JPH09106433A - Character recognizing processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a result in a quick response by selecting one kind of combination of a character recognizing part and a dictionary whenever one character is inputted and using one of the combinations as a most suitable one. SOLUTION: Characters are inputted in an order from the first one through the use of a character input part 5. Here, the stroke number of the character, namely a picture, and the stroke shape of it, namely the string of coordinate points, are compared with a dictionary 13 to recognize the character. A character recognition part 10 extracts the stroke number and the shape of the stroke from the inputted character and compares them with data in the dictionary 13 to select the character which is the most similar to the piece of data as the recognizing result. Then several characters from the most similar one are outputted as the recognizing result. In addition recognizing result from a first to fifth candidates concerning each character are displayed. Then a handwriting user inputting characters through the use of a character input part 5 collates the results with the characters inputted with a keyboard, a mouse, a pen, etc., to specify correctly recognized characters.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition processing device for recognizing characters by reading sentences input in various character shapes even if they are the same characters such as handwritten characters.

[0002]

2. Description of the Related Art As a method of inputting various data and the like to an information processing apparatus, there is a method of reading an image of a character and recognizing and processing the image to convert it into data. This applies to documents printed in type and the like, and documents input by handwriting. For example, in a device for recognizing a handwritten character, there is a device that draws a character on a tablet with a pen, accepts the position coordinate sequence as a stroke of the character, and collates and recognizes the stroke. For inputting such handwritten characters, a dictionary suitable for the character recognition of the writer is prepared in advance. There are various methods for recognizing characters, such as a method of cutting out a character and a method of comparing with a pattern. When combining such various recognition methods and dictionaries, the recognition rates differ.

Therefore, a writer usually selects in advance a dictionary or a combination of recognition methods for appropriately recognizing a character input by the writer. On the other hand, in order to automate such selection work, all combinations of the recognition program and the recognition dictionary are prepared, and the recognition processing is performed by all the combinations every time one character is input, and the result is the most suitable. A method of selecting a desired one is also introduced (Japanese Patent Laid-Open No. 5-3249).
03 publication).

[0004]

The conventional character recognition processing apparatus as described above has the following problems to be solved. When several recognition methods and dictionaries for character recognition are prepared, there are many combinations of these, so performing recognition processing for all these combinations each time one character is input increases the processing amount, It takes a very long time to output the recognition result. Therefore, when there are multiple users and many dictionaries are prepared, the response from the writer's input of characters until the recognition result is displayed is poor, and in terms of operability. There was a problem.
Moreover, since a large number of recognition results are output, a memory having a large capacity for temporarily storing a large amount of recognition results is required.

[0005]

The present invention employs the following structure to solve the above problems. <Structure 1> The character recognition processing device of the present invention includes a character input unit for inputting a character to be recognized, a plurality of character recognition units for performing recognition processing by referring to a dictionary prepared in advance, A combination selection unit that selects a combination of a plurality of prepared dictionaries and a character recognition unit each time a character is input, in a predetermined order, a display unit that displays the recognition result as a list of candidate characters, and this display unit. The evaluation unit that evaluates the displayed recognition result, and a control unit that specifies a correct recognition result, and a control unit that determines a combination of a dictionary and a character recognition unit with a high recognition rate for the recognition process of the document based on the specified result With.

<Explanation> The input character to be recognized may be a handwritten character or a type character. This invention, even if the same character is entered in two or more types,
It is applied when character recognition is performed by selecting and referring to two or more dictionaries that store patterns corresponding to each character shape. The plurality of character recognition units perform character recognition by different recognition processing methods. Each time one character is input, one of the combinations of the dictionary and the character recognition unit is selected and referenced, and the recognition result is output. Therefore, each time the character is recognized by the combination of the dictionary and the character recognition unit. The response of the recognition process is faster than when processing is performed. Also, the memory for the output buffer may have a small capacity. The dictionary or character recognition unit to be selected may be narrowed down in advance from all dictionaries.

The combination of the dictionary and the character recognition section is switched to a different one each time one character or several characters are recognized, and is displayed each time or every time several characters are recognized and evaluated by the operator. As a result of repeating this processing for an appropriate amount of characters, the combination of the dictionary and the character recognition unit having a high recognition rate is used as the optimum one for recognition of the entire document. The combination used as the optimum one at this time may be two or more kinds instead of one kind. The candidate character is a result of picking up the most similar characters in order as a result of collation with the dictionary, and there may be no candidate character when it is unrecognizable.

<Structure 2> After recognizing a preset number of characters, the recognition result displayed on the display unit is evaluated, and a recognition unit that specifies a correct recognition result and a recognition rate based on the specified result A control unit that determines a combination of a high-performance dictionary and a character recognition unit for recognition processing of the document. <Explanation> When a recognition result is obtained by sequentially applying prepared dictionaries for an appropriate number of characters, the work for determining a combination of dictionaries and the like is completed. Therefore, an appropriate number of recognized characters and the like are set in advance.

<Structure 3> After recognizing a preset number of characters, the recognition result displayed on the display unit is evaluated to specify a correct recognition result, and a dictionary having a low recognition rate as a result of the specification. It is preferable to include a combination selection unit that excludes the combination of the character recognition unit and the character recognition unit from the selection target. <Explanation> Appropriate dictionaries are narrowed down by repeating recognition using a plurality of dictionaries and evaluation thereof, but in that case, inappropriate dictionaries may be excluded.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to specific examples. <Specific Example> FIG. 1 is a schematic view showing a specific example of the character recognition processing apparatus of the present invention. This device has a display 1 and
A storage unit 4, a character input unit 5, and the like are connected to an information processing apparatus including a keyboard 2 and a control unit 3. In this case, the display 1 is used for displaying and evaluating the character recognition result. In the present invention, this display 1 is called a display unit. The keyboard 2 is also used for inputting evaluation results and other operations. In the present invention, this is called a designation unit. The control unit 3 includes a processor and the like, and includes a character recognition processing program and other control mechanisms. The storage unit 4 is composed of a hard disk, a floppy disk or the like, and stores various dictionaries 13 referred to for character recognition therein. It should be noted that these dictionaries 13 are prepared in advance in order to recognize handwritten input by a plurality of writers or characters of various fonts. A method of generating such a dictionary is a well-known method in the related art, and a detailed description thereof will be omitted here.

A plurality of character recognition units 10 are incorporated in the control unit 3 for the character recognition processing of this apparatus. These character recognition units 10 are composed of processing programs and the like for performing character recognition processing by different recognition methods,
For example, various well-known methods such as a character cutting method, a stroke recognition method, and a dictionary matching method are used. A plurality of these are prepared so that the most suitable one can be selected from the characteristics of the input characters. Selector 1
Reference numeral 1 denotes a portion for selecting a combination of the character recognition unit 10 and the dictionary 13 and executing the character recognition process in a preset order or the like at the time of character recognition. When inputting a character using the pen 6, the character input unit 5 is a well-known device that detects the stroke as a sequence of coordinate points and sends it to the control unit 3.

Although the apparatus of the present invention is particularly suitable for inputting such handwritten characters online, it can be similarly applied even when the handwritten characters are already recorded in a storage device such as the floppy disk 7. In addition to this, it is also applicable to the case of inputting data using an optical reading device such as an image scanner or a CCD camera. Further, the device of the present invention can be applied to a case where the character input unit 5 is replaced with a pointing device such as a keyboard or a mouse, and a character is input on the display 1 by handwriting.

FIG. 2 is an explanatory view of a combination of the dictionary and the character recognition section. That is, FIG. 2 is a diagram for explaining what kind of result is obtained by the combination of the character recognition unit and the dictionary. For example, suppose there are five writers A, B, C, D, and E here. In this case, it is assumed that three types of character recognition units P, Q, and R and three types of dictionaries X, Y, and Z are prepared.
At this time, for the writer A, if the character recognition unit is Q and the dictionary is Y, and the recognition rate is the highest, this is a suitable combination. For the writer B, the character recognition unit is Q, the dictionary is X, and the writers C, D, and E are
Combinations such as P, R, Q for the character recognition unit and X, Z, Y for the dictionary are suitable combinations. In the present invention, the main configuration is to determine such a combination of the character recognition unit and the dictionary for each writer automatically and as quickly as possible.

FIG. 3 is an explanatory view of combination numbers of the dictionary and the character recognition section. If there are three types of character recognition units P, Q, and R and three types of dictionaries X, Y, and Z as described above,
As shown in this figure, there are nine combinations of the two. In this embodiment, a combination number S is assigned to each of these combinations.
Each time one character is input, the combination number S is selected and the recognition process is executed. This selection method includes, for example, a method in which the combination number S is sequentially changed from 1 to 2, 3, and 4, and a sequence of combination numbers is randomly determined in order to prevent a certain combination from being the last. There is a method to change the order freely. Also, unnecessary combinations can be excluded in advance.

FIG. 4 shows an example explanatory diagram of an input document. In this example, the characters "I am a member of society" are sequentially input from the first character to the eighth character. On the other hand, in FIG.
For example, an explanatory view of a candidate character display example when a certain combination is selected for the second character is shown. 2 for display 1
As a result of recognizing the th character, for example, five types are displayed as shown in this figure. Of these five types, for example, the ones with high possibility are arranged in order from the leftmost. That is, the target character is collated with the dictionary, and distance data corresponding to the difference between each character in the dictionary and the input character is obtained. The one having the closest distance data is the first candidate and is the leftmost. Put on. Hereinafter, the second to fifth candidates are displayed. While seeing such a display, the writer uses the keyboard 2 shown in FIG. 1 to specify what matches the corresponding character of the actual input document. In this way, the character recognition result is evaluated.

<Operation> The specific operation of the apparatus of the present invention for the above-described input document will be described below with reference to FIG. FIG. 6 is a diagram illustrating a recognition result display example. First, FIG.
The characters shown in FIG. 4 are input in order from the first using the character input unit 5 of the device shown in FIG. At this time, character recognition is performed by comparing the number of strokes of the character, that is, the number of strokes, and the shape of the stroke, that is, the coordinate point sequence, with a dictionary. The dictionary stores data obtained by collecting characters to be recognized from a plurality of people, detecting the number of strokes and stroke shapes, and performing averaging processing. The plurality of dictionaries are generated from, for example, data collected from different persons, or based on various data. The character recognition unit 10 extracts the number of strokes and the stroke shape from such input characters, compares them with the dictionary data, and sets the character having the highest degree of similarity as the recognition result.
Then, as shown in FIG. 5, several characters are output as a recognition result from the character having high similarity. The character recognizing unit 10 recognizes the number of strokes of a character, a vector connecting a start point and an end point of a stroke, a vector connecting an end point of a stroke and a start point of the next stroke, and the like in a matching order with a dictionary. As shown in this figure, the combination numbers of the character recognition unit and the dictionary are set to 1, 2, 3, 4 up to the 8th character in the order of the 1st character, the 2nd character, and the 3rd character, and this is repeated twice. There is. Then, for each character, the recognition results from the first candidate to the fifth candidate are displayed. The writer who has input a character using the character input unit 5 collates the input character with a keyboard, a mouse, a pen or the like, and designates the correctly recognized character. For example, for the second character, the combination with the number 2 is used, and the first candidate is erroneously recognized. Therefore,
The writer specifies the second candidate, "ha". Therefore, the first character is the first candidate, the second character is the second candidate, the third character is the first candidate, the fourth character is the first candidate, the fifth character is the third candidate, the sixth character. Is the first candidate, the seventh character is the first candidate, the eighth character is the third candidate, and so on. The control unit 3 shown in FIG. 1 receives this input and calculates the goodness of fit. Note that such correct answer designation may be omitted for the first candidate, and the burden can be reduced.

FIG. 7 shows an explanatory diagram in which the result of such evaluation is calculated as the goodness of fit. Here, the correct character is the i-th
In the case of a candidate, the matching degree of the character is expressed as V (i).
If there is no correct answer in the candidates, the goodness of fit is “0”. Therefore, V (1) becomes 1.0 and V (5) becomes 0.2. In this way, if the goodness of fit is provided from the result shown in FIG. 6, as shown in this figure, the first, third, fourth,
The sixth and seventh characters have a goodness of fit of 1.0, while the other characters have a goodness of fit of 0.8 or 0.6. Here, since the combination of the character recognition unit and the dictionary is switched for each character, the matching degrees when the same combination is used are summed up.

FIG. 8 shows an explanatory diagram of the result of the sum of the goodness of fit. As shown in this figure, combination numbers 1, 2, 3,
It can be judged that the combination number 3 is the most suitable when looking at the total result of the matching degrees of 4. In this way, the combination of the character recognition unit and the dictionary is determined. In the above case, the number of combinations was set to 4, and the combination was changed by one character, and the final evaluation was performed when 8 characters were recognized in order to repeat this twice. In this way, the optimum combination can be determined by comparing the total of the matching degrees after recognizing only certain characters. In this case, the number of characters is a positive multiple of the number of combinations.

In the above example, an example in which another combination is selected for each character has been described, but a method of using the same combination for several characters and another combination for the following several characters is also possible. . Further, in addition to using the degree of conformity as described above, various evaluation methods may be adopted. Further, when there are two or more combinations having a high degree of conformity, it is possible to use not only one combination but also two combinations for recognition processing of the document. Further, although an example in which a certain number of characters is set in advance and a certain combination is selected when the recognition of the number of characters is finished has been described, an inappropriate combination is deleted one by one each time evaluation is performed, or It is also possible to decide the final combination in several stages by a method of leaving the combination of the upper half for each evaluation and deleting the other half. By doing so, it is possible to determine the most suitable combination of the dictionary and the character recognition section and facilitate the subsequent processing.

The combination once decided is returned to the initial state after an appropriate time, the selection process of the combination is restarted again, or the same selection process is performed when more than a certain number of characters are input. A method of restarting is also possible.

<Effect> The character recognition processing apparatus described above selects a combination of one type of character recognition unit and a dictionary each time one character is input, displays the recognition results in order, and stores this result a predetermined number of times. Since any combination is repeatedly used as the most suitable one, the recognition processing for each character becomes one combination, the processing amount can be reduced, and the result can be obtained with a fast response. Further, if a buffer memory for temporarily storing the recognition result is provided on the character recognition unit side, a small amount of memory is required to hold the recognition result for each character, and it is possible to reduce the size and cost of the memory. . The device of the present invention is
(Optical character reader), computer, electronic notebook,
It can be widely applied to electronic book readers and other devices that recognize various characters.

[Brief description of the drawings]

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

FIG. 2 is an explanatory diagram of a combination of a dictionary and a character recognition unit.

FIG. 3 is an explanatory diagram of combination numbers of a dictionary and a character recognition unit.

FIG. 4 is a diagram illustrating an example of an input document.

FIG. 5 is a diagram illustrating a candidate character display example.

FIG. 6 is a diagram illustrating a recognition result display example.

FIG. 7 is an explanatory diagram of a goodness of fit.

FIG. 8 is an explanatory diagram of a result of total fitness.

[Explanation of symbols]

 1 Display (display unit) 2 Keyboard (designating unit) 3 Control unit 4 Storage unit 5 Character input unit 6 Pen 10 Character recognition unit 11 Selection unit 13 Dictionary

Claims (3)

[Claims] 1. A character input unit for inputting a character to be recognized, a plurality of character recognition units for recognizing the character by referring to a dictionary prepared in advance, and a character prepared in advance each time one character is input. A combination selection unit for selecting and selecting a combination of the plurality of dictionaries and the character recognition unit in a predetermined order, a display unit for displaying the recognition result as a list of candidate characters, and an evaluation of the recognition result displayed on this display unit. And a control unit that determines a combination of a dictionary having a high recognition rate and the character recognition unit for recognition processing of the document based on the result of the designation. Characterized character recognition processing device. 2. A recognition unit which evaluates the recognition result displayed on the display unit after recognizing a preset number of characters, and specifies a correct recognition result, and a recognition rate of the recognition rate based on the specified result. The character recognition processing apparatus according to claim 1, further comprising a control unit that determines a combination of a high dictionary and the character recognition unit for recognition processing of the document. 3. A designation unit for designating a correct recognition result by evaluating a recognition result displayed on the display unit after recognizing a preset number of characters, and a dictionary having a low recognition rate as a result of the designation. The character recognition processing device according to claim 1, further comprising a combination selection unit that excludes a combination with the character recognition unit from selection targets.