JPS63157292A - Hand-written kanji ocr device - Google Patents

Abstract

PURPOSE:To accurately discriminate similar KANJIs (Chinese character) by constituting the titled device so that employed candidate characters are replaced with discrimination-unable characters in the case of phrases of allowable combination, storing the results in a text memory, and the results are inputted for a confirmation work by an operator by conversa tional format through CRT screen after ending the processing of the discrimination-unable characters of whole text. CONSTITUTION:An original 1 is set in a hand-written KANJI OCR part 2, and the KANJIs are optically read. A processor 4 discriminates an output character from the part 2 by means of pattern aligning method as referring to a standard pattern memory 5. A discriminated character is stored in the form of character code in a text memory 6, but for a character that is not discriminated, a discrimination-unable code is stored in the memory 6. The proces sor 4 obtains a phrase which is the combination of respective candidate characters listed up and characters before and after the discrimination-unable character concerned in the text memory, and checks whether or not this phrase is registered in a dictionary memory 3. When it is the phrase of existing combination, the candidate character is employed, and it is substituted in the part of a discrimination-unable code in the text memory 6. When the phrase is not present in the dictionary memory, the discrimination-unable code is left as it is in the text memory, then, it is confirmed by an operator.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a handwritten kanji OCR (OCR is an abbreviation for optical character reader) device, and particularly relates to an improvement in the character discrimination function.

[Prior Art] Devices for reading handwritten kanji using OCR are well known. By the way, there are many examples of kanji in which different types of characters are expressed by the presence or absence of dots, minute differences in line elements, or differences in the positions of similar line elements.

Examples of the presence or absence of points are ``king'' and ``tama,'' examples of minute differences in line elements are ``sign'' and ``micro,'' and examples of differences in position of similar line elements are ``large'' and ``. "Length" etc.

[Problems to be Solved by the Invention] The Kanji character shapes for OCR must be such that even such subtle differences can be clearly distinguished, and there are strong restrictions on the distinguishable Kanji character shapes.

The problem was that it was difficult to use.

The present invention has been made in view of the above points, and it is an object of the present invention to accurately distinguish similar kanji as described above (herein, kanji includes kana characters, symbols, etc.) An object of the present invention is to provide a handwritten kanji OCR device that can relax the restrictions on handwritten kanji characters.

[Means for solving problems] In order to achieve such an objective, the present invention.

A handwritten kanji OCR unit that optically reads kanji on a manuscript, and a dictionary memory that stores kanji and words that are combinations of two or more kanji characters.

For an unidentifiable character, find a combination of a candidate character and the characters before and after the unidentifiable character, check the word in the dictionary memory, and if it is a combination of words that is allowed, adopt the character of that word. However, if the word is not in the dictionary, a processor that attaches a specific code to the unidentifiable character and stores the result in a text memory, and a CRT display device that displays the characters and symbols on the screen. and a CRT interface for displaying the contents of the text memory on a CRT display device.

The present invention is characterized by comprising a keyboard for inputting characters, symbols, and control information to the processor.

[Operation] In the present invention, a kanji manuscript is read by OCR, and the processor analyzes the characters read from the manuscript and stores the corresponding character code in the text. For characters that cannot be distinguished, candidate characters and their surroundings are The word/phrase in combination with characters is checked in the dictionary memory, and if the word/phrase is in an allowed combination, the adopted candidate character is replaced with an indistinguishable character and stored in the text memory.

After the processing of indistinguishable characters in the entire text is completed, the operator further confirms the text on the CRT screen and in a conversational manner and converts the text into appropriate characters by inputting information from the keyboard.

[Example code] Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a handwritten kanji OCR device according to the present invention. In FIG.

3 is a dictionary memory that stores a kanji dictionary; 4 is a processor; 5 is a standard pattern that stores standard character shape patterns of kanji and symbols (hereinafter collectively referred to as kanji); memory, 6 is a text memory in which the character code of the kanji of the read manuscript is stored;
7 is a CRT interface, 8 is a CR that displays kanji
T display device, 9 is a keyboard for inputting information.

The operation in such a configuration will be explained next with reference to the flowchart of FIG. A manuscript 1 is subjected to a handwritten kanji OCR section 2 to optically read kanji. The processor 4 refers to the standard pattern memory 5 and uses a pattern matching method to identify the characters output from the handwritten kanji OCR section 2. Note that the pattern matching method is performed using a known method.

Discriminated characters are stored in the text memory 6 as character codes, and characters that cannot be discriminated are stored as undiscernible codes in the text memory 6.

After reading into the text memory 6 in this manner, the processor 4 selects a plurality of candidate characters for the 6 unidentifiable characters, which are subjected to the following processing for the unidentifiable code. A word combination including each of the listed candidate characters and the characters before and after the unidentifiable character in the text memory (these characters are distinguishable characters) is found, and the combination word is stored in the dictionary memory 3. It is checked to see if it is registered, and if the combination of words exists on the dictionary memory, the candidate character is adopted and replaced with the unidentifiable code part on the text memory 6.

If the word is not in the dictionary memory, all other combinations of words are checked.
If it is still not in the dictionary memory, the unidentifiable code is left in the text memory as is, and the I! code in the text memory 6 is highlighted for the operator to confirm. I's information is added.

A specific example of processing for unidentifiable characters is as follows. For example, as shown in FIG. 3, if the character 31 is an unrecognizable character, it is identified as a candidate character using a pattern recognition technique, and is identified as "length".

List ``r sentence'' and ``r large''. Then, it is assumed that there are "oaza", "character", and "oaza" as words and phrases in combination with "1 character after 1 unidentifiable character", and the processor checks each of these words and phrases. First of all, ``r large letter'' is a word that is not found in the dictionary, and is the primary r character that is determined to be incorrect.''
is in the dictionary and is determined to be the correct word. "Oaza" is in the dictionary and is determined to be the correct word, but in this case, the frequency of occurrence is "
Since ``letter'' is higher, we use ``r letter'' as j11a complement, and ``
``Oaza'' has 12 candidates. In the text memory, one character sentence of the word adopted as the first candidate is replaced with an unidentifiable character.

The final decision as to whether to use "r character" for "r sentence" or "large character" for "r large" is left to the operator's choice.

The CRT interface 7 displays the characters corresponding to the character codes in the text memory 6 on the CRT display device i! Display on 8. Characters that are indistinguishable and have control information for highlighted display are displayed with high brightness on the CR7 screen of the CR7 display device 8, or in a special display format such as inversion or blinking.

After completing the search for the entire text as described above, the operator checks these characters on the CR7 screen in conversation mode.

If appropriate, press “YES (or Y)” from keyboard 9.
If the text is incorrect, enter "No (or N)" and complete the text by selecting the next candidate or by directly inputting (inputting characters or character codes).

Note that a known method can be used to select the next candidate or directly input it.

[Effects of the Invention] As explained in detail above, according to the present invention, for an unidentifiable character, it is checked whether the combination of a candidate character and the characters before and after the unidentifiable character is valid. Since the adoption was decided based on the above, accurate readings can be expected. In this way, even characters that cannot be identified by conventional OCR can be identified or confirmed by an operator, so restrictions on handwritten characters can be relaxed.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram showing an embodiment of handwritten kanji 0CR1 according to the present invention, FIG. 2 is a flowchart for explaining the operation, and FIG. It is a figure showing a concrete example. 1...Handwritten manuscript, 2...Handwritten kanji OCR department. 3...Dictionary memory, 4...Processor, 5...Standard pattern memory, 6...Text memory, 7...
CRT interface, 8... CRT display device, 9
···keyboard.

Claims (1)

[Scope of Claims] A handwritten kanji OCR unit that optically reads kanji on a manuscript; a dictionary memory that stores kanji and words consisting of combinations of two or more kanji characters; Find the combination with the characters before and after the unidentifiable literature, check the word in the dictionary memory, and if it is a combination of words that is allowed, use the characters of that word, and if the word is not in the dictionary, A processor that attaches a specific code to the unidentifiable characters and stores the result in a text memory, a CRT display device that displays the characters and symbols on the screen, and a CRT display device that displays the contents of the text memory. a CRT interface for displaying on the screen, and a keyboard for inputting characters, symbols, and control information into the processor, and the processor analyzes the characters read from the original and stores the corresponding character code in the text. However, for unrecognizable characters, the word combination of the candidate character and the characters before and after it is checked in the dictionary memory, and in the case of a combination of words that is allowed, the adopted candidate character is replaced with the unrecognizable character. After the entire text has been processed for unintelligible characters, the operator can confirm the text on the CRT screen and in a conversational manner, and input information from the keyboard to convert the text into appropriate characters. A handwritten kanji OCR device featuring the following.