character of the character code 901 into each part for each radical,
Each characteristic point is stored. For example, the character code 95A7 (France) indicates that the character code is composed of two parts.

Next, the operating procedure will be described. Figure 5
It is an example of a screen of the conventional handwritten character recognition processing. The figure shows an example in which a plurality of characters are successively input.

FIG. 5A is a diagram showing an example in which the operator inputs the writing data with the intention of inputting the character string "time machine". Next, a segmentation process is performed to determine a character break ((b) in FIG. 5). However, the character breaks are internal data and are not displayed to the operator. Here, a case is shown in which the delimiter between "a" and "mu" fails and the character is delimited as one character. Subsequently, the handwritten character recognition processing is performed on the handwritten character data for five characters delimited as shown in FIG. 5B, and the result is displayed on the screen (FIG. 5C).

As described above, in the conventional handwritten character recognition processing, the operator has no opportunity to correct the segmentation delimiter error during the state transition from FIG. 5 (a) to FIG. 5 (c). . Therefore, the operator was forced to input the writing data of the partial character string "im" again.

FIG. 6 is an example of a screen of the handwritten character recognition processing according to the present invention. 6A and 6B are similar to FIG.
Is the same as However, according to the present invention, even if the segmentation process causes a similar segmentation error, the segmentation process segmentation error can be corrected. In the present invention, the segmented writing data is subjected to the feature point extraction processing by the segmentation processing and displayed on the screen ((c) of FIG. 6).

Here, as shown in FIG. 6 (c), characters are displayed with an interval between each delimiter position which is the result of the segmentation processing. Thereby, the operator can identify the result of the segmentation process by the character spacing.
If the delimiter position of the segmentation process is different from the operator's intention as shown in FIG. 6C, a part of the coordinates extracted from the feature points of the operator is displayed as shown in FIG. 6D. Move to suit your intent. In the case of (d) of FIG. 6,
By separating the feature point corresponding to "mu" from "a",
The operator can modify the result of the segmentation process.
Handwritten character recognition processing is performed based on the corrected division position, and the result is displayed on the screen as shown in FIG.

As described above, in the handwritten character recognition processing of the present invention, the operator can correct the result of the segmentation processing, and even if the result of the segmentation processing is incorrect, the re-input of the writing data can be prevented. .

Next, a processing procedure for realizing the above screen example will be described. FIG. 7 is a diagram showing a processing procedure of conventional handwritten character recognition processing. When writing data of a plurality of characters is input (step 701), segmentation processing is performed as shown in FIG.
2) Next, the feature point extraction process (step 703) and the dictionary matching process (step 704) are successively performed, and the recognition result is displayed as shown in FIG. 5C (step 70).
5).

FIG. 8 is a diagram showing a processing procedure of the handwritten character recognition processing of the present invention.

The difference from the prior art is that after the characteristic point extraction processing (step 803), the characteristic point display processing for connecting and displaying the characteristic points by straight lines is performed as shown in FIG. 6C (step 8).
04), as shown in FIG. 6D, the feature point editing process (step 805) for correcting the segmentation error of the segmentation process is performed using the displayed feature points.

Here, the movement of the feature points by the feature point editing processing may be performed only in the lateral direction as shown in FIG. 6 (d). This means that when the writing data is input horizontally, it is possible to prohibit the writing data of a plurality of characters from being arranged in the vertical direction by dividing the column with ruled lines. This is because the segmentation processing can be limited to the horizontal direction.

Further, if the movement of the above-mentioned characteristic points is carried out in units of radicals, the burden on the operator can be reduced. For this purpose, it is necessary to acquire the radical structure of the feature points, and for this purpose, a character structure dictionary is used. The feature points after the segmentation processing are collated once with the dictionary, the character code is acquired, and the radical structure is acquired from the character structure dictionary. These processes are internal processes that are not displayed to the operator. As a result, it becomes possible for the operator to move each radical structure in order to correct an error in the segmentation process.

Note that the number of extracted coordinates of feature points may differ depending on the number of strokes of one character. In this case, after the feature points are extracted with a certain number of feature points, if the operator corrects the delimiter position, the number of character feature points at the new delimiter position may not be correct. In this case, when the number of extracted feature points differs depending on the number of strokes of the character, the feature points are set to the maximum (or least common divisor) feature points in each stroke until the operator finishes correcting the division position. The above problem can be solved by recreating the feature points for each character when the operator finishes the correction of the delimiter position by extracting and storing it.

[0030]

As described above, in the handwritten character recognition processing of the present invention, the operator can correct the result of the segmentation processing, and even if the result of the segmentation processing is incorrect, the re-input of the writing data is prevented. be able to.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of a character input device according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram schematically showing a program and data stored in a storage unit of FIG.

FIG. 3 is an explanatory diagram showing an example of feature point extraction processing.

FIG. 4 is an explanatory diagram schematically showing an example of data stored in the recognition dictionary of FIG.

FIG. 5 is an explanatory diagram showing a procedure of handwritten character recognition processing according to a conventional technique.

FIG. 6 is an explanatory diagram showing a procedure of handwritten character recognition processing according to the present invention.

FIG. 7 is a flowchart showing a processing flow of handwritten character recognition processing according to a conventional technique.

FIG. 8 is a flowchart showing a processing flow of handwritten character recognition processing according to the present invention.

FIG. 9 is an explanatory diagram schematically showing an example of data stored in the character structure dictionary of FIG.

[Explanation of symbols]

101 Input means such as a tablet for inputting writing data of an operator 102 Display means for displaying recognition results 103 Storage means for storing recognition results and various programs 104 Control means for controlling each of the above means 105 Supplying power to each of the above means Power supply

Front page continued (72) Inventor Hiromi Tomita, 292 Yoshida-cho, Totsuka-ku, Yokohama, Kanagawa, Ltd. Hitachi, Ltd. Multimedia System Development Division (72) Inventor Hiroyuki Kumai, 292, Yoshida-cho, Totsuka-ku, Yokohama, Kanagawa Ceremony Company Hitachi, Ltd.Multimedia Systems Development Headquarters (72) Inventor Akira Nakajima 292 Yoshida-cho, Totsuka-ku, Yokohama, Kanagawa Prefecture Stock Company Hitachi Systems Multimedia Systems Development Division (72) Inventor Yasumasa Matsuda Totsuka, Yokohama-shi, Kanagawa Prefecture 292 Yoshida-cho, Ward, Ltd. Hitachi, Ltd. Multimedia System Development Division

Claims (10)

[Claims] 1. An input unit for inputting a plurality of characters of writing data, a segmentation unit that divides the writing data input from the input unit into writing data for each character, and one character that is an output result of the segmentation unit. Feature point extracting means for extracting feature points from the handwriting data, feature point displaying means for connecting and displaying the feature points which are the output results of the feature point extracting means by straight lines, and the feature points displayed by the feature point displaying means. The feature is edited by the operator to move the points, the character recognition dictionary that stores the standard pattern and the character code as a set, the extracted feature points and the standard pattern of the character recognition dictionary are collated, and the pattern is compared. A dictionary collating means for calculating the distance between, and a display for displaying the character code paired with the standard pattern having the closest distance based on the processing result of the dictionary collating means. Comprising a stage, and by moving the feature point by using the feature point editing means, the character input device characterized by modifying the separation position of the writing data of the segmentation means. 2. The character input device according to claim 1, wherein the feature point display unit displays the writing data as an output result of the segmentation unit at intervals for each delimiter position. 3. The feature of the feature point editing means according to claim 1, wherein a straight line is displayed at an input position of the input means in a writing data input direction, and continuation of writing data other than the input direction is prohibited. A character input device characterized in that points are moved only in the input direction. 4. The method according to claim 1, wherein when the number of coordinates of the feature points, which is the output result of the feature point extracting means, is not constant depending on the number of strokes of the character, the operator's delimiter position by the feature point editing means is determined. A character input device, characterized in that the characteristic points are held at the coordinate number of the characteristic points of the maximum or least common multiple in each stroke until the correction is completed. 5. The character input device according to claim 4, wherein when the correction of the break position of the operator by the feature point editing means is completed, the feature point is recalculated for each corrected break position. 6. An input means for inputting a plurality of characters of writing data, a segmentation means for dividing the writing data input from the input means into writing data for each character, and one character which is an output result of the segmentation means. Feature point extracting means for extracting feature points from the handwriting data, feature point displaying means for displaying the feature points, which are the output results of the feature point extracting means, by connecting them with straight lines, and a character storing the radical structure of the character The structure dictionary, the feature point editing means by which the operator moves the feature points displayed by the feature point display means, the character recognition dictionary in which the standard pattern and the character code are stored as a set, the extracted feature points and the above A dictionary matching means for matching the standard pattern of the character recognition dictionary and calculating the distance between the patterns, and a standard pattern having the closest distance based on the processing result of the dictionary matching means. And a display means for displaying a character code paired with the above, wherein the characteristic point editing means is used to move the characteristic points to correct the delimiter position of the writing data of the segmentation means. Character input device. 7. The character input device according to claim 6, wherein the feature point display means displays an interval at each delimiter position of the writing data which is the output result of the segmentation means. 8. The feature of the feature point editing means according to claim 6, wherein a straight line is displayed at an input position of the input means in a writing data input direction to prohibit continuation of writing data in a direction other than the input direction. A character input device characterized in that points are moved only in the input direction. 9. The method according to claim 6, wherein when the number of coordinates of the feature points, which is the output result of the feature point extraction means, is not constant depending on the number of strokes of the character, the operator delimiter position by the feature point editing means is set. A character input device, characterized in that the characteristic points are held at the coordinate number of the characteristic points of the maximum or least common multiple in each stroke until the correction is completed. 10. The character input device according to claim 9, wherein when the correction of the break position of the operator by the feature point editing means is completed, the feature point is recalculated for each corrected break position.