JPH07282198A - Typed character recognizing device - Google Patents

Abstract

PURPOSE:To select an item to be given importance at the time of executing recognition processing actually, that is, one of the improvement of a recognition rate and the reduction of recognition time so that recognition can be executed by a recognizing method to meet the said item. CONSTITUTION:A recognition processing selecting means 17 is provided in a typed character recognizing device. In the recognition processing selecting means 17, the improvement of the recognition rate or the reduction of the recognition time is selected as the item to be given importance. As for a method for improving the recognition rate, the character registration and deletion to/ from a recognition dictionary 18, and the style of the character of collation result, are managed, and the said style is collated preferentially in the subsequent recognition. As a method for reducing the recognition time, either the recognition by a simplified recognition dictionary 16 in which only the necessary characters are stored or a method to recognize the character by collating limitedly only the necessary characters in the recognition dictionary 18 can be selected. In a recognizing (retrieving) means 13, the recognition is executed by the selected method.

Description

Detailed Description of the Invention

[0001]

INDUSTRIAL APPLICABILITY The present invention can arbitrarily specify whether to improve the recognition rate even if the time is slightly longer, or to shorten the recognition time even if the recognition rate is slightly worse. The present invention relates to a type character recognition device.

[0002]

2. Description of the Related Art Conventionally, when a character is recognized by a type character recognition device, a recognition dictionary in which character patterns of all types of typefaces to be recognized are registered is stored in the recognition device.
The recognition is performed by comparing and collating this recognition dictionary with the input character pattern. However, there are Mincho typefaces, Gothic typefaces, other typefaces, etc. even in the case of the typefaces of Japanese characters, and among these Gothic typefaces, there are typefaces with different fonts. Had to store the typeface, and the memory capacity of the recognition dictionary might be huge. Therefore, as a method that significantly suppresses the number of standard patterns to be compared, speeds up recognition processing, and can suppress recognition errors,
For example, a type character recognition device described in Japanese Patent Laid-Open No. 4-245390 has been proposed. In the above device, the printed characters printed on the form are digitized into a binary pattern by the photoelectric conversion unit and then stored in the quantization pattern memory. At the time of recognition, each character is cut out by the character cutout unit to detect the height and width of the character. On the other hand, the standard value memory stores the standard values of height and width for each font and each character format, and the standard pattern memory stores standard patterns grouped by the classification of these standard values. Keep it. The character height / width selection part selects which standard pattern of the input print character height is, and the recognition part performs comparison and matching with the recognition dictionary only for the selected print character pattern. .

[0003]

In the above prior art, the character patterns of all types of typefaces to be recognized are registered in the recognition dictionary, and the recognition dictionary is collated at the time of recognition. Therefore, there has been a problem that the recognition time becomes longer and the recognition error also increases in response to the increase in the typefaces to be recognized. There is also a method of suppressing the increase in recognition errors by limiting the character patterns to be collated among the character patterns registered in the recognition dictionary based on the characteristics of the characters, etc. Character patterns of many typefaces are stored in addition to the typeface. For this reason, even if limited, there are considerable matching patterns, which is an improvement compared to the method of matching with all patterns, but the recognition time is extended and recognition is performed in response to the increase in the typefaces to be recognized. There was a similar problem of increasing errors. In order to solve the above problems, an object of the present invention is to provide a type character recognizing device capable of recognizing by a selected method by selecting one to be emphasized in recognition rate improvement, recognition time reduction, etc. when recognizing. To do.

[0004]

In order to achieve the above object, the type character recognizing device of the present invention is a process selecting means for designating whether to improve the recognition rate or reduce the recognition time at the time of recognition. (21 in FIG. 2) and a form on which the typeface characters to be recognized are printed are input from the input device (11) to register the character pattern in the recognition dictionary (18), and the character pattern of the typeface that is no longer needed Recognition dictionary registration / deletion means (23) that arbitrarily deletes the recognition dictionary (18), and a collation result that registers the typeface of the collation result that matches the recognition dictionary (18) at the time of form recognition in order to improve the recognition rate. In the dictionary registration means (24) and the recognition thereafter,
A recognition dictionary collation management unit (25) that preferentially collates the typefaces registered by the collation result typeface registration unit (24), and in order to reduce the recognition time, the registered characters are limited and only the typefaces to be recognized are recognized. A simple dictionary creating means (2) for arbitrarily creating the stored simple recognition dictionary (16) from the recognition dictionary (18)
2) and a recognition form typeface managing means (26) for limiting the typeface to be collated in the recognition dictionary (18) according to the form to be recognized in order to shorten the recognition time.

[0005]

In the present invention, the item that is most important when recognizing is selected from the one of the improvement of the recognition rate and the reduction of the recognition time, and the recognition process is executed by the selected recognition method.
First, if you select as an item that emphasizes the recognition rate improvement,
Register and manage the typeface name that is the recognition result. At the time of recognition, the recognition target character and the recognition dictionary are compared and collated, and the collation result is output. In this case, the typeface name of the character that coincides as a result of collation is registered. As a result, when the next character to be recognized and the recognition dictionary are matched, the recognition error can be suppressed and the recognition rate can be improved by preferentially matching the registered typefaces. it can. In addition to the recognition process, new recognition target characters are registered in the recognition dictionary and unnecessary characters are registered and deleted to maintain the recognition dictionary, thereby improving the recognition rate. . The conventional recognition dictionary is fixed at the time of creation, and the contents cannot be arbitrarily deleted by the user or newly registered. In the present invention, the user can arbitrarily delete or register. Next, when it is selected as an item that emphasizes reduction of recognition time, the recognition dictionary is operated. Since many characters of typefaces are stored in the recognition dictionary, a simple recognition dictionary that stores only the typeface characters necessary for recognizing individual forms from this recognition dictionary is created. It is possible to shorten the recognition time by collating at the time of recognition with this simple recognition dictionary. In addition to the creation of the simple recognition dictionary, a method of limiting the typeface to be collated by the form recognized in the recognition dictionary is also used. In this case, the recognition time is shortened by using the method of registering the typeface to be recognized corresponding to the form to be recognized and collating only the typeface to be recognized in the recognition dictionary. As a result, it is possible to perform recognition by a method suitable for the point of importance at the time of recognition.

[0006]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a schematic configuration diagram of a type character recognition apparatus showing an embodiment of the present invention. In FIG. 1, 11 is an input means, 12 is a memory means, 13 is a recognition means, 14 is a code conversion means, 15 is an output means, and 18 is a recognition dictionary.
These are means provided conventionally, and in the present invention, a new recognition process selecting means 17 and a simple recognition dictionary 16 are newly added.
And. The recognition dictionary 18 and the simple recognition dictionary 16 are
It is stored in the recognition (search) means 13. Input means 1
The image information input from 1 is temporarily stored in the memory means 12, then read out and input to the recognition (search) means 13 to recognize the image information. The recognition process selection means 17 provided in the type recognition device can operate the simple recognition dictionary 16 or the recognition dictionary 18 to be collated at this time. The recognition processing selection means 17 selects improvement of the recognition rate or reduction of the recognition time as an item to be emphasized at the time of recognition. When the recognition rate improvement is selected, a processing method for improving the recognition rate in actual recognition is selected.

Characters are registered in the recognition dictionary by using a recognition dictionary registration form (41 in FIG. 4 which will be described later) in a predetermined format.
A character to be newly recognized is printed in (see) and is input from the input unit 11. The input image information is temporarily stored in the memory means 12. The recognition processing selection means 17 receives this image information from the memory means 12, and creates a standard character pattern for registration based on the image information. Then, this standard character pattern is registered in the recognition dictionary 18.
When deleting a standard character pattern that has already been registered, the recognition processing selection means 17 displays the contents of the recognition dictionary 18 and deletes the typeface name (for all characters of the specified typeface) or typeface name and character (designation). The standard character pattern is deleted by specifying the individual characters of the typeface. In this way, the character pattern is registered and deleted in the recognition dictionary 18. When the recognition processing method is selected, the recognition processing selection means 17 manages the character attributes and typefaces as a result of collation. Character attributes are types such as numbers, letters, and kanji. The recognition (search) means 13 stores the character patterns stored in the recognition dictionary 18 and the memory means 1.
The image information read from 2 is collated, and the recognition result is output to the code conversion means 14. For the result that can be recognized, the attribute of each character, the recognition dictionary 1
The name of the typeface that matches with No. 8 is output to the recognition processing selection unit 17. The recognition processing selection means 17 manages the attribute of the character and the typeface name, and if there is the same attribute of the character to be recognized thereafter, the typename that has been matched in the previous matching is recognized (searched). It outputs to 13. Recognition (search)
The means 13 preferentially collates the characters of the received typeface name. By this recognition processing, the recognition rate is improved.

When the recognition process selecting means 17 selects to shorten the recognition time, the recognition dictionary 18 is operated. As a method for shortening the recognition time, a method of creating the simple recognition dictionary 16 from the recognition dictionary 18 and a method of recognizing the characters stored in the recognition dictionary 18 by limiting the typeface to be collated by the form Can be selected. In the latter case, the simple recognition dictionary 16 is not created, and the typeface used for collation is specified in the recognition dictionary 18 so that the other typefaces cannot be collated. When the method of creating the simple recognition dictionary 16 is selected, the recognition processing selecting unit 17 selects only the characters of the typeface to be recognized by the form to be recognized from the recognition dictionary 18, creates the simple recognition dictionary 16, and recognizes ( A simple recognition dictionary 16 is designated in the search means 13 as a dictionary to be matched at the time of recognition. The recognition (search) means 13 sequentially reads out characters from the designated simple recognition dictionary 16 and performs collation. When the method of limiting the characters to be collated is selected, the recognition processing selection unit 17 outputs the typeface to be recognized by the form to the recognition (search) unit 13 before recognition. The recognition (search) means 13 reads only the characters of the received typeface from the recognition dictionary 18 and collates them. As described above, as a result of the recognition processing selecting means 17 performing the recognition processing by a method suitable for the item to be emphasized at the time of recognition, the code converting means 14 converts the code into a code and the output means 15 outputs the code.

FIG. 2 is a detailed block diagram of the recognition processing selection means in FIG. The recognition processing selection means 17 is shown in FIG.
As shown in, select whether to improve the recognition rate or shorten the recognition time,
In the case of shortening the time, a process selecting unit 21 that selects simple dictionary creation or collation typeface limiting process, and a simple recognition dictionary creating unit 22 that is started when the simple recognition dictionary creating in the case of time shortening is selected. , The recognition form typeface management unit 26 that is activated when the collation typeface limiting process for time reduction is selected
And registration in the recognition dictionary when the recognition rate improvement is selected,
A recognition dictionary registration / deletion unit 23 that performs deletion, a verification result typeface registration unit 24 that registers a typeface of the verification result when the recognition rate improvement is selected, and a recognition dictionary collation management unit that collates and manages the registered typeface. 25 and. The operation of the recognition processing selection means 17 will be described by taking four examples. As a method for improving the recognition rate, when a new recognition target character is registered as the maintenance of the recognition dictionary 18 and an unnecessary character is deleted (registration / deletion unit 24), at the time of matching with the recognition dictionary 18, A case where the matched typeface is managed and recognized (matching management unit 25) will be described. Also,
As a method for reducing the recognition time, when the simple recognition dictionary 16 storing only the characters necessary for the recognition is created and the recognition is performed (the simple recognition dictionary creating unit 22), the characters stored in the recognition dictionary 18 are used. Among the above, when recognizing by limiting the typeface to be collated by the form to be recognized (typeface management unit 2
6) will be described.

First, in order to improve the recognition rate, the recognition dictionary 18
The registration of new recognition target characters and the deletion of unnecessary characters will be described as maintenance of. Figure 3
FIG. 4 is a format configuration diagram of the recognition dictionary in FIG. 1, and FIG. 4 is a format diagram showing an example of a recognition dictionary registration form for registering in the recognition dictionary. In Gothic,
It is assumed that there is a 01 typeface as a font type. Now, in order to newly recognize the character of the Gothic type 01 as the recognition target, when the character pattern is registered in the recognition dictionary 18, as shown in FIG.
The characters of the Gothic font 01 are printed on the recognition dictionary registration form 41 having a predetermined format as shown in FIG. The typed character print area 43 in the recognition dictionary registration form 41 indicates that the image inside is the area used as the source of the standard character pattern registered in the recognition dictionary 18, and the upper part of this typed character print area 43. The corresponding character 42 in is a character corresponding to the character pattern printed in the lower type character printing area 43. The recognition dictionary registration form 41 in which the Gothic font 01 to be newly registered as a character pattern is printed in the print character printing area 43 is input from the input unit 11, and the input image information is stored in the memory unit 12. In the recognition process selection means 17, first, the process selection unit 21 shown in FIG. 2 selects registration of the recognition dictionary 18, and receives the newly registered typeface name “Gothic 01”.

Next, the image information of the recognition dictionary registration form 41 input from the memory means 12 is added to the recognition dictionary registration / deletion unit 2.
When 3 receives, the character pattern is extracted from the image in the typed character print area 43 in which the character pattern exists and is associated with the corresponding character 43. This character pattern is converted according to the dot density of the standard character pattern registered in the recognition dictionary 18 to create a standard character pattern. The typeface name “Gothic 01” is the recognition dictionary 18 shown in FIG.
The standard character pattern that is output to and created in the typeface name item 31 in the standard character pattern storage area 32 is stored in the area corresponding to the corresponding character 35 in the standard character pattern storage area 34 of the Gothic 01 (shaded line). Area). As a result, a new Gothic 01 character pattern is registered in the recognition dictionary 18. At this time, if'Gothic 01 'already exists in the typeface name item 31, either change the typeface name and register it, or register the standard character pattern as the same typeface name by overwriting, and select either one. To do.

When a nonexistent typeface name is redesignated in the recognition dictionary 18, the redesignated typeface name is output to the typeface name item 31, and a standard pattern is registered in the same manner as the Gothic type 01 registration. In the case of overwriting registration, the typeface name is not output, and the standard character pattern is overwritten to be registered in the recognition dictionary 18. When deleting the character pattern of Gothic 01 in the recognition dictionary 18 in which the character pattern of Gothic 01 is already stored, the character selection of the recognition dictionary 18 is deleted by the process selection unit 21 of the recognition process selection means 17. To be selected. In the recognition dictionary registration / deletion unit 23, the typeface name item 3 in the recognition dictionary 18
1. Read the standard character pattern stored in the standard character pattern storage area 32. Displays the read typeface name and the corresponding standard character pattern and deletes the typeface name (for all characters corresponding to the specified typeface) or the typeface name and characters (for individual characters of the specified typeface) Upon receiving the designation, the contents of the recognition dictionary 18 are changed. When the deletion designation is the font name'Gothic 01 'and all the characters belonging to that font are deleted, the'Gothic 01' in the font name item 31 in the recognition dictionary 18 is deleted and invalidated, and the Gothic 01 The standard character pattern storage area 34 is cleared to be unused. Further, when the deletion designation is the font name “Gothic 01” and the character “A” and only the character “A” of the Gothic 01 font is deleted, the Gothic in the standard character pattern storage area 32 in the recognition dictionary 18 is deleted. The standard character pattern is deleted from the recognition dictionary 18 by erasing the area 33 corresponding to the standard character pattern of the body 01 and A to make it an unused area.

FIG. 5 is a format diagram of a collation typeface management table provided in the recognition dictionary collation management unit in FIG. Next, a method for managing and recognizing a matched typeface at the time of matching with the recognition dictionary 18 in order to improve the recognition rate will be described. In the recognition process selection means 17, first, the process selection unit 21 selects a recognition process for improving the recognition rate. Next, the collation result typeface registration unit 24 creates the collation typeface management table 51 shown in FIG. 5, deletes the attribute 52 and the typeface name 53, and sets it as an unused area. The recognition dictionary matching management unit 25 specifies the recognition dictionary 18 as a dictionary to be matched by the recognition (search) means 13. The recognition (retrieval) means 13 reads the recognized image from the memory means 12 and sequentially collates it with the characters stored in the recognition dictionary 18. Every time the result of collation is output, the attribute of the matched character and the typeface name are output to the collation result typeface registration unit 24. When the first matched character is the number “3” and the typeface “Gothic 01”, the collation result typeface registration unit 24 sets the received attribute “number” and typeface name “Gothic 01” at the time of collation. It is output to the attribute 52 and the typeface name 53 of the management table 51.

The recognition dictionary collation management unit 25 refers to the contents of the collation typeface management table 51. Next, if the attribute of the character to be recognized is a number, it exists in the collation typeface management table 51 '. Recognize Gothic 01 '(search)
Output to the means 13. The recognition (search) means 13 preferentially collates the received typeface'Gothic 01 '. The collation result is output to the collation result typeface registration unit 24 with the attribute and typeface name on a character-by-character basis, but if it is already present in the collation typeface management table 51, it is not output to the table. Further, the recognition dictionary collation management unit 25 refers to the collation typeface management table 51 for each character to be collated, but if the attribute 52 does not exist in the table, collation is preferentially performed to the recognition (search) means 13. The typeface is not output, and the recognition (search) means 13 uses the recognition dictionary 18
The characters stored in are sequentially read out and collated.

FIG. 6 is a format diagram of a form-specific typeface management table provided in the recognition form typeface management unit in FIG. Next, in order to shorten the recognition time, a method for creating and recognizing the simple recognition dictionary 16 in which only necessary characters to be recognized are stored will be described by taking the case of recognizing the form 01 type as an example. In the recognition processing selection means 17, first, the processing selection unit 21 selects a method of creating the simple recognition dictionary 16 among the recognition processing for reducing the recognition time. next,
The simple dictionary creating unit 22 refers to the form-specific typeface management table 61 shown in FIG. 6, and selects form 01 from the form type item 62.
To search. Next, the attribute 63 corresponding to the form 01 and the typeface name 64'numerical Gothic 01''English letter Gothic 02 'are read. The numbers of the Gothic font 01 and the English letters of the Gothic font 02, which are the typeface characters of the read attributes, are read from the recognition dictionary 18 and registered in the simple recognition dictionary 16.
In short, only the required attributes and typeface names are read from the collation typeface management table 51 and registered in the simple recognition dictionary 16. The simple recognition dictionary 16 read out and created is designated to the recognition (search) means 13 as a dictionary to be collated at the time of recognition. The recognition (search) means 13 sequentially reads out characters from the simple recognition dictionary 16 designated as a dictionary to be matched at the time of recognition and performs matching.

Next, in order to reduce the recognition time, the recognition dictionary 18
A method of limiting the typefaces to be collated by the form to be recognized among the characters stored in the form and recognizing the typeface will be described taking the case of recognizing the form 01 type as an example. In the recognition process selection means 17, first, the process selection unit 21 selects a method of limiting the collated typeface among the recognition processes for reducing the recognition time. Next, the recognition form typeface management unit 26 refers to the form-specific typeface management table 61 and retrieves the form 01 from the form type item 62. Next, the attribute 63 corresponding to the form 01
And typeface name 64 'number Gothic 01''Alphabet Gothic 02' is read. The read attributes and the typeface name, ie, the numbers of the Gothic font 01 and the letters of the Gothic font 02 are output to the recognition (search) means 13. Recognition (search) means 1
In No. 3, when reading from the recognition dictionary 18, only the number Gothic type 01 and the English letter Gothic type 02 are limited to be read, and collation is performed. That is, in the limited read process, the recognition form typeface management unit 26 recognizes (searches) only the typefaces selected and read from the type-specific typeface management table by the program without creating a new simple recognition dictionary.
By transferring to the means 13, the recognition (search) means 13
Recognizes the input pattern by collating it with the typeface pattern.

[0017]

As described above, according to the present invention,
In the recognition by the type character recognition device, it is possible to recognize by the recognition method by selecting it as the recognition method that emphasizes either the improvement of the recognition rate or the reduction of the recognition time. Therefore, when recognizing, the recognition process can be performed by a recognition method suitable for the actual situation of the user.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a type character recognition device showing an embodiment of the present invention.

FIG. 2 is a detailed block diagram of a recognition processing selection unit in FIG.

FIG. 3 is a format configuration diagram of a recognition dictionary in FIG.

4 is a recognition dictionary registration form format diagram for registering in the recognition dictionary of FIG. 1. FIG.

5 is a format diagram of a collation typeface management table created by a collation result typeface registration unit in FIG. 2;

FIG. 6 is a format diagram of a form-specific typeface management table created by a recognition form typeface management unit in FIG.

[Explanation of symbols]

11 ... Input means, 12 ... Memory means, 13 ... Recognition (search) means, 14 ... Code conversion means, 15 ... Output means, 1
6 ... Simple recognition dictionary, 17 ... Recognition process selecting means, 18 ... Recognition dictionary, 21 ... Process selecting section, 22 ... Simple dictionary creating section, 2
3 ... Recognition dictionary registration / deletion unit, 24 ... Collation result font registration unit, 25 ... Recognition dictionary integrated control unit, 26 ... Recognition form font management unit, 31 ... Font name item, 32 ... Standard character pattern storage area, 33 ... Gothic Body 01 A standard character pattern storage area, 34 ... Gothic 01
Standard character pattern storage area, 35 ... Corresponding character, 41 ... Recognition dictionary registration form, 42 ... Corresponding character, 43 ... Printed character printing area, 51 ... Collation type management table, 52 ... Attribute, 5
3 ... font name, 61 ... font management table by form, 62 ...
Form type item, 63 ... Attribute, 64 ... Font name.

Claims (1)

[Claims] 1. A process selecting means for designating whether to prioritize improvement of recognition rate or reduction of recognition time at the time of recognition, and by inputting a form on which a typeface character to be recognized is printed from an input device. In addition to registering character patterns in the recognition dictionary, the recognition dictionary registration / deletion means that deletes unnecessary character patterns of typefaces from the recognition dictionary, and in order to improve the recognition rate, matches the recognition dictionary at the time of form recognition. A collation result dictionary registration means for registering the typeface of the collation result, and a recognition dictionary collation management means for preferentially collating the typeface registered by the collation result typeface registration means in the subsequent recognition, and reduction of the recognition time. Therefore, the number of registered characters is limited, and a simple dictionary creating means for arbitrarily creating a simple recognition dictionary that stores only the typefaces to be recognized is designed to reduce the recognition time. By recognizing the form, printed characters recognition apparatus characterized by comprising a recognition document font management means for limiting the typeface to match in the recognition dictionary.