JPH04199454A - Document input device - Google Patents

Abstract

PURPOSE:To improve the recognizing accuracy of the post-processing by carrying out the post-processing before putting a character code of a null character and a null line into a character code string. CONSTITUTION:A shaping part 27 calculates the average character and line pitches of a character string based on the coordinates of each character read out of a character information buffer 23 and then divides the coordinates of each character by the average character and line pitches for calculation of the row and the column of each character included in the character code string. In this case, the character code of a null character and a null line are put into the character code string outputted from a post-processing part 25 so as to secure the approximation to a character position on an original. Then a coded document is outputted through a terminal 28. In such a way, the post- processing is carried out before the character code of the null character and the null line are put into a character code string. So that the word of the character code string is never divided by the character code of the null character and the null line. Thus the recognizing accuracy is improved for the post- processing.

Description

[Detailed Description of the Invention] [Summary] The present invention relates to a document input device that reads a document and outputs it for storage. a position measuring means for measuring the position of each character to obtain position information; a recognition means for recognizing each character from the image data and converting it into a character code; and a position measuring means for each character obtained by the position measuring means. a storage means for storing at least a character code string obtained by the recognition means, a post-processing means for comparing the character code string obtained by the recognition means with a word dictionary and correcting unrecognized parts and erroneously recognized parts, and characters outputted by the post-processing means. and a formatting means for inserting character codes of blank characters and blank lines into the code string so as to correspond to the position of each character on the document based on the position information of each character in the storage means.

[Industrial application field]

The present invention relates to a document input device, and more particularly to a document input device that reads and stores documents.

[Conventional technology]

FIG. 2 shows a block diagram of an example of a conventional document input device.

In the figure, image data read from a document by an optical reading device 10 is stored in an image memory 11. Character position measurement section 1
2 measures the position of each character to obtain position information, and a character recognition unit 14 recognizes each character and converts it into a character code.

In order to preserve the appearance of the manuscript, the formatting unit 15 inserts blank character codes between character codes as necessary based on character position information so that each character position in the manuscript can be reproduced on the coded document. .

The recognition rate of the document recognition unit 14 is not 100%, and there are many parts that cannot be recognized in coated documents, or incorrectly recognized parts of isomorphic or similar characters such as the katakana ``工'' and the kanji ``工''. exist. The post-processing unit 16 corrects rejected parts and erroneously recognized parts by comparing the coded character string with a word dictionary 17, etc., and outputs the character code of the coded document from the terminal 18.

[Invention or problem to be solved]

In the conventional device, the formatting section 15 is placed in the middle of the recognized character code string.
Post-processing is performed after inserting a blank character code. As a result, words may be separated by blank character codes inserted between character code strings, and the words may not be matched with the word dictionary, resulting in the problem that recognition accuracy cannot be improved through post-processing.

The present invention has been made in view of the above points, and an object of the present invention is to provide a document input device that improves recognition accuracy.

[Means to solve the problem]

The document input device of the present invention includes: a reading device that reads a document to obtain image data; a position measuring device that measures the position of each character from the image data to obtain position information; and a document input device that recognizes each character from the image data to obtain the character. a recognition means for converting into a code; a storage means for storing at least the position information of each character obtained by the position measurement means; post-processing means for correcting erroneously recognized portions; and blank characters and blank lines are added to the character code string output by the post-processing means to correspond to the position of each character on the manuscript based on the position information of each character in the storage means. and a formatting means for inserting a character code.

[Effect]

In the present invention, since post-processing is performed before inserting the character codes of blank characters and blank lines into the character code string, words in the character code string are not separated by the character codes of blank characters and blank lines. Verification with the word dictionary can be performed reliably, thereby improving recognition accuracy through post-processing.

〔Example〕

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

In the figure, 20 is an optical reading device such as a CCD image scanner, and image data read from the original document is stored in an image memory 21. The character position measurement unit 22 cuts out each character from the image data, and stores the coordinates of the apex position of each cut out character in the character information buffer 23 as character position information.

Further, the character recognition unit 24 extracts features of each extracted character, such as the number and direction of lines forming the character, and converts the recognized character into a character code by comparing the extracted characters with a dictionary. Furthermore, if there are multiple candidates at the time of recognition, the character coats of all the candidates are obtained and the character coats are stored in the character information buffer 23 in association with the position information of the character.

The post-processing unit 25 reads the character code from the character information buffer 23 using, for example, a morphological analysis method, compares this character code string with the word dictionary 26, verifies the validity of the word and the validity of the connection between words, and The most suitable candidate for characters is selected and rejected parts and misrecognized parts are corrected.

The formatting unit 28 calculates the average character pitch and line pitch of the character string based on the coordinates of each character read from the character information buffer 23, and divides the coordinates of each character by the average character pitch and line pitch. calculate the rows and columns in the character string, insert character codes of blank characters and character codes of blank lines into the character code string output from the post-processing unit 25 so as to approximate the character positions on the manuscript, The encoded document is output from the terminal 28.

In this way, since post-processing is performed before inserting the character codes of blank characters and blank lines into the character code string, the words in the character code string are not separated by the character hoods of blank characters and blank lines, and the words are Collation with the dictionary can be performed reliably, thereby improving recognition accuracy through post-processing.

Of course, in addition to this, information on the height and width of the characters included in the character position information may be used to add information on the size of the characters onto the coded document.

Further, the character code obtained by the character recognition section 24 may be directly supplied to the post-processing section 25, and only the position information of each character may be stored in the character information buffer 23, and the present invention is not limited to the above embodiment.

〔Effect of the invention〕

As described above, according to the document input device of the present invention, there is no word separation during post-processing, and recognition accuracy is improved through post-processing.
It is extremely useful for practical purposes.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the device of the present invention, and FIG. 2 is a block diagram of an example of a conventional device. In the figure, 20 is an optical reading device, 21 is an image memory, 22 is a character position measuring section, 23 is a character information buffer, 24 is a character recognition section, 26 is a word dictionary, and 27 is a shaping section. Patent applicant Fujitsu Ltd. Figure 1 Block diagram of conventional Ml Figure 2

Claims (1)

[Claims] Reading means (20) for reading a document to obtain image data; position measuring means (22) for obtaining position information by measuring the position of each character from the image data; recognition means (24) for recognizing characters and converting them into character codes; storage means (23) for storing at least position information of each character obtained by the position measuring means; A post-processing means (25) corrects unrecognized parts and erroneously recognized parts by comparing the character code string with a word dictionary, and a storage means (23) for storing the character code string output from the post-processing means (25) 1. A document input device comprising: a formatting means (27) for inserting character codes of blank characters and blank lines to correspond to the position of each character on a document based on the position information of each character on the document.