JP2939945B2 - Roman character address 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 Roman character address recognizing device, and more particularly to a Roman character address recognizing device for performing character recognition on address information handwritten in Roman characters.

[0002]

2. Description of the Related Art Conventionally, in address reading of a character recognition device, an object is to prevent erroneous recognition by using a word dictionary or a character table as disclosed in, for example, Japanese Patent Application Laid-Open No. 62-247451. The main character recognition method was as follows. In the case of an English sentence, a plurality of words having a correct spelling are stored, and a spell check for extracting a description other than the word having the correct spelling is performed.

[0003]

In the above-mentioned conventional example, there is a problem that the recognition performance is not improved if the target is kanji or kana and the target is romaji. In particular, in the case of Roman characters, "H" and "N" and "U" and "J" have similar shapes, so that the recognition rate is not improved.

On the other hand, in the case of an address, unlike a general word, the number of place names is large, and it is often changed, so that it is difficult to store a correct word in advance, and furthermore, However, the automatic recognition of the address could not be performed if the candidate was generated, and therefore, the recognition rate of the address written in the Roman alphabet could not be improved.

[0005]

SUMMARY OF THE INVENTION An object of the present invention is to provide a Roman character address recognition apparatus which can solve the disadvantages of the prior art and can improve the recognition rate.

[0006]

SUMMARY OF THE INVENTION Accordingly, the present invention provides a reading section for reading a character string of an address written in Roman characters, and a recognition process for recognizing a character string of image data of the character string read by the reading section. Means. Moreover, the recognition processing means converts a character string information of the recognition result with reference to a character conversion table which records a combination of a result of the erroneous recognition and a correct character with respect to a character having a high probability of erroneous recognition. And a character replacement processing unit. This aims to achieve the above-mentioned object.

[0007] The reading unit reads a character string of an address written in Roman characters. Then, the recognition processing means performs a recognition process on the image data of the character string, and recognizes a Roman character string. At this time, the character replacement processing unit
The character string of the recognition result is converted with reference to the character conversion table. In the character conversion table, a combination of the result of incorrect recognition and a correct character is recorded for a character having a high probability of misrecognition. Convert according to the table. Then, even if erroneous recognition of a Roman character that is difficult to distinguish by the described method is made, this is corrected. At this time,
If the conversion is performed with a combination of three characters or the like, the conversion will not be performed erroneously.

[0008]

Next, an embodiment of the present invention will be described with reference to the drawings. As shown in FIG. 1, the Roman character address recognition device according to the present embodiment includes a reading unit 20 that reads a character string of an address written in Roman characters, and a character string of image data of the character string read by the reading unit 20. A recognition processing unit 10 having a character recognition processing unit 12 for recognition. This recognition processing means 10
A character conversion table that records a combination of a result of erroneous recognition and a correct character for a character having a high probability of erroneous recognition, and a character replacement processing unit 14 that converts character string information of a recognition result with reference to the character conversion table. Prepare. In one embodiment, the character conversion table has a combination of the result of misrecognition and a correct character in units of three Roman characters. Then, erroneous replacement is not performed.

The recognition processing means 10 includes a Roman alphabet rule replacement table that records characters that are incompletely written according to the rules of Japanese Roman alphabet notation and replacement candidates for the incomplete characters, To convert the character string information of the recognition result with reference to the Roman character rule replacement processing unit 16. As a result, a part of the recognition result of the character recognition processing unit 12 that does not conform to the rules for writing in Roman characters is replaced with a predetermined candidate. The Roman character rule replacement table has, for example, character string replacement candidates in which the ending of the character string is a consonant. Further, in one example, the character string has a replacement candidate for a character string in which consonants are continuous.

Further, it is preferable to have a table input section for inputting these tables and a table registration section for registering information input to the table input section in each table. According to this, erroneous recognition can be reduced according to the characteristics of the character recognition processing unit 12.

Next, the operation will be described.

As shown in FIG. 2, first, input image data 2A of a handwritten Roman alphabet address character string is input to the reading unit 20. Then, for example, the reading unit may output a candidate character indicated by reference numeral 2B as a candidate character string. Recognition performance of alphabetic characters is more difficult than kanji or kana, and handwritten "H and N" and "U and J" are often indistinguishable depending on the description method.

Therefore, the candidate character string output from the OCR reading unit is not directly output as a recognition result, but a character replacement process indicated by reference numeral 2C is performed to replace and correct the candidate character of the reading unit. In the character replacement processing unit 14, characters having a high probability of being erroneously recognized by the character recognition processing unit 12 are registered in advance in the character replacement table shown in FIG. Replace the output candidate string.

In the character replacement processing unit 14, reference numeral 2B
Are input to the character replacement processing unit, and a character string is searched by referring to the character conversion table. Further, a character string is searched with reference to the character conversion table to obtain character strings “CNI” and “SNI” corresponding to the character replacement table. The character replacement processing unit 14 performs character replacement from “CNI → CHI” to “SNI → SHI” with reference to the character replacement final result. The character string replacement unit outputs the content indicated by reference numeral 2D as the processing result.

Further, the character string output by the character replacement processing unit 14 is processed by the Roman character rule replacement unit 16. The Roman character rule replacement unit 16 replaces the character string indicated by the code 2D output by the character replacement processing unit 14 with the character string corrected by the character replacement processing unit according to the rules of the Roman character rule conversion table shown in FIG. For example, if the input of the Roman character rule replacement unit is the content indicated by reference numeral 2D, the search is performed with reference to the table of FIG. 4 to obtain “NIHOM” corresponding to the table of FIG.

[0016] In the item of the character replacement condition "when the ending of the word is other than AUEON" in the Roman alphabet rule replacement table,
“M” of “NIHOM” is a character replacement target in the Roman rule replacement unit, and is converted to “M → N” here. As described above, the character string converted and output using the character replacement processing unit 14 and the Roman character rule replacement processing unit 16 is output as the recognition result of the OCR device. As described above, when double correction processing is performed by the character replacement processing unit 14 and the Roman character rule replacement processing unit,
Since the processing of the consonant and the processing of the vowel are performed, the recognition accuracy is further improved.

As described above, according to the present embodiment, the OC
By adding character replacement processing and Roman character rule replacement processing to the character recognition result output from the R reader,
The reading accuracy of Roman character address recognition can be improved.

[0018]

According to the present invention, the character replacement processing unit converts the character string of the recognition result with reference to the character conversion table, and converts the character string into the character conversion table. Is recorded as a combination of a character with a high probability of misrecognition and a correct character. Then, even if erroneous recognition of a Roman character that is difficult to distinguish by the described method can be corrected, this can improve the recognition rate of handwritten Roman characters as compared with the conventional example. There can be provided no excellent Roman character address recognition device. Also, the Roman character rule replacement processing unit converts the character string information of the recognition result with reference to the Roman character rule replacement table. And the replacement candidate for the incomplete character are recorded, so that if the Roman character string of the recognition result is incomplete as a Roman notation in Japanese, it can be replaced with a replacement candidate. Accordingly, the recognition rate of handwritten Roman characters can be improved as compared with the conventional example.

[Brief description of the drawings]

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

FIG. 2 is an explanatory diagram showing an operation example in the configuration shown in FIG. 1;

FIG. 3 is an explanatory diagram showing an example of a character conversion table in the processing shown in FIG. 2;

FIG. 4 is an explanatory diagram showing an example of a Roman character table in the processing shown in FIG. 2;

[Explanation of symbols]

 Reference Signs List 10 recognition processing means 12 character recognition processing unit 14 character replacement processing unit 16 Roman character rule replacement processing unit

Claims (7)

(57) [Claims] 1. A Roman character address recognition apparatus comprising: a reading unit that reads a character string of an address written in Roman characters; and a recognition processing unit that recognizes a character string of image data of the character string read by the reading unit. The recognition processing unit converts a character string information of a recognition result by referring to a character conversion table in which a combination of a result of erroneous recognition and a correct character is recorded for a character having a high probability of erroneous recognition; A character address recognition device comprising: a character replacement processing unit. 2. The character conversion table according to claim 1, wherein the character conversion table is a Roman character 3
2. The Roman character address recognition device according to claim 1, wherein a combination of a result of the erroneous recognition and a correct character is provided in units of characters. 3. A table input unit to which a combination of a result of the erroneous recognition and a correct character is input, and a combination of the result of the erroneous recognition input to the table input unit and the correct character. 3. The Roman character address recognition device according to claim 1, further comprising: a table registration unit that registers the character conversion table. 4. A Roman character address recognition apparatus comprising: a reading unit that reads a character string of an address written in Roman characters; and a recognition processing unit that recognizes a character string of image data of the character string read by the reading unit. The recognition processing means refers to a Roman alphabet rule replacement table that records characters that are incompletely written according to the rules of Japanese Roman alphabet notation and replacement candidates for the incomplete characters, and refers to the Roman alphabet rule replacement table. A Roman character address recognition device, comprising: a Roman character rule replacement processing unit that converts character string information of a recognition result. 5. The Roman character address recognition device according to claim 4, wherein the Roman character rule replacement table has character string replacement candidates in which the ending of the character string is a consonant. 6. The Roman character address recognition device according to claim 4, wherein the Roman character rule replacement table has a replacement candidate for a character string in which consonants are consecutive in the character string. 7. A table input unit, wherein the recognition processing unit inputs a character that is incompletely written according to the rules of the Japanese Roman alphabet notation and a replacement candidate for the incomplete character, A table registration unit for registering, in the character conversion table, a character that is incompletely written according to the rules of Japanese Roman alphabet input and a replacement candidate for the incomplete character. Item 7. The Roman character address recognition device according to item 4, 5 or 6.