JPH02212990A - Character reader - Google Patents

Abstract

PURPOSE:To execute character reading processing by using a dictionary corresponding to sub set and to easily change the contents of the dictionary by reading a standard pattern, with which collation is executed in correspondence to the sub set, at random from the dictionary in which the dictionaries of plural categories are stored without being overlapped. CONSTITUTION:In a standard pattern designation information storage part 16, standard pattern designation information are classified and stored for each sub set to designate the standard pattern of the specified category out of the plural standard patterns of the dictionary for large group classification processing stored in a large group classification dictionary memory 15. The plural standard patterns corresponding to a reading candidate character group, which is designated by character group designation information, are read from this storage part 16 based on the standard pattern designation information. Then, a character written to a slip is read by collating the read standard pattern with a character pattern which is the object of character reading processing. Thus, the number of the sub set dictionaries is increased without widely increasing the capacity of the dictionary and the maintenance of the sub set dictionary can be easily executed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a character reading device that reads characters by comparing character patterns with preset standard patterns.

(Prior Art) Generally, a form on which character reading processing is performed by a character reading device is provided with a plurality of reading fields indicating positions where characters are written. Then, the character reading device performs a character reading process on information specifying a specific group consisting of the character type of the characters to be written (hereinafter referred to as a subset) according to each reading field provided in the form. preset before. The character reading device executes character reading processing using a dedicated dictionary according to the subset (hereinafter referred to as a subset dictionary).

For example, if the reading field of a form where character reading processing is performed is specified to contain only numbers, such as an amount field, by setting a subset that specifies numbers in advance, a standard pattern for numbers can be created. Character reading processing is performed using the subset dictionary in which the characters are stored. In addition, if the reading field is specified to be filled in only katakana, such as in the furigana column, by setting a subset that specifies katakana in advance, it is possible to use a subset dictionary that stores standard patterns related to katakana. Character reading processing is performed.

In addition, subsets are not limited to those that specify character types;
Regarding the reading field where kanji are written, it is possible to specify a specific group according to the content of the entry. For example, if a reading field is specified to be filled with kanji, such as a name field or an address field, a subset dictionary designed for the kanji characters often used in names will be used for the name field, and a subset dictionary designed for the kanji characters often used in names will be used for the name field, and a dictionary for the address field will be used for the name field. Each subset is preconfigured to specify a dictionary designed for characters commonly used in addresses.

This is because there are so many kanji characters (JIS-1
In order to improve the processing speed, reading accuracy, etc. of character reading processing, a subset dictionary is used for a specific group according to the content entered in the reading field.

(Problems to be Solved by the Invention) However, if subset dictionaries are configured according to the contents entered in each reading field, duplicate dictionary information may be included in each subset dictionary. A specific example of this will be explained. When reading a transfer request form from a financial institution (bank) as a form to be processed, a dictionary regarding bank names and branch names is used. Here, the dictionaries for reading the bank name and for reading the branch name each contain 7.
If a dictionary capacity of 00 characters and 1500 characters is required, the total dictionary capacity of all subset dictionaries will be about 2200 character types. Since there are usually many characters commonly used in bank names and branch names, all subset dictionaries include
There is about 400 duplicate dictionary information out of 2200 characters.

In this way, when performing the reading process of Chinese characters, if subset dictionaries are configured to correspond to each reading field, common dictionary information may be stored redundantly in a plurality of subset dictionaries. For this reason, when dealing with a plurality of subset dictionaries, there is a problem in that the dictionary capacity increases significantly. Furthermore, when changing the contents of dictionary information regarding a certain character, it is necessary to change all subset dictionaries that include the dictionary information of the character to be changed.

The present invention has been made in view of the above points, and provides a character reading device that can increase the number of subset dictionaries without significantly increasing the dictionary capacity, and can easily maintain the subset dictionaries. The purpose is to

[Structure of the Invention] (Means for Solving the Problems) The present invention uses standard pattern storage means in which standard patterns to be compared with character patterns to be processed are stored without duplication. Based on the standard pattern specification information that specifies a specific standard pattern classified into each character group specified by the character group specification information, multiple standard patterns corresponding to the reading candidate character group specified by the character group specification information are specified. and read it out,
The system is configured to read characters written on a form by comparing the read standard pattern with the character pattern to be subjected to character reading processing.

(Operation) In the character reading device configured in this manner, a plurality of standard patterns specified by the character group specification information are classified into standard pattern specification information classified by character group specified by the character group specification information. It is read out based on the For this reason, instead of preparing a standard pattern dictionary (subset dictionary) for each character group in which some characters often overlap, the only dictionary that stores the same standard pattern without duplication (standard pattern storage means ) is used, it will not be a change in any way.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a portion of a character reading device that performs character reading processing according to the same embodiment. In the figure, a reading control unit it controls the various parts constituting the device and executes character reading processing. In the main memory 12,
A processing program for character reading processing executed in the reading control unit 111, various information, etc. are stored. Reading control section 1
1 reads a processing program stored in the main memory 12 and executes processing according to the program. The major classification sampling memory 13 stores standard patterns to be subjected to character reading processing, which are obtained by performing preprocessing, sampling, etc. on input character patterns.

The major classification unit 14 performs individual recognition processing similar to the sampling pattern by comparing the dictionary stored in the major classification dictionary memory 15 to be described later with the sampling pattern stored in the major classification sampling memory 13. Performs major classification processing to select candidate characters. The major classification dictionary memory 15 stores a dictionary for major classification processing. The dictionary for major classification processing stored in the major classification dictionary memory 15 stores standard patterns designed for each type of character (hereinafter referred to as a category) to be compared with the sampling pattern without duplication. has been done. The standard pattern designation information storage unit 1B classifies standard pattern designation information for designating a standard pattern of a specific category into subsets from among the plurality of standard patterns of the dictionary for major classification processing stored in the major classification dictionary memory 15. is stored. The individual recognition sampling memory 17 stores the sampling pattern stored in the major classification sampling memory 13 for which the major classification process has been completed, as a target of the individual recognition process to be performed after the major classification process has been completed. The individual recognition unit 1g includes an individual recognition dictionary memory 19, which will be described later.
By comparing the dictionary stored in the memory 17 with the sampling pattern stored in the individual recognition sampling memory 17, individual recognition processing is performed to determine one character indicated by the sampling pattern. The individual recognition dictionary memory 19 stores a dictionary for individual recognition processing, and is read out based on the contents stored in the major classification candidate list memory 20, which will be described later. The major classification candidate list memory 20 stores information for specifying, from the individual recognition dictionary memory 19, the standard pattern of the category corresponding to the candidate character for the individual recognition process obtained by the major classification process. It is written by the reading control unit 11 according to the result.

FIG. 2 is a diagram showing the relationship among the major classification sampling memory 13, the major classification section 14, the major classification dictionary memory 15, and the standard pattern designation information storage section 18 shown in FIG. The sampling pattern storage unit 31 of the major classification sampling memory 13 stores sampling pattern data, and
According to the value of 2, the main classification i14 is read out. The standard pattern storage section 33 of the major classification dictionary memory I5 stores standard pattern data of characters to be processed, and is read out to the major classification section 14 according to the value of the address counter 34.

The detailed configuration of this standard pattern storage section 33 is shown in FIG. The address counters 32 and 34 are incremented by the dimensional clock output from the major classification unit 14 when the comparison with the standard pattern for one character in one category is completed. The subset table memory 35 of the standard pattern designation information storage section 16 is connected to the standard pattern storage section 3.
The start address (standard pattern designation information) of the standard pattern data of a specific category stored in No. 3 is stored according to the subset. The detailed structure of this subset table memory 35 is shown in FIG. The address counters 32, 34, and 36 receive the character end clock output from the major classification section 14 when the comparison with the standard pattern for one category is completed. The address counter 3B receives a load signal and an address signal from the reading control section 11, and specifies address data stored in the subset table memory 35 in accordance with each signal.

Next, the operation of this embodiment will be explained.

First, the subset table memory 35 shown in FIG.
The relationship between this and the standard pattern storage section 33 shown in FIG. 4 will be explained. The standard pattern storage section 38 stores standard pattern data for each category without duplication. Further, the dictionary for each category stores standard pattern data for a plurality of characters included in the category. In other words, for multiple identical characters with different shapes,
Standard patterns are designed for each. The dictionary (standard pattern) for each category is specified by its start address ("aaaa bbbb"...). The subset table memory 35 stores addresses (11 quasi-pattern designation information) that designate dictionaries for each category stored in the standard pattern storage section 33, classified for each subset. In addition, the set of addresses classified into each subset is the first address (“
``AAAA'', ``BBBB''). Furthermore, in the area starting from the address ``AAAA'' of the subset table memory 35, a dictionary storing standard patterns of kanji related to the ``color'' specified by the subset is stored. It is assumed that an address to be specified is stored, and an area starting from the address "BBBB" stores an address that designates a dictionary storing standard patterns of Kanji characters related to "day of the week" designated by the subset.

First, before reading the characters written on the form,
A group of characters to be written in each reading field is specified in advance by a subset. Here, it is assumed that a kanji related to "color" is specified for a certain reading field. The reading control unit 1 specifies a dictionary of a specific category in the standard pattern storage unit 33, as shown in FIG. 3, in the subset table memory 35 of the standard pattern designation information storage unit 18 according to the specified subset. Set address.

When the character reading process is started, the character pattern of the characters written in the reading field is detected from the form image obtained by optically scanning the form, for example, in a scanning mechanism section (not shown). The detected character patterns are preprocessed and sampled by the reading control unit 11 and stored in the major classification sampling memory 13. When a sampling pattern is stored in the sampling pattern storage section 81 of the major classification sampling memory 18, a major classification process is performed on this sampling pattern. This major classification process will be explained with reference to FIG. The reading control unit 11 reads the subset table memory 3 corresponding to the subset specified for the reading field in which the character pattern to be processed is written.
Address signal indicating address 5 (here, °AAAA
”) and a load signal to the standard pattern designation information storage unit IB.Address counter 3B
When each signal from the reading control unit 11 is input manually, "aaaa" which specifies the dictionary stored in the area indicated by "AAAA" of the subset table memory 35 is read out, and the address counter of the major dictionary memo I No. 15 is read out. 34. When the address "aaaa" is loaded into the address counter 34, when the address "aaaa" is specified, the dictionary of the kanji red category in the standard pattern storage section 33 is read out to the major classification section 14. be done. In addition, the reading control unit 0
causes the major classification unit I4 to read out the sampling pattern data stored in the sampling pattern storage unit 31. Major classification section 14
compares the standard pattern data read from the standard pattern storage section 33 with the sampling pattern data from the sampling pattern storage section 31.

The major classification unit 14 reads the reading result (similarity value with the standard pattern data) obtained by performing the matching #A
It is output to the s section 11. When the major classification unit I4 finishes matching one character in the "red" category with the standard pattern, it outputs a dimensional clock to the address counters 32 and 34. As a result, the values of each counter 32, 34 are incremented, and the sampling pattern storage section 31. The next address data stored in the standard pattern storage section 33 is then read out to the major classification section 14 and processed. That is,
The standard pattern data indicating other features in the "Red ° category" in the standard pattern storage section 33 are sequentially read out and processed. In this way, when the processing for the "Red 1" category dictionary is completed, the main classification section 14 is address counter 3
Output character end clock for 2.34.35.

As a result, the address counter 3B is incremented and "stored in the area of the next address of AAAA="
Specify bbbb'. And address counter 34
“b” is input by inputting the character end clock.
bbb” is loaded. Also, the address counter 32
is cleared by the character end clock.

Next, when the address counter 34 is loaded with "the first address of the dictionary for the blue 2 category"bbbb', the major classification unit 14 selects the blue category as described above.

The standard pattern in the category and the sampling pattern are sequentially compared, and the reading result is output to the reading control unit ti.

In this way, when the comparison with the standard pattern of kanji related to "color" specified by the subset (°white° "yellow" water "°green"...) is completed, the reading control section 11 starts the main classification section. Based on the reading results from 14, categories (candidate characters) to be subjected to individual recognition processing whose similarity to the sampling pattern is higher than a predetermined value are selected. The reading control unit 11
The starting address indicating the dictionary for individual recognition processing (standard pattern for each category) corresponding to each candidate character stored in the individual recognition dictionary memory 19 is written into the major classification candidate list memory 20. Further, a sampling pattern for individual recognition processing is stored in the individual recognition sampling memory 17. Individual recognition processing is
The individual recognition unit 1B compares the standard pattern read out based on the address specifying the dictionary of a specific category stored in the major classification candidate list memory 20 with the sampling pattern stored in the individual recognition sampling memory 17. It will be held in The individual recognition unit 18 outputs the processing result of the individual recognition process to the reading control unit 11. The reading control unit 11 determines the one with the highest degree of similarity from the processing results from the individual recognition unit 18 as the character reading result.

Here, a case will be described in which kanji related to "day of the week" are set as a subset for a certain reading field. In this case, the reading control section 11 stores the starting address "" in the standard pattern designation information storage section 1B as shown in FIG.
Address “PPPP” qqqqq that specifies a dictionary of categories related to “day of the week” in the area starting with “BBBB”
Set... This set address includes the address eeee'' that specifies the dictionary of the character ``water'' that is common to the group of kanji related to ``color'' and ``day of the week'' specified by the subset. From the dictionaries of each category stored in the pattern storage section 33, a dictionary of "water" is read out at random.

That is, the dictionary for "water" stored in the area of address "eeee" shown in FIG.
eeee" stored in each area from the start address 'AAAA'°BBBB". Therefore, it is not necessary to set a plurality of dictionaries of characters common to a plurality of groups specified by the subset in the standard pattern storage section 33.

Note that in the above embodiment, a two-stage configuration was used in which detailed individual recognition processing was performed after performing major classification processing.
Final individual recognition may be performed in one step.

[Effects of the Invention] As described above, according to the present invention, a subset table is created in which information specifying a dictionary of a specific category is set according to a subset from a dictionary in which dictionaries of a plurality of categories are stored without duplication. Since standard patterns to be checked are read out at random based on the contents of the memory, character reading processing can be performed using a dictionary corresponding to a subset without increasing the capacity of the dictionary. Furthermore, even if it is necessary to change the contents of a dictionary for a certain character, it is possible to do so easily because it is only necessary to correct the dictionary to be changed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a part that performs character reading processing of a character reading device according to an embodiment of the present invention, FIG. 2 is a block diagram showing details of a part of the configuration shown in FIG. 1, and FIG. FIG. 3 is a diagram showing the contents stored in the subset table memory, and FIG. 4 is a diagram showing an example of the standard pattern storage section. 11...Reading control unit, 12...Main memory, 13
...Large classification sampling memory, 14...Large classification section, 15
...Main classification dictionary memory, IB...Standard pattern designation information storage section, 17...Individual recognition sampling memory, 18.
...Individual recognition unit, 19...Individual recognition dictionary memory, 20
... Major classification candidate list memory, 31... Sampling pattern storage section, 33... Standard pattern storage section, 35...
- Subset table memory.

Claims (1)

[Scope of Claims] A preset character for specifying a character written in a form to be subjected to character reading processing as one of a plurality of character groups consisting of a character pattern of the character and a reading candidate character group. In a character reading device that reads by comparing group designation information with a plurality of standard patterns, the standard pattern storage means stores the standard patterns without duplication, and the identification information stored in the standard pattern storage means standard pattern designation information storage means storing standard pattern designation information that designates standard patterns for each of the character groups; and a plurality of standards corresponding to the reading candidate character groups designated by the character group designation information. standard pattern reading means for reading out a pattern from the standard pattern storage means based on the standard pattern designation information stored in the standard pattern designation information storage means; A character reading device that reads characters written on the form by comparing the pattern with a character pattern to be subjected to character reading processing.