JPS60220481A - Character recognizer - Google Patents

Abstract

PURPOSE:To lighten the labor for inputting by repeating an input of a sample character and a category name after a category name input start signal is inputted and by registering feature data of plural arbitrary sample characters. CONSTITUTION:After a category name input start signal is inputted from an input part 5, a sample character category name on a sample form is inputted to a registration part 6. Each sample character entered in a sample form 1 is quantized in a character input part 2 and its feature values are extracted in a recognition part 3. After the category name input start signal is inputted from the input part 5, the registration part 6 receives the category name input. Then it combines said category name with the extracted feature values and registers the result as a combination dictionary factor in a dictionary 7. A display part 4 displays the sample character pattern inputted from the input part 2 and the category name inputted from the input part 5. The registration part 6 terminates the dictionary registration by means of a category name input termination signal from the input part 5.

Description

[Detailed Description of the Invention] (Industrial Application Field) The present invention relates to an optical character recognition device (hereinafter abbreviated as OCR). The present invention relates to a character recognition device that uses a dictionary created from feature data.

(Prior art and its problems) In conventional OCR, a feature series that is used as a criterion for judgment is created from the feature values obtained from many learning characters, and a general-purpose stationary sequence created by integrating it with the category name is created. has been used.

However, when the recognition target characters include characters that have similar shapes but belong to different categories, or characters that have individual quirks, these feature values take on similar values, which can have a large impact on the judgment result. Therefore, in order to solve the above-mentioned problem, a personal dictionary created from the features extracted from sample characters written by the person filling out the form was devised.

This personal dictionary has already been used once in speech recognition and real-time character recognition, but it has a long history.

With the spread of computers, the necessity of character recognition devices using character recognition has increased, and it is becoming more important to improve the recognition rate of OCt+. It is created by combining the percentile data obtained from the characters on the sample form and the category name of the same character, but conventionally, when giving the category name, the sample character entry frame How to set the category name corresponding to the printed character ketemple character mentioned above, or input the category name human power signal from the human power department when reading sample characters, input the category name, and perform this input operation for all sandal characters. An iterative method was used. In either method, the category name sent to the registration unit is combined with the percentage value extracted from the sample characters and registered as a dictionary element.

However, the former method ``C'' requires the form filler to fill out a plurality of 1il-sample forms in order to create a personal dictionary. Furthermore, the sample form may include characters and symbols that the person filling out the form does not use on the form. On the other hand, there is a drawback in that the characters and symbols used on the form may not be included in the sample form.

The latter method has the disadvantage that it is necessary to input a category master signal every time a category master is performed, and it requires a great deal of manual effort when a large number of category masters are performed.

(Object of the Invention) An object of the present invention is to provide a character recognition device that requires only the category master's human power after inputting the category master's human power start signal, thereby reducing both the time and effort required for manual operation of the human power section. .

(Structure of the Invention) The character recognition device of the present invention runs an input character and generates a quantized image.
A character input section to be converted into a bell, a word (t, and an index of this word 1) in which several pieces of feature data extracted from the category name of the input character and the quantized stroke are registered as w-book elements. A recognition unit that determines the input character from the above-mentioned Kenko-ized picture iM#, a category master power start signal, and a category name of a sample character for dictionary creation.

Category name manual end (input section for inputting No. 8 in this order, input of MjJ category name according to the category name manual start signal, and input of feature data extracted from the category name and the sample character) a registration unit that starts registration in the dictionary as an element, and ends input of the category name and registration in the dictionary in response to the category name input end signal; and a display unit that displays the category name and the character button. After inputting the start signal within the category when creating a dictionary, you can create any number of sample characters by repeating inputting the sample characters into the character input section and inputting the category name into the human input section. The feature is that the feature data of is registered in the dictionary.

(Example) Hereinafter, the present invention will be described with reference to drawings showing examples.

FIG. 1 shows a first embodiment of the present invention, in which after inputting a category name manual start signal from an input section 5, a sample character category name on a sample form is inputted to a registration section 6. At the same time, each sample character written on the sample form 1 is quantized in the character input section 2, and the % signature value is extracted in the recognition section 3. When the registration unit 6 receives the category master input after inputting the category master input start signal from the input unit 5, it thereafter registers the same category name and the extracted feature value in the dictionary 7 as a combination dictionary element. Here, the display section 4 displays the sample character ``bang'' input from the character input section 2 and the category name input from the input section 5.

Registration s6 ends dictionary registration in response to the category expert end signal from input @Bs.

FIG. 2 is a block diagram showing the first embodiment of the present invention in more detail. In FIG. 2, 2 is a character input section, 3 is a recognition section, 6 is a registration section, 7 is a dictionary, 4 is a display section, and 5 is an input section.

The human characters are first quantized in the character input section 2, and the recognition section 3 extracts the percentage value from the quantization button.

In addition, the category expert power start signal is input from the input unit 5 to the registration unit 6.
Then, the category name is inputted from the doujinshi department into the registration section 6 and displayed on the display section 4 at the same time. The registration unit 6 receives the category name and registers the feature value and the category name in the dictionary 17 as a combination dictionary element.

After registering in the dictionary, when reading a form, the recognition unit 3 performs determination processing using the feature values extracted by the recognition unit 3 from the input character stamps quantized by the character input unit 2 and the feature values registered in the dictionary. .

It is assumed that a combination of many known methods will be adopted for Tanishi Seibu, and a known method using a dictionary will be used as the recognition method.

Further, as the recognition method using a personal dictionary, a known method is used, and its details are described in [Character Recognition Method (Japanese Patent Application No. 1988-088674) J.

FIG. 3 shows a category input start signal, a category name, a category name manual end signal, and sample characters on a sample form that are input from the input device in the present invention. In Figure 3, 31 is a sample form, 32 is a sample character, 3
3 indicates an input signal.

Here, sample characters "A", "IJ", "U", "EJ", and "J" are entered in the sample form 31, and "UJ" is a character that does not need to be registered in the dictionary. First, input the category master power start signal r$ SJ from the input section to the sample character category master power, and then input the category name "a" and "i".
Enter. For the characters that do not require registration, enter rXJ, which is predetermined to indicate that registration is not required, and finally enter "Category Meijinriki End Tsukuda Misaki E" to end the category Meijinriki.

FIG. 4 shows an example of the display on the display unit in the second embodiment of the present invention. In Fig. 4, 40, 41, 42°, 43.
44, 45, 46, 47, 48, and 49 are category names displayed on the display unit and registered in the dictionary.

FIG. 5 shows an example of the display on the display unit in the third embodiment of the present invention. In FIG. 5, 50, 51, 52°, 53.
54.55 are category names that have been displayed on the display section and registered in the dictionary, and 9.56 and 57.58 are category names that have been input multiple times in the input section.

FIG. 6 shows an example of the display on the display unit in the fourth embodiment of the present invention. In Fig. 6, 60, 61, 62°63.
64 is the category name registered in the dictionary displayed on the display section, 9.67.68.69 is the category name input from the input section, and 9.65.66 is the category name that has been input many times by the human resources department. This is a category name that has not been registered in the dictionary.

FIG. 7 shows an example of the display on the display unit in the fifth embodiment of the present invention. In Fig. 7, 70, 71, and 72 are category names displayed on the display and registered in the dictionary.
73, 74, and 75 are character slams corresponding to the same category name.

FIG. 8 shows an example of the display on the display unit in the sixth embodiment of the present invention. In FIG. 8, 80 is a category name registered in the dictionary displayed on the display section, and 81 is a character button corresponding to the same category name. 82 is the category name input from the input section, and 83 is a character button corresponding to the same category name.

FIG. 9 shows an example of display on the display unit in the seventh embodiment of the present invention. In FIG. 9, 92, 94, 95°96.
97, 98.99 are category names registered in the dictionary displayed on the display section, 9.100, 101, and 102 are category names that have been input multiple times in the input section, fi, 90°91.
93 is a category name registered in the dictionary that is the same as the category name input from the input section. When character slam itself is used as feature data, the input category name and the same character slam are combined, and the dictionary By using the dictionary as an element, a dictionary is created, and a device 4 for reading a form using the dictionary is obtained. Furthermore, by displaying already registered category names on the display device when performing category masters, waste of category masters can be avoided and dictionary creation can be made more efficient. The performance of the dictionary can be improved by displaying the category name and same character button of already registered characters on the display device and selecting whether to register new sample characters of the same category. moreover,
The person filling out the form is freed from the traditional burden of filling in multiple sample forms for creating personal dictionary, and only the character characteristic values and category names necessary for reading the form are registered in personal dictionary 1. It is possible to shorten the time required to make a profit.

(Effects of the Invention) According to the present invention, when creating a personal dictionary 1, compared to a method in which a number of pre-prepared sample forms are filled in and character and character characteristics on the same forms are extracted and created. For the form filler to read the form, thick characters are written on the form, features are extracted, and the category name of the same character is input by the input device, so the dictionary capacity is smaller than before and the judgment can be made better. Processing speed is also greatly reduced because there are fewer candidate characters.

The present invention not only improves the handwritten character recognition device using the above-mentioned personal dictionary, but also improves the efficiency of dictionary creation in terms of printed character recognition.

[Brief explanation of the drawing]

1 and 2 are block diagrams showing a first embodiment of the present invention, and FIG. 3 shows a series of inputs (
An explanatory diagram showing the number M, FIG. 4 is a display diagram showing a display example of the display section in the second embodiment of the present invention, and FIG. 5 is a display diagram showing a display example of the display section in the third embodiment of the present invention. Display diagram, 6th
FIG. 7 is a display diagram showing a display example of the display unit in the fourth embodiment of the present invention, FIG. 7 is a display diagram showing a display example of the display unit in the fifth embodiment of the present invention, and FIG. FIG. 9 is a display diagram showing a display example of the display unit in the sixth embodiment of the present invention, and FIG. 9 is a display diagram showing a display example of the display unit in the seventh embodiment of the present invention. In the figure, 1.31 is a sample form, 32 is a sample character, 2 is a character input section, 3 is an authentication Bk section, 5 is an input section, 6
is the registration section, 7 is the dictionary, 4 is the display +'S: 5s33 is entry number 8, 40, 41, 42, 4.3, 44, 45, 46
゜47.48, 49.50, 51.52, 53, 54,
55,60,61゜62.63,64,70,71,7
2,80,92,94,95,96゜97.98.99
is the category name already recorded in the dictionary displayed on the display %5
6,57,58.67.68,69,82゜100.1
01.102 is the category name that has been input multiple times in the input section, 65
, 66 is a category name that is not registered in the dictionary among the category names input from input s, 73, 74゜75.81.8
3 is the category name 70, 71, 72゜80.82 respectively
The character bangs corresponding to 90, 91, and 93 are the category names input from the input section among the category names already registered in the dictionary, and the category names that have been deleted. Agent Patent Attorney Nihara Uchihara, -1

Claims (1)

[Scope of Claims] 1. A character input section that scans input characters and converts them into quantized images, and a plurality of category names of the input characters and % sign data extracted from the quantized images are registered as dictionary elements. A dictionary, a recognition unit that determines the input character from the quantized picture based on the index of this dictionary, a category master human power start signal, a category name of a sample character for dictionary creation, and a category master human power end signal in this order. inputting the category name using the human power section, inputting the category name using the category name human power starting number, and starting registration in the dictionary using the category name and the percentage index extracted from the sample characters as dictionary elements, and A registration section for terminating input of the category name and registration in the dictionary in response to a master's power end signal, and a display section for displaying the category name and the character button, After inputting the symbol, the characteristic data of any number of sample characters can be registered in 2#1 by repeatedly inputting the sample character into the character input section and inputting its category name into the input section. Character recognition device. 2. Displaying the category name registered in the Erasing Dictionary on the front dC display section %1: Character recognition device according to claim 1,
. 3. A character recognition device for HC notebooks in claim 1, which displays the category name registered in the dictionary and the category name input in the upper part of the nil entry on the front H display section. 4. Registered in the dictionary. Displaying the category name on the display section, and displaying the same category name on the display section if the category name input through the input section is unregistered. Character recognition device. 5. The character recognition device according to claim 1, wherein all or some of the category names registered in the dictionary and a character button corresponding to the category name are displayed on the display unit. 6. Display on the display section the character slam corresponding to the full strength lever name or some category names and the open force category name registered in the dictionary, and the same character pattern as the name of category 1 input from the input section. A character recognition device according to claim 1 which is displayed. 7. Claim No. 7, wherein all category names to be registered input from the input section are displayed on the display section, and category names that are the same as the category names input from the input section are erased from the display section. Character recognition device according to item 1.