JP2994932B2 - Handwritten character recognition device - Google Patents

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 for recognizing handwritten input characters, such as an electronic organizer, a word processor, a personal computer, and the like.

[0002]

2. Description of the Related Art In a conventional handwritten character recognition apparatus, Japanese Patent Laid-Open No. 611-1 / 1986 discloses a method for reducing erroneous recognition of a recognition result.
There is one described in JP-A-14387. This is to reduce erroneous recognition due to habit peculiar to the writer by learning. When a handwritten character is erroneously recognized and one of the candidate characters is selected as a correct character, an average pattern is obtained by associating the basic pattern in the dictionary of the correct character with the input character pattern. Next, this averaged pattern is replaced with a basic pattern in the dictionary, and the new pattern is used as a basic pattern for the next recognition. Thus, by selecting a character in the candidate, the basic pattern of the dictionary is rewritten, so that the individual custom character can be easily recognized.

[0003]

[0007] The shape of the handwritten input character is individual difference is large, for example, "1" and "1" and "/" even if I'm the person who writes in shape, such as, "/" and "1" Some write like "/". For this reason, if the character is written in the correct shape, the character can be correctly recognized. However, if the character is deformed and becomes close to “/” due to the habit of each individual, the character is recognized as “/” instead of “1”. Since the value of the evaluation value is viewed as a difference from the basic shape of the character, it cannot cope with the habit of each individual, and the recognition rate is deteriorated.

[0004]

SUMMARY OF THE INVENTION The present invention has been made for the purpose of solving the above-mentioned problems. According to the present invention, there is provided a handwriting input means for inputting handwritten characters,
Basic pattern storage in which basic patterns of characters are stored
Means and a sentence stored in the basic pattern storage means
Basic pattern of characters and input by the handwriting input means
Calculate the distance to the handwritten character shape and evaluate it as a recognition result.
Evaluation value calculating means as a value, and the evaluation value calculating means
Regarding the calculated evaluation value, the one with the smallest value
Handwriting having candidate character output means for arranging candidate characters in order
In the character recognition device, the candidate character output means
The first candidate of the handwritten character obtained is not a correct character,
When the character is corrected to a correct character,
Value of the first candidate of the handwritten character obtained by
Correction value calculating means for calculating a correction value from an evaluation value of a character
And the correction value calculated by the correction value calculation means.
A correction information record for storing the first candidate character and a correct character.
Storage means and newly handwritten characters
When input, the evaluation value is calculated by the evaluation value calculation means.
Output to the candidate character output means based on the evaluation value.
Therefore, the first candidate of the handwritten character is obtained, and the obtained
And the first candidate stored in the correction information storage means.
One candidate character is compared and determined.
Correct characters using the correction value stored in the
Is corrected by the candidate character output means.
By rearranging candidate characters in ascending order of value, handwritten characters
Control means for correcting the recognition result of
This is a handwritten character recognition device .

[0005]

In a handwritten character recognition apparatus for determining the order of recognition result candidates based on an evaluation value obtained by comparing a handwritten input character with a prestored basic pattern, the first candidate character of the recognition result is a desired character. If it is not the character, the character is corrected to the desired character by indicating the desired character.
The evaluation value of the first candidate character and the evaluation value of the desired character
Correction value, and the first candidate character along with the correction value
And the desired character, and newly input by handwriting
When handwritten characters are entered,
Recognition is performed and candidate characters are output in the same manner as above.
When the first candidate is the same as the character of the first candidate stored in advance.
In this case, the evaluation value of the desired character is corrected by the correction value,
The recognition result candidates are output again in ascending order of the evaluation value .

[0006]

FIG. 1 is a system block diagram of a handwritten character recognition device of the present invention, and FIG. 2 is a flowchart of character recognition of the present invention. FIG. 3 is a diagram illustrating an example of a basic pattern stored in a ROM, and FIG. 4 is a diagram illustrating an example of character recognition. FIG. 5 is a flowchart of the learning of the present invention, and FIG. 6 is a flowchart of the correction of the recognition result of the present invention. 7 and 8 are diagrams illustrating an example of learning and correction according to the present invention.

In FIG. 1, a handwritten character recognition apparatus according to the present invention includes a panel-type display element 1, a pressure-sensitive tablet 2 as an input unit mounted on the surface of the display element, an A / D converter 3, and a display controller driver 4. And ROM (read only memory) 5 and RAM (random access memory,
Read / write memory 6) and CPU as control unit
(Central processing unit) 7.

The display element 1 is, for example, a liquid crystal display panel, and can display characters according to a received signal. The pressure-sensitive tablet 2 converts pressure into an electric signal (analog signal) for each stroke of a character handwritten by the user with the pen 8 and outputs the signal. The A / D converter 3 converts an analog signal from the pressure-sensitive tablet 2 into a digital signal and outputs the digital signal to the CPU 3. The display controller driver 4 controls the display element 1 based on a control signal from the CPU 3.
Drive. The ROM 5 includes a basic pattern of characters and a control program storage unit. The RAM 6 can temporarily store various data required for the handwriting recognition device. The CPU 3 performs various processes of handwritten character recognition including calculation.
And controls the entire operation of the handwritten character recognition device.

The recognition of handwritten characters will be described with reference to FIG. First, the distance between the character input by handwriting and the basic pattern of the character stored in the ROM 5 is determined (S
1). Next, the basic pattern having the shortest distance from the handwritten input character is set as the first candidate of the recognition result of the handwritten input character, and the recognition results are set in the order of smaller distances (S2). The recognition result candidates obtained in this way are displayed in response to key operations in the order of the candidate intensities,
The list is displayed, and the operator is made to select it, and the recognition result of the handwritten character is determined.

Next, referring to FIGS. 3 and 4, a description will be given of a method of obtaining the distance between a handwritten input character and a basic pattern, using "1" (number) and "/" (slash symbol) as an example. . Input from the pressure-sensitive tablet 2 is assumed to be represented by a 5 × 5 dots. FIG. 3 shows basic patterns of "1" and "/" stored in the ROM 5, and the distance between the basic pattern and the input pattern is calculated.

FIG. 4 shows a method of calculating a distance between an input pattern and a basic pattern. ○ mark indicates the input pattern, X marks indicate the basic pattern. ,
Shows respectively If you enter the If you enter the basic pattern the, the basic pattern than a little out of the pattern of "1" (ambiguous pattern) ,,
Is ,, a comparison of the basic pattern of "1" indicates each of the comparison of the basic pattern of "/". The distance between the input pattern and the basic pattern is determined for each horizontal line, and the sum is defined as the distance between the patterns. Since the distance for each horizontal line is 0, the total is also 0. Then, the difference between the distances of the first and fifth horizontal lines is 2, the difference between the second and fourth horizontal lines is 1, and the difference between the distances of the third horizontal lines is 0, so that the total is 6. In the input pattern of “,” the distance of the basic pattern of “1” is 0 and the basic pattern of “/” is 6, so “1” is the first candidate and “/” is the second candidate. , And the input pattern has a distance of 4 from the basic pattern of “1” and a distance of 2 from the basic pattern of “/ ” , the first candidate is “/” and the second candidate is “2”.
The candidate becomes “1”. The present invention has been conceived in order to obtain a correct recognition result in the subsequent recognition when the recognition result thus obtained, particularly the first candidate is incorrect.
Hereinafter, description will be made with reference to FIGS.

FIG. 5 is a flowchart of the learning process according to the present invention.
Thus, the first candidate A of the recognition result is corrected to the correct character B.
(S11), the evaluation value of A (the flowchart of FIG. 2)
The evaluation value of B is subtracted from the distance obtained in the figure), and 1 is further obtained.
Subtract the value X(Value of variable X)age,The value of this variable X
It is stored in the RAM 6 together with the characters A and B (S12).

Using the value of the variable X obtained in this manner, correction is made in subsequent recognition to obtain a correct recognition result. This will be described with reference to the flowchart of FIG. 6 for correcting the recognition result of the present invention. As a result of the recognition, it is determined whether the first candidate is the same as the character A stored at the time of learning (S2).
1) If not A (NO), the correction of the recognition result ends. On the other hand, if the first candidate A (YES), by adding the value of the variable X to the evaluation value of the candidate character B (S22), rearrange the candidate characters in ascending order of evaluation value (S23) the recognition result Is completed.

FIG. 7 shows a specific example of learning and correction. FIG.
Of the <br/> if "1" is correct character for the input character. In the learning mode, the value of the variable X is expressed as X = (evaluation value of “/”) − (evaluation value of “1”) − 1 = 2-4-1 = −3 .
Then, the value of the variable X (X = −3) is stored together with the characters “1” and “/”. Next, when the same pattern is input as originally first candidate is a "/",
When the correction mode of the recognition result by adding the value of the variable X to the evaluation value of "1" the result of the calculation is 4 + (- 3) = 1 and Do Ri,
Evaluation value after correction is "1" is 1, since the "/" is 2, the first candidate is "1", the second candidate "/".

FIG. 8 shows another specific example of learning and correction. In this case, the input character is "/" from the input character in the specific example of FIG.
Therefore, even if the correction is performed, the characters of the first candidate are not replaced , “/” remains the first candidate , and “1” is the second candidate.
Will remain .

[0016]

According to the present invention, characters input by handwriting
Is recognized, and the first candidate character of the recognition result is a desired correct character.
If it's not the right letter,
When a new handwritten character is input,
To be able to recognize written characters correctly
It is. For example, although the correct character when you enter the same character shape as characters entered when instructed the correct character is recognized in the first candidate, sometimes close to the input and recognition shape erroneous than shape, the recognition result is erroneous The recognized character is the first candidate. In addition, if a character other than the erroneously recognized character is the first candidate, no correction is performed, so that the recognition result is not affected. For this reason, it can be recognized according to the writing style of each individual.

[Brief description of the drawings]

FIG. 1 is a system block diagram of a handwritten character recognition device of the present invention.

FIG. 2 is a flowchart of character recognition according to the present invention.

FIG. 3 is a diagram showing an example of a basic pattern stored in a ROM.

FIG. 4 is a diagram illustrating an example of character recognition.

FIG. 5 is a flowchart of learning according to the present invention.

FIG. 6 is a flowchart of the correction of a recognition result according to the present invention.

FIG. 7 is a diagram illustrating an example of learning and correction according to the present invention.

FIG. 8 is a diagram illustrating an example of learning and correction according to the present invention.

[Explanation of symbols]

 Reference Signs List 1 display element 2 pressure-sensitive tablet 3 A / D converter 4 display controller driver 5 ROM 6 RAM 7 CPU 8 pen

Claims (1)

(57) [Claims] A handwriting input for inputting a handwritten character.
Force means and basic pattern storage in which basic patterns of characters are stored
Means and basics of characters stored in the basic pattern storage means
Pattern and handwritten sentence input by the handwriting input means
The distance from the character shape is calculated and used as the evaluation value as the recognition result.
Evaluation value calculation means, and the evaluation value calculated by the evaluation value calculation means,
Candidate characters that arrange candidate characters in order from the one with the smallest value
And a handwritten character recognition device having an output unit.
If one candidate is not a correct character but is corrected to a correct character
The handwritten sentence obtained by the candidate character output means.
Corrected from the evaluation value of the first candidate character and the evaluation value of the correct character
It said correction value calculating means for calculating a value, together with the correction value calculated by said correction value calculating means
Correction information storage means for storing first candidate characters and correct characters
A new handwritten character is input by the handwriting input means.
And an evaluation value is calculated by the evaluation value calculating means.
By the candidate character output means based on the evaluation value of
A first candidate for a handwritten character is obtained, and the obtained first candidate is obtained.
And the first candidate character stored in the correction information storage means.
And, if they are the same, stored in the correction information storage means.
Correcting the evaluation value of the correct character using the correction value,
The character output means arranges candidate characters in ascending order of evaluation value.
Control to correct the recognition result of handwritten characters by reviewing
Handwritten character recognition apparatus characterized by including a means.