JPS6336389A - Character reader - Google Patents

Abstract

PURPOSE:To improve the rate of a correct reading and to speed-up processing speed by character-recognizing a KANA (Japanese syllabary) attached to Chinese character recorded on a slip word-recognizing based on the result of said character recognition, and then, character-recognizing a KANJI (Chinese character) limiting candidate characters for the KANJI recorded on the slip in accordance with the result of the work-recognition. CONSTITUTION:A carrier scanning part 100 sends each quantized pattern, which corresponds to the KANA 200 attached to the Chinese character and to the KANJI 230 recorded on the slip to a preprocessing part 110. A character recognition processing part 120 character- recognizes by a character unit after selecting either the dictionary (ANK dictionay) 121 of English letter, numericals and KANA or a KANJI dictionary 122 by the aid of a dictionary selection control signal d1, and outputs the recognition result to a word recognizing part 130. The KANJI dictionary 122 enables a selection of the short of the stored character by a selection control signal d2 for the sort of recognizing character. The word recognizing part 130 word-recognizes by a character unit after selecting an ANK word dictionary 131 and a KANJI word dictionary 132 by the signal d1. A control part 140 decides a final recognition result according to the result of the word recognition of the KANA attached to the Chinese character, and to the result of the word recognition of the KANJI, and outputs the result of the decision.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention (Field of Industrial Application) The present invention particularly relates to a character reading device for recognizing handwritten Chinese characters.

(Prior art) In recent years, character reading devices (OCR) for handwritten kanji have been developed to recognize kanji using the reading results of furigana characters (usually katakana characters) corresponding to handwritten kanji as a guide in order to improve recognition accuracy. A method of processing has been developed.

In this type of character reading device, furigana characters corresponding to kanji are written on the form in advance, and based on the reading results of these furigana characters, the answer is recognized from among multiple candidates obtained from the kanji recognition results. The results are configured to be selected.

However, in the above-mentioned type of character reading device, the accuracy of kanji recognition depends on the recognition result of furigana characters (katakana characters), and it is only effective if the recognition process of furigana characters is performed with high accuracy. Currently, it is impossible to completely recognize katakana characters, which are furigana characters, so it is difficult to read kanji with high accuracy.

Therefore, in the past, in order to recognize kanji with high accuracy,
A character recognition means that recognizes each character character by character for a form in which kanji and furigana characters corresponding to the kanji characters are recorded, and a character recognition means that recognizes the above kanji and furigana characters based on the recognition result of the medium character output from this recognition means. A word recognition means for recognizing words made up of each character is provided, and the word recognition means collates the recognition results of each word of kanji characters and furigana characters (q), and calculates mutually corresponding kanji and furigana characters. The final answer was a combination of letters.

However, in the above-mentioned character reading device,
When there are many candidate characters as a result of character recognition of furigana characters recorded on a form, the number of characters to be compared inevitably increases, which takes time to process, leading to a decrease in the correct reading rate.

(Problems to be Solved by the Invention) As described above, in the conventional character reading device, when there are many candidate characters as a result of character recognition of furigana characters recorded on a form, processing takes a long time. , which led to a decline in the rate of legitimacy.

The present invention has been made in view of the above points, and is a character reading device that makes it possible to improve the correct reading rate and speed up processing in a character reading device that recognizes kanji by referring to the recognition results of furigana characters. The purpose is to provide a reading device.

[Structure of the Invention] (Means for Solving the Problems) In other words, in the character reading device according to the present invention, each character is and a word recognition means for recognizing each word on a word-by-word basis based on the character recognition results output from the character recognition means. controlling the character recognition for the kanji in the character recognition processing means based on the word recognition result for the furigana character, and controlling the word recognition result for the kanji character outputted from the word recognition result for the furigana character and the word recognition result for the furigana character. When the recognition results for each word of furigana characters and kanji correspond to each other,
A feature of the present invention is that each recognition result is output as a final answer.

(Function) In the character reading device configured as described above,
First, character recognition is performed on furigana characters recorded on a form, and word recognition is performed based on the recognition results. Next, character recognition is performed for the kanji characters recorded on the form, and in this case, the character recognition for the kanji characters is controlled by the word recognition result for the furigana characters.

That is, the candidate characters for the kanji part are limited based on the word recognition results for the furigana characters, and the kanji character On is performed using the limited candidate characters.

Therefore, the number of characters to be compared between the word λΣ recognition results for furigana characters and the word recognition results for kanji characters is reduced, and as a result, the success rate can be improved and the processing speed can be increased.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram for explaining the configuration of a character reading device according to the present invention.

In FIG. 1, 100 is a conveyance scanning section, and this field conveyance scanning section 100 conveys and scans a form (shown in FIG. 2), and photoelectrically converts each of the kanji characters and furigana characters recorded on this form. Then, an ω child pattern (a character pattern consisting of a binary signal) is output to the preprocessing section 110. This preprocessing unit 110 performs preprocessing on the input dumpling pattern,
The preprocessed quantization pattern (character pattern) is output to the character recognition processing section 120 via the data line 117.

The character recognition processing unit 120 receives a dictionary selection control signal d, which will be described later.
1, the data line 12 has the ability to select an alphabetic, numeric, and katakana character dictionary (hereinafter referred to as ANK dictionary) 121 and a kanji dictionary 122 to perform medium-level character recognition.
3, the recognition result of each character is sent to the word recognition unit 13.
Output to 0.

The kanji dictionary @122 is a kanji dictionary that allows selection of stored character types (hereinafter, this selection of character types is referred to as a character set) using a recognition character type selection control signal d2, which will be described later.

Word recognition section? 30 has an R function that selects the ANK word dictionary 131 and the Kanji word dictionary 132 using the dictionary selection system Usin@d1 and performs word recognition character by character. The furigana character word recognition results and the kanji word recognition results are sent to the control unit 1.
Output to 40.

The ANK word dictionary 131 is a dictionary storing ANK word dictionaries and their attributes, and the Kanji word dictionary 13 is a dictionary that stores ANK word dictionaries and their attributes.
2 is a dictionary in which a kanji word dictionary is stored.

The control unit 140 has the ability to control the entire device, and sends the recognition character type selection system 1i (I signal d2 and the dictionary selection control signal @d1 to the signal line 141).
and the kanji dictionary 122, the character recognition processing section 120, and the word i\! via the signal line 142. It is output to the identification section 130. Further, the control unit 140 determines the final recognition result based on the furigana character moderation Σ recognition result and the kanji word recognition result from the word recognition unit 130, and outputs the determined result to the data line 143.

The operation of the character reading device having the above configuration will be explained below. First, a form as shown in FIG. 2 is conveyed and scanned by the conveyance scanning unit 100, and the kanji and furigana characters recorded on the form are respectively converted into maternalization patterns and sent to the preprocessing unit 110. .

Figure 2 shows an example of a form, and 200 and 210 are character areas for writing ANK characters (particularly katakana) and kanji, respectively, and both are printed in colors that cannot be detected by OCR. . Also, 220 and 230 are characters written by a scribe, and the furigana are AN.
It is assumed that ``Yamamoto'' and 4 characters in katakana are entered in the K field, and ``Sansui'' and 2 kanji characters are entered in the kanji field.

Therefore, the conveyance scanning unit 100 sends each quantization pattern corresponding to the one-line kana characters 220 and kanji characters 230 on the form to the preprocessing unit 110. The preprocessing unit 110 performs a FIL operation on the input quantization pattern, and sends the preprocessed quantization pattern to the character recognition processing unit 120 via the data line 111.

Here, the character recognition processing unit 120 performs character-by-character recognition processing based on the adoption pattern of furigana characters 220 and kanji 230 from the preprocessing unit 110.
In this case, according to the dictionary selection control signal d1 from the control unit 140, the ANK dictionary 121 is referred to and the one-curved kana character 22 is selected.
The character 0 is recognized and the character recognition result as shown in FIG. 3 is output to the word recognition section 130.

FIG. 3 is an example of the furigana character recognition results for the furigana characters recorded in the form shown in FIG.
? ” indicates a reject character, and 310 “Uya” and 311 “kaJ” are candidate characters for the above-mentioned reject character 300, respectively.

When the word recognition unit 130 receives the result of character recognition of furigana characters as shown in FIG. Character word recognition results and the same kana characters! ! The kanji word that is the attribute of the one-word recognition result is sent to the control unit 1.
Output to 40.

Here, FIG. 4 is an example of the recognition result of a kana character word for the furigana character recognition result of the above-mentioned FIG. 3,
410, 420, and 430 correspond to the first, second, and third place in the recognition results. Also, FIG. 5 refers to ANKIi word recognition! This is an example of an 11-word dictionary, and 510 is a katakana word dictionary string, and in this example, four types are shown. Further, 520 is a Kanji word corresponding to each katakana (or furigana), and is ranked as an attribute of the word dictionary string 510.

When the control unit 140 inputs the furigana character/word recognition result as shown in FIG. 4 and its attribute, the kanji word string 520, the control section 140 first performs a logical wheel operation on the kanji word string 520 to generate the result as shown in FIG. 6. A kanji character string 61 (consists of l).

Next, the control unit 140 controls the kanji character string 6 shown in FIG. 6 above.
10 is output as a recognition character type selection control signal d2 to the kanji dictionary 122 as a character set for the kanji 230 of the form shown in FIG. 1
30 outputs a dictionary selection control signal d1 so as to refer to the Kanji word recognition 132.

In this way, when the recognition character type selection control signal d2 and the dictionary selection control signal d1 are output from the control unit 140,
The character recognition processing unit 120 performs character &X recognition for the kanji 230 on the form shown in FIG.
Output to. In this case, since the character of the kanji dictionary 122 has been set by the recognition character type selection control signal d2, the character recognition processing section 120 can perform the character Hy4 by referring to the kanji dictionary 122 in which the character has been set.

That is, when performing kanji recognition for ``Fyama'' and ``Book J'' recorded in the form shown in FIG. Kanji recognition can be performed by referring only to candidate characters such as ",""gen,""sawa," and "yama."

In this way, when the kanji recognition by the character recognition processing unit 120 is completed, the kanji recognition result is input to the word recognition unit 130, and the kanji word dictionary @13 is inputted to the word recognition unit 130.
Kanji word recognition is performed by referring to 2.

The control unit 140 receives the kanji word recognition results from the word recognition unit 130 via the data line 134, and uses the input 11 kanji recognition results and the furigana character word ranked No. 1 shown in FIG. 4 above. Compare with recognition results. and,
When both the furigana characters and kanji of each word correspond to each other, it is output as the final content. That is, in this case, the control section 140 outputs the answer shown in FIG.

In this way, in the character reading device of the present invention, first, the furigana characters recorded on the form are the characters Ll 'I
A, and word recognition is performed based on the result of this character recognition. Next, character recognition is performed for the kanji characters recorded on the form, and in this case, the character recognition for the kanji characters is controlled by the word recognition result for the furigana characters. That is, the candidate characters for Kanji are limited from the word recognition for the furigana characters, and the Kanji character HF2a is performed using these limited candidate characters.

Therefore, for example, even if furigana characters recorded on a form are recognized and the number of candidate characters increases as a result of character recognition, the number of kanji candidate characters when recognizing kanji recorded on a form is limited. Therefore, character recognition for kanji does not take much time.

Furthermore, since Kanji character recognition is performed using the limited candidate characters in this manner, the subsequent Kanji word recognition does not take much time, and the number of candidate characters in the word recognition result is reduced. Therefore, when comparing the final furigana word recognition results with the kanji word recognition results, the number of characters to be matched is reduced, resulting in an improved continuation rate and faster processing speed. .

[Effects of the Invention] As described above, according to the present invention, furigana characters recorded on a form are recognized, and word recognition is performed based on the result of the character recognition. By limiting the candidate characters for the recorded Kanji characters and performing character recognition for the Kanji characters, it is possible to improve the success rate and increase the processing speed.

[Brief explanation of the drawing]

The drawings show a practical example of the present invention; Fig. 1 is a block diagram showing the circuit configuration, Fig. 2 is a diagram showing an example of a form, and Fig. 3 is character recognition for furigana characters on the same form. Figure 4 shows an example of word recognition results for the same character recognition results. Figure 5 shows an example of a single gR dictionary that is referenced when recognizing ANK words and Σ. Figure 6 shows kanji. FIG. 7 is a diagram showing an example of a character string, and FIG. 7 is a diagram showing an example of the final result for the form shown in FIG. 2 above. 120... Character recognition processing unit, 121... ANK dictionary, 122... Kanji dictionary, 131... ANK word dictionary, 132... Kanji word dictionary, 140... Control unit.

Claims (1)

[Scope of Claims] A form having a character area for recording a kanji and a character area for recording furigana characters corresponding to the kanji, and the above-mentioned kanji recorded on the form by conveying and scanning the form. and a photoelectric conversion unit that photoelectrically converts the furigana characters and outputs a character pattern, and recognizes each of the furigana characters and kanji character by character based on the character pattern output from the photoelectric conversion unit. a character recognition means; a word recognition means that recognizes each word on a word-by-word basis based on the character recognition results outputted from the character recognition means; The character recognition processing means controls the character recognition for the above-mentioned kanji, and compares the word recognition result for the above-mentioned furigana characters with the word recognition result for the above-mentioned kanji output from the word recognition means, and in the comparison result, the furigana characters and 1. A character reading device comprising: control means for outputting each recognition result as a final answer when the recognition results for each character of a kanji character correspond to each other.