JPS616780A - Recognizing method of character - Google Patents

Abstract

PURPOSE:To reduce the frequency of rejects by regarding an unrecognizable character having one and the same similarity order as that of plural reference characters as one and the same character, and counting up its generating frequency, recognizing the character while referring the reference character when the counted value exceeds a prescribed value. CONSTITUTION:A non-recognized word storage area in a memory of a character recognizing device has storage areas 10-30 for unrecognizable characters and respective areas 10-30 are provided with similarity storage columns 12-32, preceding/succeeding character storage columns 13-33, generating frequency storage columns 14-34, and correct character storage columns 15-35. The character recognizing device regards an unrecognizable character having practically one and the same similarity order as that of plural reference characters which are references of recognition to count up the generating frequency of the unrecognizable character, and when the counted value exceeds a prescribed value, specifies the unrecognizable character as a reference character while referring the contents stored in the non-recognized word storage area.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a character recognition method, and more specifically, the present invention relates to a character recognition method. Regarding weaving methods.

[Prior art]

In recent years, character recognition devices have been put into practical use, but there are cases where similar characters, especially handwritten characters, cannot be recognized.In such cases, recognition is rejected to avoid misrecognition. It is often configured to do so.

In general, character recognition is performed by calculating the degree of similarity between each of multiple standard characters (characters that serve as the basis for recognition) that have a high degree of similarity to the character to be recognized or woven, using a predetermined calculation process, and then determining the standard character with the highest degree of similarity. The degree of similarity (maximum similarity) with the standard character is greater than or equal to a predetermined value, and the difference between this maximum degree of similarity and the degree of similarity (the next highest degree of similarity) with the standard character with the next highest degree of similarity is a predetermined value. It is recognized as a standard character with the maximum similarity only if it is above a threshold value, and if the maximum similarity is below a predetermined value or the difference between the maximum similarity and the corporate similarity is within the dark value. , rejection is performed to avoid misrecognition.

By the way, as a method to avoid the above-mentioned rejection, for example, Japanese Patent Publications No. 52-19416 and No. 57
-49940 etc. have been proposed.

In Japanese Patent Publication No. 52-19416, by making the darkness value for recognition determination variable according to the difference between the above-mentioned maximum similarity and corporate similarity, recognition becomes impossible without increasing the rate of false recognition. I am trying to reduce the proportion of characters. However, the recognition accuracy increases as the darkness value for determining between the maximum similarity and the next-largest similarity increases, so it is possible that the recognition accuracy decreases with the above method. Even if it is made smaller, if the difference between the maximum similarity and the corporate similarity is within the range of the dark value, it will still be rejected.

On the other hand, in Japanese Patent Publication No. 57-49940, a standard pattern for rejection is newly established in the center of the area where misrecognition is likely to occur, a certain range of rejection area is set around this pattern, and an input unknown pattern (recognition target If a character (character) enters this area, it will be rejected. As a result, it is said that it is possible to almost completely reduce the erroneous UA weave area as a reject area without almost reducing the conventionally correct recognition iIl+ area, but in this invention, the erroneous
Although this makes it possible to avoid more errors than before, the number of rejects increases more than before, which increases the number of re-inputs or requires other configurations to prevent rejects, which increases costs. Problems such as this arise.

Further, in any of the above-described inventions, when recognizing handwritten characters, there is a possibility that rejections may occur frequently due to individual habits.

[Purpose of the invention]

The present invention has been made in view of the above-mentioned problems, and each time a reject character, that is, an unrecognizable character to be recognized, occurs, the similarity ranking and degree of similarity are determined for a plurality of standard characters that have a high degree of similarity to that character. and the characters written before and after that character are memorized, and the frequency of occurrence of an unrecognizable character having substantially the same similarity ranking as this unrecognizable character is counted, and this count value is determined as a predetermined value. When the value is reached, it is identified as one of the standard characters by referring to the similarity ranking, the similarity, and the characters written before and after that character, and from then on it is recognized as the specified standard character. To provide a character recognition method that can reduce the occurrence of rejects, reduce the effort of re-inputting, eliminate the wasted time required for this, and improve the recognition rate.

[Structure of the invention]

The character recognition device according to the present invention provides a method for recognizing a recognition target character based on the degree of similarity between the recognition target character and a plurality of standard characters, when the recognition target character is not recognized as any standard character. , each time, for a plurality of standard characters with a high degree of similarity to this recognition target character, store the rank of W4 colleagues, the degree of similarity, a code specifying the standard character, and the characters written before and after the recognition target character, Every time an unrecognizable character to be recognized that has substantially the same rank as the above-mentioned rank occurs, the number of occurrences is counted, and when this count value is greater than or equal to a predetermined value, the stored similar rank, similar The unrecognizable recognition target character is identified as one of the standard characters by referring to the degree and the characters written before and after the recognition target character, and the unrecognizable recognition target character is then used as the specified standard character. Characterized by recognizing characters.

〔Example〕

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below based on drawings showing embodiments thereof.

The drawing is a schematic diagram showing the contents of an unrecognized word storage area provided in the memory of a character recognition device used to implement the present invention.

This unrecognized word storage area is, for example, a predetermined area in the memory divided into predetermined locations, each of which is a storage area 10, 2 for unrecognized characters (hereinafter referred to as unrecognized characters).
0.30...It is a closed item. Each storage area 10 etc. has a similarity storage field 12□22.32..., a preceding and following character storage field 13, 23.33..., an occurrence number storage field 14, 24.34..., and a correct character. Memory column 15.25.
35... are provided.

Similarity storage columns 12, 22, 32, . . . store the degree of similarity of each unrecognized character with a plurality of standard characters according to its rank, and also store the JIS code of the standard character.

The preceding and following character storage columns 13, 23, 33, . . . store the characters that were stored before and after each unrecognized character when it occurred. This can be used as a reference when identifying the unrecognized character by memorizing the Q-okuri kana, etc. if the unrecognized character is a kanji and composes a compound word, or if it is read in kun-yomi. It's for a reason.

In the occurrence count storage columns 14, 24, 34, etc., each time an unrecognized character assigned to each storage area 10, 20, 30, etc. re-occurs, the number of occurrences is stored. is the similarity storage column 1 of each storage area 10, 20, 30...
When an unrecognized character having a rank that is considered to be substantially the same as the similarity rank stored in 2, 22, 32, etc. occurs, it is counted as the occurrence of the same unrecognized character. Note that the similarity rankings are substantially the same not only when they are exactly the same, but also when, for example, most of the combinations of stored similarity rankings are the same and the top two or three similarity rankings are the same. Cases etc.

In the correct character storage columns 15, 25, 35, . . . , the correct answers to the unrecognized characters assigned to the respective storage areas 10, 20, 30, . . . are manually input and stored. That is, when the number of occurrences stored in the number of occurrences storage columns 14, 24, 34, . Each data stored in ... is output. In this output data, as mentioned above, the JIS codes of standard characters are stored for each unrecognized character in descending order of class A or higher, and each time the unrecognized character is stored, the preceding and following characters are stored. (constituent characters of idioms, kanji, etc.) are also stored, so the user refers to these to specify unrecognized characters and inputs them into the correct character storage columns 15, 25, 35, . . . .

As described above, unrecognized characters that occur at a certain frequency are identified, and the correct characters (standard characters) are stored in the correct character storage column 1 of each storage area 10 (or 20, 30, etc.).
5 (or 25.35...), it is thereafter stored in the similarity storage column 12 (or 22.32...) of each storage area 10 (or 20.30...). When an unrecognized character having substantially the same similarity rank as the previous similarity rank occurs, the storage area 10 (or 20, 30, etc.) is
) is recognized as the correct character stored in the correct character storage column 15 (or 25, 35...).

In addition, if there is no more room in the unrecognized word storage area,
The data of unrecognized characters that occur less frequently (and have been stored for a longer time) are deleted, and newly generated unrecognized characters are stored.

〔effect〕

As described in detail above, the character recognition method according to the present invention considers unrecognizable characters that are substantially the same in the rank of similar or higher than a plurality of standard characters that are recognition standards as the same character. The frequency of occurrence is counted, and when this count value exceeds a predetermined value, the unrecognizable character is ranked in similarity to the stored standard characters, the degree of similarity, the code of the standard character, and The character is specified as one of the standard characters by referring to the characters written before and after the character, and is thereafter recognized as the specified standard character. Therefore, the recognition rate is improved and the frequency of rejections is reduced, which saves the effort of re-inputting and reduces the time required for character recognition.

Furthermore, according to the present invention, it is possible to avoid rejections due to individual quirks, so it is highly effective in recognizing handwritten characters, and it is also possible to know what kind of characters the quirky characters are judged to be similar to. Therefore, the user of the character recognition device should try to avoid inputting idiosyncratic characters.

[Brief explanation of drawings]

The drawing is a schematic diagram illustrating the contents of an unrecognized word storage area provided in a memory of a character recognition device, for illustrating an embodiment of the present invention.

Claims (1)

[Claims] 1. In a method for recognizing a character to be recognized based on the degree of similarity between a character to be recognized and a plurality of standard characters, if the character to be recognized is not recognized as any standard character, In each case, the similarity ranking of multiple standard characters with high similarity to this recognition target character,
The degree of similarity, the code specifying the standard character, and the characters written before and after the character to be recognized are memorized, and each time an unrecognizable character to be recognized that has substantially the same ranking as the above-mentioned ranking occurs, that character is memorized. The number of times the recognition target character is recognized is counted, and when this count value is greater than or equal to a predetermined value, the unrecognizable recognition target is identified by referring to the stored similarity rank, degree of similarity, and characters written before and after the recognition target character. A character recognition method, characterized in that a character is specified as any standard character, and then the unrecognizable recognition target character is recognized as the specified standard character.