JP2530659B2 - Optical character reading system - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character reading system, and more particularly to an optical character reading system suitable for a user to form a desired word dictionary in the optical character reading system. Regarding the system.

[Conventional technology]

In the conventional optical character reading system, the word dictionary referred to when performing word matching is fixed in advance in the system,
Further, the types of the words are limited, for example, prefecture names, municipalities names, and the like. Therefore, in the conventional optical character reading system, only the word in the same field as the word type attribute of the word dictionary can be matched, and it is impossible to match any word.

Examples of devices related to this type include, for example, JP-A-60-584 and 59-197974.

[Problems to be solved by the invention]

In the above-mentioned conventional technique, a word dictionary to be collated is not prepared for a word of a type peculiar to a user, for example, a word name such as a bank name in the banking industry and a product name code in the distribution industry, and word collation cannot be performed. There was a problem.

Further, in the word matching in the conventional optical character reading system, there is a problem that it takes a lot of time to process because all the combinations of a plurality of candidate characters that have been character-recognized are matched.

It is an object of the present invention to provide an optical character reading system that enables word matching with respect to a word of a user-specific type, and that can speed up word matching.

[Means for solving problems]

The optical character reading system of the present invention reads a character pattern on a form and recognizes characters on a character-by-character basis, a word dictionary including registered words, and a word recognized as a registered word in the word dictionary. Is applied to an optical character reading system that outputs a word matching result, and has the following features.

That is, the word dictionary comprises a word table and an address table. Here, in the word table, each word of an arbitrary word type source is rearranged according to a certain standard, and for each rearranged word, at least the second character of the word is the same as another word. It is composed of a pointer indicating the address on the word table and the code of the word.

Further, the address table is configured to include an address in the word table of the word in the first character of each word in the word type source and an address in the word table in the second character of each word.

Further, the word matching unit obtains the address of the first character or the second character of the character-recognized word on the word table from the address table, and matches the address with the word table.

[Work]

According to the present invention, a user uses a desired word type source to create a desired word table and an address table, which are registered as a word dictionary. As a result, it becomes possible to perform word matching on the word type peculiar to the user as in the field of the word type peculiar to the device.

Further, according to the present invention, the address table is searched for the first character or the second character of the character-recognized word, and the word in the word table existing at the searched address is collated with the character-recognized word. , It is possible to speed up word matching.

〔Example〕

Hereinafter, the present invention will be described in more detail with reference to the embodiments shown in the accompanying drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention.
As shown in FIG. 1, this optical character reading system is composed of an optical character reading device 1 and a general-purpose computer 2. The characters entered on the form 3 are optically scanned by the photoelectric conversion unit 11 to convert the characters into a binary pattern. The character cutout unit 12 cuts out the binarized pattern for each character,
The character recognition unit 13 performs character recognition on a character-by-character basis, and stores the result in the character recognition result candidate buffer 15. The word matching unit 14 matches the contents of the character recognition result buffer 15 with the word dictionary 17, and stores the best word matching result in the word matching result buffer 16.

Specific examples of word matching are shown in FIGS. 2 (a), (b), (c),
This will be described in more detail with reference to (d). The reading of the prefecture field 21 of the form 3 shown in FIG. 2A will be described as an example. The character "Kanagawa" entered in the prefecture field 21 is recognized on a character-by-character basis, and as shown in FIG. 2B, the first to fourth characters from the first to third characters are selected. Each candidate character is stored in the character recognition result candidate buffer 15.

The word matching unit 14 matches each candidate in the character recognition result candidate buffer 15 with a word registered in the prefecture table 22 of the word dictionary 17 shown in FIG. The available words are stored in the word matching result buffer 16. In the case of this example, as shown in FIG. 2D, the word "Kanagawa" is output as the most matched word. Here, the word field is
When the city-town is hierarchically structured, word matching can be performed based on the information of the hierarchical structure.

The feature of this embodiment is that the user can freely set the contents of the word dictionary 17 shown in FIG. 1 based on the word type source 4 by the processing of the general-purpose computer 2 and the control unit 18. Next, the processing will be described. The word type source 4 shown in FIG. 1 is created by a database or the like. The contents of the word type source 4 are, in the general-purpose computer 2, the character code order of the first character of each word (for example, software JIS code order).
Are sorted into, and after assembling, are registered in the word dictionary 17 via the control unit 18.

The details of word registration will be described below. First, the user-specific word types are sorted in the order of character codes.
This is because, when searching for a word containing the corresponding character, its storage address can be uniquely determined by a certain defining expression. This is also because, in order to use the word type source 4 created by a database or the like here, generally, the above-mentioned hierarchical structure of words can be easily obtained.

Next, the sorted word types are assembled to create a word dictionary. The word dictionary is an address table as shown in Fig. 3.
It consists of 31 and word table 32. Address table 31 is a word table
For fast access to 32, the first character word table pointer 33 when the first character of the word is the key and the second character word table pointer 34 when the second character is the key Composed. Access to the address table is as follows. If the character code is Shift JIS (2-byte configuration), the upper byte is C _U , the lower byte is C _L , and the hexadecimal display shows C _L = C _L − (40) _H …… (1) C _U = C _U − (81) _H ... (2). However, when C _U <(AO) _H , C _U = C _U −
(40) Substituting the value of C _U obtained by the equation of _H into the right side of equation (2)
_Ask for C _U. If the first character word table pointer 33 and the second character word table pointer 34 is a table of 1 byte configuration, (1), from the beginning of the address table 31 using a C _L, C _U obtained in (2) The relative address A of is calculated by the following equation (3).

A = (C _U · 192 + C _L ) · 2 (3) By searching the word table with the word table pointer stored at the address A, the word having the first character as the corresponding code can be known. It is also possible to search for a word using the second character as a key, and the method of obtaining the relative address in the address table from the character code of the second character is the same as the case of the first character. The address of the word appearing first from the beginning of the word table among the words having the corresponding character as the second character is stored in the second character word table pointer 34 at address A + 1 of the address table 31. The key to the second character is the first
This is especially used when the letter characters cannot be used due to rejects. In addition to the word 37, the word table 32 includes a word length 35 indicating the length of the word and a second character link pointer 36 used when the second character is used as a key. The second character link pointer 36 stores the relative address of the word having the same second character from the word, and the second character is linked by linking until the relative address becomes zero. Can get the same word.

Next, FIGS. 4A, 4B, and 4B show a specific example of sorting and assembling the contents of the word type source 4.
This will be described with reference to (c) and (d). First, FIG. 4 (a)
Based on the Shift JIS code of the first character of each word registered in the word type source 4 shown in FIG. 4, the words are rearranged in the order of smaller Shift JIS code, and the sorted word type source 41 shown in FIG. To create. Next, the word table 32 shown in FIG. 4C is created based on the sorted word type source 41.
The length of each word 37 in the sorted word type source 41 is calculated in bytes and stored as a word length 35. Also, the second word
When the same character as the first character matches the second character of another word, the difference between the storage address of the word and the storage address of the word including the same character is the second character link pointer 36.
Is stored as In this example, the second letters of Tokushima and Fukushima are the same, so (22) _H- (12) _H = (10) _H is the second
It can be calculated as the character link pointer 36. If there is no word including the same character, 0 is stored as the second character link pointer 36. Further, addresses 0, A, 12 ... In the word table 32 indicate addresses in hexadecimal.

Finally, the word table 32 to the address table shown in FIG. 4 (d)
Create 31. Address table 31 arranges each character in word table 32 in order from the smallest Shift JIS code, and shift JIS
There is a one-to-one correspondence with the code. The method for creating the address table 31 is as follows. First, the first character word table pointer 33 and the second character word table pointer 34 for each character are all initialized to (FF) _H. Next, the word table address of the corresponding word is stored in the first character word table pointer 33 of the address table 31 corresponding to the code of the first character of the word 37 of the word table 32. Here, if the content of the address table 31 is not (FF) _H , the content is not updated. In the case of this example, the content (1A) _H of the address table corresponding to Fuku is not updated by the word Fukushima. Because it was already set by Fukuoka before that. For the second character, similarly, the second character word table pointer 34 is obtained, and this operation is performed for all the words 37 in the word table 32.
The address table 31 can be created by performing the above.

As described above, the address table 31 for high-speed word access and the word table 32 to which the word length 35 and the second character link pointer 36 are added
The word dictionary object consisting of
Is registered as a word dictionary 17 in the optical character reader. At the time of registration, as shown in FIG. 3, a word dictionary management table 38 for storing the head addresses of the word dictionary address table 31 and the word table 32 and their capacities is prepared for each word type.

The word collating unit 14 shown in FIG. 1 collates the contents of the character recognition result candidate buffer 15 with the word dictionary 17 as follows. That is, it is assumed that the first candidate character in the character recognition result candidate buffer 15 is God. In this case, since the content of the first character word table pointer 33 corresponding to the god of the address table 31 is 0, the content of the address 0 of the word table 32 is referred to. Thereby, the word 37 of Kanagawa is obtained as the matching result.
In the case of this example, since there is no other word whose first character is God, it is uniquely determined. If there is another address (in this case, the address exists at the next address), the next address is also referred to. Also, if the collation with the first candidate character fails,
Matching using the second candidate character is performed. in this case,
The address of the word table 32 is obtained by using the second character word table pointer 34 of the address table 31, and the address is compared with the word 37 existing at that address. If the second character pointer link 36 of the corresponding address is not 0, the second character pointer link
It is also matched with the word 37 at the address indicated by 36. As described above, according to the present embodiment, it is not necessary to match the candidate characters in the character recognition result candidate buffer 15 with all the words in the word dictionary, and the matching process can be significantly speeded up.

In addition, since the user can create the word dictionary 17 using the word type source 4, it is possible to provide a highly versatile optical character reading system.

Although the word collation unit 14 and the word dictionary 17 are provided in the optical character reading device 1 in the above embodiment, the word collation unit 14 and the word dictionary 17 are provided in the general-purpose computer 1 and the character recognition result candidate buffer is provided. It is also possible to easily send the contents of 14 to the general-purpose computer 1 and perform word matching in the general-purpose computer.

〔The invention's effect〕

According to the present invention, even for a word type peculiar to a user, by assembling the word type and registering a word dictionary consisting of a word table and an address table, the word matching can be performed similarly to the field of the word type provided by the apparatus. It is possible to achieve high recognition accuracy.

Further, the address table and the word table described above enable high-speed word matching.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIGS. 2 (a), (b), (c), and (d) are explanatory views showing the outline of the processing of the embodiment shown in FIG. FIG. 3 is an explanatory view showing an example of the word dictionary shown in the embodiment of FIG. 1, FIG. 4 (a),
(B), (c), (d) is explanatory drawing which shows the specific example of word dictionary creation of the Example shown in FIG. 1 ... Optical character reader, 2 ... General-purpose computer, 4 ... Word type source, 14 ... Word collation unit, 17 ... Word dictionary, 31 ...
… Address table, 32 …… Word table, 33 …… Word dictionary management table.

Claims (1)

(57) [Claims] 1. A character recognition means for reading out a character pattern on a form for character recognition on a character-by-character basis, a word dictionary having registered words, and a word for collating the registered words in the word dictionary with the recognized words. In an optical character reading system that includes a collating unit and outputs a word collating result, the word dictionary is configured to include a word table and an address table, and the word table is a constant word for each word type source. For each word rearranged according to the criteria, rearranged, at least the second character of the word is composed of a pointer indicating the address on the word table of another word and the code of the word, and The address table includes an address in the word table of the word at the first character of each word in the word type source and an address in the word table at the second character of each word. An optical character characterized in that the word matching unit obtains the address of the first character or the second character of the character-recognized word on the word table from the address table, and matches the address with the word table. Reading system.