JPS6331832B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character reading device that recognizes and reads characters written on a form or the like.

A character reading device scans a form, etc., generates an electrical signal corresponding to the character pattern written on the form, etc., and recognizes each character based on the characteristics of the character pattern of each character to be recognized. It is equipped with a dictionary, and reads characters by referring to the recognition dictionary based on the pattern of electrical signals generated by the above-described scanning.

By the way, conventional character reading devices are mainly focused on reading forms filled out by an unspecified number of people, and are therefore considered to represent the average deformation of character patterns filled in by many people. Since it is configured to perform character recognition using only one type of recognition dictionary, some users may write unique and highly deformed characters, and such characters cannot be recognized by conventional character reading. The device has a high probability of refusing recognition or erroneous recognition, which has been a drawback of conventional devices. In addition, especially for characters with similar shapes, such as "ku" and "wa", "so" and "n", the correspondence between the character type and the character pattern often involves the subjective opinion of the person filling them out. It is essentially difficult to objectively recognize the characters in . Therefore, in the past, a standard character pattern was determined, and all fillers were required to fill in using this standard character pattern as a model, which was a major drawback when putting character reading devices into practical use. It was becoming.

This invention was made in order to eliminate the above-mentioned drawbacks of conventional devices, and is capable of reflecting individual character pattern information for each person writing in a recognition dictionary. Explain an example.

FIG. 1 is a block diagram showing an embodiment of the present invention, in which 1 is a form, 2 is a scanning device, 3 is a recognition device, and 4 is a block diagram showing an embodiment of the present invention.
is a first storage device, 5 is a second storage device, 6 is a control device, and 7 is an external storage device. The form 1 has characters written on it (in this specification, characters include not only ordinary characters but also symbols, codes, and other figures), and the scanning device 2 reads the characters written on the form 1. A recognition dictionary for recognizing characters from the character pattern is stored in the first storage device 4, and a recognition dictionary for recognizing characters from the character pattern is stored in the second storage device 5. The external storage device 7 is made up of, for example, a magnetic disk, and stores a table unique to the person who filled in the data, in which pattern classification information indicating information on a character pattern unique to the person who filled in the data corresponds to the character information.
A recognition dictionary to be stored in the storage device 4 and a table unique to each filler to be stored in the second storage device 5 are stored for all the fillers, and the recognition device 3 stores the recognition dictionary to be stored in the first storage device 4 based on the output of the scanning device 2. Character recognition is performed with reference to the second storage device 5, and the control device 6 controls the operations of the scanning device 2, the recognition device 3, and the external storage device 7. Further, among the signal lines shown in FIG. 1, hollow lines indicate the flow of character data, solid lines indicate transfer and reference of a recognition dictionary and a table unique to the filler, and broken lines indicate control signals.

Figure 2 is a diagram showing an example of characters written on form 1, where 8 is an identification character string that is represented by the person's unique name or number, and 9 is a character string that can be read by this character reader and read by a computer, etc. This is input character information to be input.

The operation of this device will be explained below with reference to FIGS. 1 and 2. The scanning device 2 scans the form 1 under the control of the control device 6, and first sends an electric signal corresponding to the character pattern of the identification character string 8 to the recognition device 3.
Since the recognition device 3 is informed by the control device 6 that the signal is obtained from the identification character string 8, the recognition device 3 performs character recognition for the identification character string 8 in the same manner as the conventional method. That is, recognition is performed by referring only to the recognition dictionary stored in the first storage device 4, and the result is sent to the control device. By the way, the identification character string 8 is generally composed of a string of numbers when handwritten, and is often machine printed, so
Similar to the conventional method, the recognition rate can be sufficiently high even if recognition is performed by referring only to the recognition dictionary.

When the reading of the identification character string 8 is completed and the filler is identified, the control device 6 reads a filler-specific table for the identified filler from the external storage device 7, and reads this table via the recognition device 3. Transfer to the second storage device 5. Generally, reading from the external storage device 7 is not fast, but since the capacity of the filler-specific table for the identified filler is several kilobytes or less, the scanning device 2
The table specific to the filler can be stored in the second storage device 5 before scanning of the next input character information 9 is started.

FIG. 3 is a diagram illustrating a part of a filler-specific table for each filler stored in the external storage device 7, and FIG. Figure b shows part of the table for filler B. Each table includes a character type, a standard character pattern for that character, and symbols 〇 and × indicating whether or not the person writing the character generates the character pattern. Since only the digital codes corresponding to the character types, standard character patterns, and symbols ○ and x are stored, the capacity of the second storage device 5 may be small. To explain in more detail using the character type (X) as an example, the first storage device 4 stores four types of dots shown as standard character patterns in Figure 3, corresponding to the character code representing the character type (X). The pattern is memorized. These four types of dot patterns are numbered from left to right and temporarily numbered 1st, 2nd,
These are called the third and fourth patterns. In the second storage device 5 corresponding to filler A, in the character code column representing (S), the first pattern is invalid (x), the second pattern is invalid (x), the third pattern is valid (〇), and the fourth pattern is invalid. A flag indicating pattern valid (〇) is set (Fig. 3a), and in the second storage device 5 corresponding to person B, the first pattern is valid (〇) in the character code column representing (SO). Flags such as 2nd pattern valid (〇), 3rd pattern invalid (x), and 4th pattern invalid (x) are set (Fig. 3b), and these dot patterns themselves are stored in the second storage device 5. It hasn't been done. When the filler is identified, only the dot patterns for which the valid flag (◯) is set in the second memory device 5 corresponding to the filler are selected from the first memory device 4 and used for character reading.

When the scanning device 2 scans the input character information 9 and sends the corresponding electrical signal to the recognition device 3, the recognition device 3 refers to the first storage device 4 and the second storage device 5 to perform character recognition. conduct. In other words, when character recognition is performed by referring to the first storage device 4 based on the output of the scanning device 2, among all the standard character patterns stored in the first storage device 4, 〇 marks are selected in the second storage device 5. Character recognition is performed by referring only to standard character patterns with . For example, person A
It is known that there is no hook in the "S" written by A, and when it is identified that the person who wrote it is A, a table, a part of which is shown in FIG. 3a, is stored in the second storage device 5. Therefore, for the input character information 9, it is only necessary to use the pattern for "SO" marked with a circle in FIG. good.

FIG. 4 is a diagram showing an example of a form to show the effect of this invention, in which FIG. 4a is a form filled out by the above-mentioned filler A, and FIG. 8a and 8b are identification numbers for identifying fillers A and B, respectively, and 9a and 9b are respective input character information. Both are written as “Ason”. As is clear from FIG. 4, the input character information 9a "So" and 9b "N" have similar character patterns, and if they coexist in the same document, they cannot be identified and recognition must be rejected. I don't get it. However, according to the present invention, since it is known from the number 8a that the characters in the form shown in FIG. 4a were filled in by person A, the second memory shown in FIG. The device shows that Person A's "S" does not have a hook, and "N" does have a hook, making it easy to identify these characters.
Similarly, the form filled out by person B shown in FIG. 4b can also be read correctly. By identifying the person filling the form in this way, character patterns that were conventionally rejected can now be correctly recognized, and a high recognition rate can be ensured. In addition, for the above-mentioned table unique to the person filling in the table, a reading test using all the standard character patterns is performed, and for the standard character pattern that is correctly recognized, the standard character pattern that the person writing in question can generate (the first (corresponding to the pattern marked with a circle in Figure 3)), and for other character patterns, the character pattern is displayed on a cathode ray tube display, and registration is determined by the operator's judgment. It can be created with .

Although Fig. 2 shows examples of alphanumeric characters and kana characters, this invention is not limited to this, but can also be applied to reading kanji, which have more complex character shapes and a large number of standard character patterns for one type of character. You can also.

As described above, according to the present invention, the person who filled in a form, etc. is identified by reading the character string for identification, and when referring to the recognition dictionary for recognizing characters, the person who filled out the form is identified. information can be used,
This method has the advantage of maintaining a high recognition rate even when the character shape changes depending on the person writing it. Furthermore, the total number of fillers across all forms that one character recognition device must actually read is relatively small, and it is possible to create a filler-specific table stored in the external storage device relatively easily.

[Brief explanation of the drawing]

Figure 1 is a block diagram showing an embodiment of this invention, Figure 2 is a diagram showing an example of characters written on a form, and Figure 3 is a part of a table unique to the filler used in this invention. FIG. 4 is a diagram showing an example of a form for showing the effects of the present invention. 1... Form, 2... Scanning device, 3... Recognition device, 4...
First storage device, 5... Second storage device, 6... Control device, 7... External storage device, 8... Character string for identification.

Claims (1)

[Claims] 1 A character reading device used when the written characters include a character string for identifying the person who wrote the characters, scans the surface on which the characters are written and generates an electric signal corresponding to the pattern of the written characters. a scanning device that generates a signal; a first storage device that stores a recognition dictionary for recognizing a corresponding character according to a character pattern; and a standard character that is different for each character in the recognition dictionary. Pattern classification information that is added to distinguish between patterns, and pattern classification information that indicates a standard character pattern predetermined for the person who wrote the character when the person who wrote the character is identified, are matched with the character information unique to the person who wrote the character. a second storage device that stores a table for each filler; an external storage device that stores a table specific to the filler for each filler information for all predetermined fillers; Input the output and recognize the character string that identifies the filler by referring to the first storage device based on the signal pattern corresponding to the character string that identifies the filler, and the filler identified in this way. The table specific to the filler is transferred from the external storage device to the second storage device, and the first storage device and the second storage device are referred to to write characters according to the output signal pattern of the scanning device. a recognition device for recognizing; the scanning device;
A character reading device comprising: the recognition device; and a control device that controls the operation of the external storage device.