JPS62200483A - Character reader - Google Patents

Abstract

PURPOSE:To execute the identifying processing of a habitual character peculiar to a slip writer without rejection and to improve a character reading ratio by registering the pattern of the habitual character classified by respective persons beforehand besides a dictionary memory to store a usual character pattern. CONSTITUTION:In the first identifying part 16, the characteristic of the regularized character pattern and a standard character pattern stored in a dictionary memory 17 are matched, and as the result, the character, in which a matching degree is maximum, is outputted as an answer. When at the identifying step by the first identifying part 16, the matching degree of the character pattern of the first candidate is a certain threshold or below and the difference in the matching degree of the first candidate and the second candidate is a certain threshold or below, in the second identifying part 18, the character pattern classified by a person registered and stored in a pattern registering part 19 classified by the person is referred, the matching of the rejected character pattern and the pattern of the habitual character corresponding to the first candidate character when the character is discriminated to be rejection by the first identifying part 16 is obtained, and when the matching degree of a certain threshold or above obtained, the character is outputted as an answer.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Application Field) The present invention optically provides a character ii! l! The present invention relates to a character recognition device that performs character recognition.

(Prior Art) A conventional character g* device photoelectrically converts characters recorded on a form, outputs the character pattern, and stores the output character pattern in a dictionary memory that stores standard patterns in advance. The text is compared with the content of the text to identify characters. At this time, if the above character pattern matches the character pattern stored in the dictionary, the determination result is output as the answer, but if it does not match, the result is output as the answer (
(refuse to read) and became unidentifiable.

However, when examining the characters that are rejected as described above, there are many characters that are unique to the person filling out the form, that is, quirky characters. Therefore, in order to improve the reading rate, we are improving the dictionary memory that stores standard character patterns, but in order to read these quirky characters, it is necessary to reduce the recognition rate of other character types. There are various problems such as the need for large capacity memory.

(Problem to be Solved by the Invention) When performing character discrimination as described above, characters unique to the person filling out a form, i.e., quirky characters, cannot be distinguished by the dictionary memory that stores standard patterns in advance, and are rejected. There were many cases.

The present invention has been made in view of the above-mentioned points, and it is an object of the present invention to provide a character reading device that can perform identification judgment without rejecting even the idiosyncratic characters peculiar to a person who fills out a form, and improves the reading rate. It is something to do.

[Structure of the invention] (Means for solving the problems) That is, the character reading device according to the present invention includes a photoelectric conversion means that photoelectrically converts characters recorded on a form and outputs a character pattern; a first identification means for identifying a character pattern based on the standard pattern; and at least a character pattern for identification processing corresponding to a character pattern rejected by the first identification means. The apparatus includes a specific pattern registration means for registering a specific pattern, and a second identification means for performing identification based on the specific pattern registration means.

(Operation) That is, in the character reading device configured as described above, an input character pattern is first identified by the first identifying means. Here, the input character pattern is matched by referring to the standard character patterns stored in the dictionary memory, and if the first candidate character has a matching degree equal to or higher than a certain threshold, that character is selected as the answer. and if it is different, the character is rejected. In this case, the rejected character is identified by the second identification means. In this second identification means, the pattern of habitual characters of the person filling out the form is registered in advance, and the rejected character pattern is matched by referring to the contents of the registration section. Characters will now be output as answers.

In this way, even when a quirk character peculiar to a person filling out a form is input, the character is not rejected and the character is identified, thereby improving the character reading rate.

(Embodiment) Hereinafter, an embodiment of a character reading device according to the present invention will be described with reference to the drawings. FIG. 1 is a schematic diagram of a character reading device according to an embodiment of the present invention. That is, this character reading device includes a quantization section 12, a verification section 13, a preprocessing normalization section 14, and a feature extraction section 1, which are each controlled by a control section 11.
5, a first identification section 16, a dictionary memory 17, a second identification section 18, an individual pattern registration section 19, and a floppy disk device 20.

The quantization unit 12 has a function of photoelectrically converting a character pattern written on a form (not shown) into black β and white 21[ signals. The cutting section 13 has a function of cutting out one character based on character pitch information given in advance and storing the character pattern. The preprocessing normalization unit 14 performs preprocessing such as noise removal on the photoelectrically converted character pattern, and determines the height, width, etc. of the preprocessed character pattern.
It has the function of keeping the slope etc. constant. Feature extraction unit 15
has the ability to extract features useful for character identification. The first identification unit 16 has a function of comparing the normalized patterns sent from the feature extraction unit 15 with reference to the memory contents of a dictionary memory 17, which will be described later, and outputting characters with a high degree of similarity as an answer. have.

The dictionary memory 17 stores standard patterns for character recognition. The second identification unit 18 refers to the registered contents of the individual pattern registration unit 19, which will be described later, and discriminates the character pattern rejected by the first identification unit 16, and outputs the discrimination result as an answer. It has a function. The individual pattern registration section 19 has a function of registering character patterns for each person filling out a form, and the quirky character patterns of each person filling out a form are registered using a registration sheet, a floppy disk, etc., which will be described later. The floppy disk device 20 has a function of storing and holding character patterns registered in the individual pattern registration section 19.

The operation of the character reading device configured as described above will be explained below. That is, a character pattern written on a form (not shown) is photoelectrically converted and binarized by the quantization unit 12, and then stored in a memory such as a line buffer. Then, the character pattern stored in this memory is cut out character by character by the cutter 13 based on the character pitch information, and sent to the preprocessing normalization part 14. The preprocessing normalization unit 14 performs preprocessing such as noise removal and normalization to make the height, width, slope, etc. constant, and sends the normalized character pattern to the feature extraction unit 15. Feature extraction unit 15
The normalized character pattern input to the normalized character pattern is sent to the first identification unit 16 after extracting characteristic parts useful for character identification. In this first identification unit 16, the characteristics of the normalized character pattern are matched with the standard character pattern stored in the dictionary memory 17, and as a result, the character with the highest degree of matching is output as the answer. do.

In this case, in the identification step by the first identification unit 16, the matching degree of the first candidate character pattern is equal to or less than a certain value, or the difference in the matching degree between the first candidate and the second fa complement is less than or equal to a certain threshold value. In some cases, the determination of the character pattern is entrusted to the second identification section 18.

The second identification section 18 refers to the individual character patterns registered and stored in the individual pattern registration section 19 to determine whether the character pattern rejected by the first identification section 16 is identified. . That is, the rejected character pattern is matched with the habitual character pattern corresponding to the first candidate character determined to be rejected by the first identification unit 16, and the degree of matching is greater than a certain threshold. If so, output that character as the answer.

Here, the registration method of the individual pattern registration section 19 will be explained. For example, necessary numbers, letters, symbols, etc. are entered for each individual on a registration sheet as shown in FIG. 2, and correspondence between standard characters and each individual's idiosyncratic characters is established.

By reading this sheet in advance with a character reading device, the patterns of quirky characters are registered in the individual pattern registration section 19 for each category.

The idiosyncratic patterns registered in the individual pattern registration section 19 are stored and held in a floppy disk or the like set in the floppy disk device M20.

In addition, a floppy disk is prepared for each individual, and each individual's quirk pattern is stored in the personal floppy disk, and each time a form is read, the floppy disk for the person filling out the form is set in the floppy disk device 20. Then, the data on the floppy disk may be written into the individual pattern registration section 19.

FIG. 3 is a flowchart 1 for explaining the above processing operation.
− is. That is, first, in step 101, a condition determination of the registration mode is performed, and when the registration mode is set, a registration sheet input determination is performed in step 102.

Here, if you do not use a registration sheet, step 1
Using the personal floppy disk registered in 03,
The idiosyncratic pattern stored on the personal floppy disk is registered in the individual pattern registration section 19.

In addition, if a registration sheet is used in step 102 above, in step 104, necessary numbers, letters, etc. are written on the registration sheet, read by a character reading device, and then the pattern of the idiosyncrasies is converted into an individual pattern. Register in the registration section 19.

After registering and setting the idiosyncratic pattern in the individual pattern registration section 19 in this way, the reading mode from step 105 is performed. First, in step 105, a first identification is performed. Here, the contents stored in the dictionary memory 17 are referred to and matching with the input character pattern is performed. Next, in step 106, a direct determination is made based on the degree of matching. That is, as a result of the identification in step 105, if the matching degree of the first candidate character is equal to or higher than a certain threshold value, in step 107, the character is used as the answer and the character code corresponding to the character is output. ·on the other hand,
In step 105, if the degree of matching of the first candidate character is equal to or less than a certain value, or if the degree of matching of the first candidate and the second candidate is less than a certain value, step 10
A reject command is output in step 6, and a second identification is performed in the next step 108. That is, the rejected character pattern is matched with the idiosyncratic character pattern registered in the individual pattern registration section 19. Then, in step 109, if the degree of matching is equal to or higher than a certain threshold, the character is used as the answer and the character code corresponding to the character is output in step 107. If the degree of matching is below a certain threshold, a reject code is output in step 100.

Note that when a reject code is output in step 110, character confirmation can be performed by displaying the reject character on a display. In addition, the above-mentioned reject characters can be added to step 103 or step 104.
By registering the pattern in the individual pattern registration section 19 using the means described above, it is possible to prevent glue jetting from the next time.

[Effects of the Invention] As described above, according to the present invention, in addition to the dictionary memory storing normal character patterns, patterns of quirky characters for each individual are registered in advance, so that unique character patterns peculiar to the person filling out the form can be recorded. It becomes possible to perform identification processing without rejecting idiosyncratic characters, and the character reading rate improves.

[Brief explanation of drawings]

FIG. 1 is a schematic configuration diagram of a character reading device according to the present invention, FIG. 2 is a diagram showing an example of a registration sheet, and FIG. 3 is for explaining the flow of operation of the character reading device shown in FIG. 1 above. This is a flowchart. DESCRIPTION OF SYMBOLS 11... Control part, 12... Quantization part, 13... Testing part, 14... Preprocessing normalization part, 15... Feature extraction part, 16... First identification part , 17... dictionary memory,
18... Second identification section, 19... Individual pattern registration section, 20... Floppy disk device. Applicant's agent Patent attorney Takehiko Suzue Figure 2 Figure 3

Claims (1)

[Scope of Claims] A photoelectric conversion means for photoelectrically converting characters recorded on a form to output a character pattern, a dictionary memory storing standard patterns in advance, and a method based on the standard patterns stored in the dictionary memory. a first identification means for identifying a character pattern obtained from the photoelectric conversion means; a specific pattern registration means for registering at least a specific pattern for identification processing corresponding to a character pattern rejected by the first identification means; , and second identification means for identifying the character pattern rejected by the first identification means based on the specific pattern registration means.