JPH0594565A - Hand-written character recognizing system - Google Patents

Abstract

PURPOSE:To execute highly accurate hand-written character recognition without previously registering user's hand-written characters. CONSTITUTION:A reference patter is stored in a pattern database 6 and hand- written character characteristic information of each user is stored in a characteristic database 3. A character recognizing part 5 compares similarity between the hand-written input pattern and the reference pattern with similarity between a corrected input pattern corrected by a pattern correcting part 4 in accordance with hand-written character characteristic information and the reference pattern and adops the more similar pattern. On the other hand, the hand-written input pattern recognized as characters is stored in the pattern data base 6 as a learning character pattern in order to utilize it for a user's individual reference pattern and the characteristic information of the database 3 is updated.

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 system, and more particularly to a handwritten character recognition system for inputting a handwritten character pattern and outputting a character code of a corresponding character.

[0002]

2. Description of the Related Art Conventionally, this type of handwritten character recognition method is designed to handwrite a standard character shape, compare the input pattern with the standard pattern, and perform character encoding. In order to further improve the recognition accuracy, the user registers a pattern unique to the user in advance as a standard pattern and compares the standard pattern with the handwriting input pattern.

[0003]

In the conventional handwritten character recognition method described above, it is necessary to register the character pattern of the user in advance in order to improve the recognition accuracy. In addition to the limited number of users, there is the complexity of requiring advance registration work each time the user changes.

[0004]

A handwritten character recognition method of the present invention is a handwritten character recognition method of inputting a handwritten character pattern and outputting a character code of a corresponding character, and a pattern database storing standard patterns. A characteristic database storing characteristic information of handwritten characters, pattern correction means for correcting an input handwritten character pattern according to the specific information stored in the specific database to obtain a corrected character pattern, The first similarity is calculated by comparing the handwritten character pattern with the standard pattern stored in the database, and the corrected character pattern obtained from the pattern correcting means and the standard pattern stored in the pattern database are compared. The second similarity is calculated by comparison, and the character pattern is selected from the first and second similarities according to the standard pattern with the higher similarity. And a characteristic learning means for updating the characteristic information of the handwritten character stored in the characteristic database according to the result of the character recognition by the character recognizing means. It is constituted by having.

[0005]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention. The embodiment shown in FIG. 1 comprises a user code identifying section 1, a characteristic learning section 2, a characteristic database 3, a pattern correcting section 4, a character recognizing section 5, a pattern database 6, and a pattern learning section 7.

In the above structure, the handwriting input pattern is compared with the standard pattern stored in the pattern database 6 to perform character recognition. Further, the handwritten character characteristic information of the user is stored in advance in the characteristic database 3, and the pattern correction unit 4 corrects the handwriting input pattern according to the characteristic information. Character correction is performed by comparing the corrected input pattern obtained by correction with the standard pattern stored in the pattern database 6. Next, a good one is adopted from the two recognition results. On the other hand, the character-recognized handwriting input pattern is stored as a learning character pattern by the pattern learning unit 7 and used by the characteristic learning unit 2 to be used as a standard pattern unique to the user. Used to optimize property information.

FIG. 2 is a flow chart of the character recognition operation of FIG. 1. The operation of handwriting input pattern recognition in FIG. 1 will be described below with reference to FIG. A user identification code that can identify the user who wrote the handwritten character is input, and the user code recognition unit 1 confirms whether or not it has already been registered. If it is not registered, an area for storing the handwritten character characteristics of the user is secured in the characteristic database 3 (step 1
01). Next, a handwriting input pattern is input (step 102). The input handwriting input pattern is compared with the standard pattern stored in the pattern database 6 in the character recognition unit 5, and the similarity between the standard pattern and the handwriting input pattern, that is, the input pattern similarity is calculated ( Step 103). On the other hand, in the characteristic database 3,
Information about user's handwritten characters, for example, handwritten character inclination information is stored. The pattern correction unit 4 confirms whether or not the handwritten character characteristics of the corresponding user are stored in the characteristic database 3 (step 104). When the handwritten character characteristics are stored in the characteristic database 3, the pattern correction unit 4 corrects the handwritten input pattern according to the handwritten character characteristics (step 105).
Corrected handwriting input pattern and pattern database 6
The standard pattern of No. 1 is compared with the character recognition unit 5 to calculate the correction pattern similarity (step 106). If the handwritten character characteristic of the corresponding user code is not stored in the pattern database 6 in step 104, the correction pattern similarity is set to zero in the character recognition unit 5 (step 107). The two similarities obtained in the above process, that is, the input pattern similarity and the corrected pattern similarity are compared with each other in the character recognition unit 5 (step 108), and as a result, the handwriting input pattern is determined to have a higher similarity. It is recognized as a character of the standard pattern and the corresponding character code is output (step 109). When the character code is output, the presence / absence of the next handwriting input pattern is confirmed (step 110).

FIG. 3 is a flowchart of the learning operation of FIG. 1. The learning operation of FIG. 1 will be described below with reference to FIG.
In the recognition process, the conversion result indicating which character code the handwriting input pattern has been converted is input (step 2).
01) and the conversion result is notified to the pattern learning unit 7. The handwriting input pattern at that time is stored as a learned character pattern in the pattern database 6 by the pattern learning unit 7 so as to be used as a standard pattern in the subsequent recognition processing (step 202). The conversion result is also notified to the characteristic learning unit 2, and the inclination of the handwriting input pattern at that time is calculated (step 203). The handwritten character characteristics of the corresponding user already registered in the characteristic database 3 and the average value of the results in step 203 are calculated (step 204), and the handwritten character characteristics of the corresponding user are re-registered in the characteristic database 3. (Step 205). This allows
The handwriting character characteristics are optimized to correct subsequent handwriting input patterns.

[0010]

As described above, according to the present invention, in addition to the comparison between the handwriting input pattern and the standard pattern, the handwriting character characteristic information of the user is stored in advance, and the handwriting input pattern is stored in this characteristic information. According to the correction, the comparison between the corrected input pattern and the standard pattern is examined, and the input pattern is recognized as the character of the higher standard pattern,
Characters unique to the user can be recognized without registering in advance, and each time the user uses the automatic learning function, the character characteristic information of the user can be updated.
This has the effect of increasing the accuracy of successive recognition.

[Brief description of drawings]

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

FIG. 2 is a flowchart of the character recognition operation of FIG.

FIG. 3 is a flowchart of the learning operation of FIG.

[Explanation of symbols]

 1 User code identification unit 2 Characteristic learning unit 3 Characteristic database 4 Pattern correction unit 5 Character recognition unit 6 Pattern database 7 Pattern learning unit

Claims (2)

[Claims] 1. A handwritten character recognition method for inputting a handwritten character pattern and outputting a character code of a corresponding character,
A pattern database that stores standard patterns, a characteristic database that stores characteristic information of handwritten characters, and a pattern that corrects an input handwritten character pattern according to the specific information stored in this specific database to obtain a corrected character pattern. The correction means compares the input handwritten character pattern with the standard pattern stored in the database to calculate the first similarity, and the corrected character pattern obtained from the pattern correction means and the pattern database are stored in the pattern database. And a character recognition means for calculating a second similarity by comparing with the stored standard pattern and obtaining a character code from the first and second similarities to the higher standard pattern. Characterized handwritten character recognition method. 2. The handwritten character recognition method according to claim 1, further comprising a characteristic learning unit that updates the characteristic information of the handwritten character stored in the characteristic database according to the result of character recognition by the character recognition unit. ..