JPS58182790A - Character discriminator - Google Patents

Abstract

PURPOSE:To have just one time of correction for each character when no reading is performed, by indicating a correct category to a temporary storing circuit to discriminate character. CONSTITUTION:A fixed dictionary F allots the only fixed category to each pattern. In addition to the dictionary F, a variable dictionary V having a variable category is provided along with a temporary storing circuit 3 which stores temporarily the contents of the dictionary V. The possible plural categories are given previously to each pattern of the dictionary V. When characters are discriminated, a correct category is indicated to the circuit 3 from outside based on the correcting category information and then fixed temporarily for discrimination of characters.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character discrimination device that discriminates a character category by comparing information obtained from an input character pattern with a pre-stored dictionary in an optical character reading device.

In recent years, with the rapid spread of electronic computers, there has been a remarkable demand for optical character reading devices. As demand increased, the scope of applications expanded, and the range of reading categories expanded from initially reading numbers to reading a mixture of numbers and kana, and finally numbers, kana, alphabets, and symbols. There is. However, as the number of categories increases, it is not easy to maintain or improve reading accuracy. Furthermore, due to historical factors in which each type of character (English, numbers, kana, and symbols) has been established independently, many characters have similar forms, and this may be due to differences in the number of people writing the characters or the printing style. Depending on the difference in the machine, the number of character sets with higher degrees of similarity will increase.

When humans read these characters, feedback from the entire text, forms, etc. should be included as a judgment factor.
However, in the case of a reading device, this type of feedback cannot be used, and it is extremely difficult to ensure reading accuracy.

In order to ensure reading accuracy, the way characters are written is determined by JIS, but in reality this has not been resolved. Therefore, in the past, in order to avoid misreading, similar characters were strictly checked and marked as unreadable for the time being, and later corrected by humans using the keyboard, and characters that were misread were also checked in the same way. , a method to correct it is being taken.

On the other hand, for each form, one person generally fills out one page, and this may span several pages. It is a general phenomenon that the characters written there reflect the habits of the person writing them.For example,
Depending on the level of training, 95chi can be filled in correctly according to the filling method prescribed by JIS, but 5cs can be filled out correctly.
appears and similar characters appear. Moreover, the content of that 5% varies from person to person.

For this reason, in the above-mentioned correction work using the keyboard, several of the same characters on a certain form become illegible, and it becomes necessary to perform key operations many times each time. father,
Similar problems arise for other characters when the form or the person filling it out changes.

The present invention has been made in view of the above-mentioned problems, and makes it easier to read and correct unreadable characters by adapting the dictionary by taking advantage of the habit of the person writing the characters. We provide a character recognition device that can identify characters with "wacky" characters without having to make corrections each time the same or similar pattern occurs, and improves processing speed and operability. The purpose is to

That is, the present invention provides a device for determining character categories using an optical character reading device or the like, which has a memory configured to store and set a fixed dictionary to which only one fixed category is assigned and a variable dictionary to which multiple categories are assigned. a temporary storage circuit that reads and stores a variable dictionary from the storage circuit; a discrimination circuit that compares discrimination information obtained from an input character pattern with the fixed dictionary and the variation dictionary to variable classify the character category; and a discrimination circuit. control to temporarily fix the variable dictionary by rewriting the content of the temporary memory circuit to adapt to the change in the person who filled it out, according to the correction category information given from the outside as necessary, in response to the output of It is characterized by being equipped with a circuit.

The principle of the present invention is to make the categories variable by allocating a plurality of categories to the dictionary for character discrimination, especially those with similar patterns as variable dictionaries, and to make the categories variable by using a fixed dictionary for categories with no similar patterns. If one of the documents to be read falls under this variable dictionary, it is unreadable and closed, and the human gives a correction category from the outside to temporarily fix the variable dictionary.
This allows characters to be distinguished.

The correction category given from the outside changes the category of the characters entered, including the person's 6 habits, and serves as a temporary category of the corresponding variable dictionary. It can be used almost deterministically and fixedly across multiple images. Therefore, even if the same pattern appears again during reading, the variable dictionary has already been temporarily fixed, so the category can be uniquely determined and discriminated. In this case, the device operates adaptively on a case-by-case basis, and does not learn probabilistically.In other words, categories that are temporarily fixed are re-established when the person who fills in the information changes. Going back, variable or multiple categories will be assigned.

Hereinafter, the present invention will be explained based on embodiments shown in the drawings.

FIG. 1 is a block diagram showing one embodiment of the character discrimination device of the present invention. In the figure, the character discrimination device of the present invention includes a dictionary section 1 consisting of a memory circuit 2 and a temporary memory circuit 3, a discrimination circuit 4, and a control circuit 5.

Memory circuit 2F'i, for example, Institute of Electronics and Communication Engineers journal DVOL
, J64-D A 9 PP877-884 No. 16
It is constructed from a memory such as the data memory shown in the figure, and contains a fixed dictionary F in which categories of characters without similar patterns are fixedly set, and categories V -t- of characters with similar patterns. A variable dictionary V is provided in which the corresponding number is assigned and set as category variable. The output 21ti of the memory circuit 2 and the input of the temporary memory circuit 3 are used to control the temporary memory circuit 3 to temporarily store the contents of the variable dictionary (2).

This control is performed from outside the dictionary section 1, and at the beginning of reading a document or at the timing of the first reading of a group of documents each consisting of a plurality of pages, the variable dictionary V is controlled to enable adaptive control, which will be described later. Transport. This temporary memory circuit 3
Assuming that all the variable dictionaries transferred to v are the contents of this dictionary V, when a new transfer occurs, the contents of the dictionary V
Updated by. Although the variable dictionary V changes adaptively and sequentially as necessary, the contents of the variable dictionary V do not actually change.

The discrimination circuit 4 includes, for example, a central processing unit and a microprogram memory, and a similar circuit is shown in FIG. 16 of the above paper. The discrimination circuit 4 includes a dictionary section output 11 which is a combination of the outputs of the storage circuit 2 and the temporary storage circuit 3.
and character pattern information to be determined 61 from the outside are input. In this discrimination circuit, input character pattern information 6
1 and the fixed dictionary F or the variable dictionary V are compared and discriminated, and a discrimination result 41 is output. This determination result 41 is output to the outside and is also guided to one input of the control circuit 5.

In addition to the determination result 41, the corrected category information 62 is manually input to the control circuit 5, and its output 51 is led to the input of the -next memory circuit 30.

When correct correction category information 62 regarding illegibility or misreading is input from the keyboard, for example, in response to the determination result 41 outputted to the outside, the control circuit 5 stores the temporary storage circuit based on the determination result 41 and the information 62. The contents of the variable dictionary V of No. 3 are controlled to be changed.

Note that the configuration of the present invention is not limited to that of the above-described embodiment; for example, in the configuration shown in FIG. 19 of Japanese Patent Application No. 51-115084, the discrimination procedure is configured as a dictionary part, and the rest is configured as a discrimination circuit. It is also possible.

Next, the operation of the discrimination device of the present invention will be explained with reference to FIG.

FIG. 2 is a table showing an example of the dictionary adaptation process.

In the figure, 41 and A2 are representative examples of the fixed dictionary F, and A3 is a representative example of the variable dictionary V.

The category of pattern ``Ge'' of 41 is ``WAJ'', and the category of pattern 17 of 42 is ``7''. On the other hand, /I
For 63 patterns 1゛7'', there are multiple categories ``
7" is given. Here, the category of the n-th pattern Pn is expressed as C(i, D) as shown in the following equation.

Pm=C(i, D........ (1) However, i=
When j, i is Pm=C(wa, 7) in the second saving link, and this is the value stored in the variable dictionary V in FIG. In the figure, at the beginning when V is transferred to V, -th storage circuit 3
Also in the variation dictionary ■ in Figure 2, the pattern A3 is
The initial value is Pa-C (wa, 7).

Therefore, in FIG. 1, the character pattern information to be determined is
If it is for "7"', the discrimination circuit 4 selects P
Matched with a = C (wa, 7), no unique result was obtained,
It becomes unreadable and the determination result 41 is output.

For this non-reading, the correction category information 62 is C (wa).
When inputted, as in temporary fixing example 1 in FIG. 2, an adaptive value is obtained from the variable word 111: v Ic b, which temporarily becomes Pg-C (wa) and is fixed. After that, a new V
Hold this until it is moved from ■. On the other hand, when the corrected category information 62 is input as C(7), an adaptive value is obtained in the variable dictionary V, as shown in temporary fixation example 2 in FIG. , immobilized.

In this way, the variation dictionary is adapted for the entire discriminator. As described above, when a form or a new group of forms with a different filler arrives, V is updated with V, the initial value is given again, and a new adaptation operation begins.

In addition, in the present invention, it is extremely easy to increase the number of categories in the variable dictionary to more than two, which is practical. Furthermore, it is clear from the structure that there is no particular control over the target character type according to the present invention.

If it can be expressed as in equation (1), it can be handled. Furthermore, instead of manually adapting the modified category one time at a time, it is also possible to quantify the adaptability in correspondence with the degree of deformation of the pattern, and to modify it to include probability calculation to improve the adaptability.

As explained above, the present invention provides a fixed dictionary in which a unique fixed category is assigned to each pattern, a variable dictionary with variable categories, a temporary memory circuit and a metal fitting for temporarily storing the contents of the variable dictionary, and a variable dictionary. A plurality of possible categories are assigned to each pattern of IJ-4-, and when character discrimination is performed, the correct category variable is indicated to the temporary memory circuit using corrective category information from the outside, and all of this is temporarily solidified. By configuring to perform the discrimination, it is possible to simplify the correction work when each character is illegible by completing it once for each character, and to reduce the number of key operations. As a result, according to the present invention, characters with ``<se'' can be identified without having to be corrected every time the same similar pattern occurs to prevent unreadability due to confusion, thereby improving processing speed and operation. It is possible to improve sexual performance.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the character discrimination device of the present invention, and FIG. 2 is a table showing an example of the adaptation process of the dictionary section in the above embodiment. 1... Dictionary section 2... Storage circuit 3... Temporary storage circuit 4... Discrimination circuit 5... Control circuit Applicant: NEC Corporation Figure 1 Figure 2

Claims (1)

[Scope of Claims] A device for determining character categories using an optical character reading device, etc., which stores a fixed dictionary that assigns only one fixed category and a variable dictionary that assigns a plurality of categories. a temporary storage circuit that reads and stores a variable dictionary from the storage circuit; a discrimination circuit that compares discrimination information obtained from an input character pattern with the fixed dictionary and the variable dictionary to discriminate character categories; and a discrimination circuit. a control circuit for controlling the variable dictionary to be temporarily fixed by rewriting the contents of the temporary memory circuit in response to the output of the above to adapt the contents of the temporary memory circuit to changes in the entry person using correction category information provided from the outside as necessary; A character discrimination device characterized by comprising: