JPH09237322A - Candidate correcting system for recognizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve a recognition rate by utilizing operator's recognizing result correcting information for later recognizing processing. SOLUTION: A recognizing means 2 compares an inputted character image with the standard feature of each category of a recognizing dictionary 3 to obtain plural high-ordered recognizing candidate categories of which distance value being a difference between both of them is small. When a pair including an erroneous reading category coinciding with a first recognizing candidate category is stored in a correcting table 5, a candidate correcting means 4 subtracts a value corresponding to information for correction in the pair from the distance value of the recognizing candidate category coinciding with a right answer category in the pair and after then, sort it by the distance value to change a candidate order. A result correcting means 6 outputs a first-order candidate after this change to a display means 6 as a recognizing result. When the recognizing result is wrong and the operator inputs the right answer category, a table updating means 9 additionally registers the right answer category and the pair of the distance difference of them in the correcting table 5.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data recognition device such as a character recognition device for recognizing characters written on a form,
More specifically, when an operator corrects a correct recognition result due to an error in the recognition result, the correction information is utilized in the subsequent recognition processing to improve the recognition accuracy, so that the recognition accuracy is improved. Regarding

[0002]

2. Description of the Related Art It is extremely difficult to achieve a recognition rate of 100% in a character recognition device, especially a handwritten character recognition device. For this reason, a semi-automatic system is generally adopted in which the recognition result is displayed, the operator visually checks it, and replaces the erroneously read character with a correct character. Various techniques have been conventionally proposed to support the recognition result correction work by such an operator.

FIG. 14 is a block diagram showing an example of the configuration of a conventional character recognition device, and is a functional block diagram of the configuration of the character recognition device described in Japanese Patent Laid-Open No. 61-74081 (hereinafter referred to as Document 1). It has been transformed.

In FIG. 14, a character written on a form or the like is input as a character image image by the data input means 10 such as a scanner and given to the recognition means 22.
The recognition dictionary 21 stores standard features of character patterns for each character category, and the recognition unit 22 extracts recognition features from a character image image, and the standard features for each character category in the recognition dictionary 21. And the recognition candidate categories up to the fifth candidate are calculated, and the candidate adding means 50 is added.
Output to

The candidate adding means 50 performs a candidate adding process, which will be described later, if necessary, and then outputs all recognition candidate categories to the result correcting means 40, and the result correcting means 40 recognizes the first candidate recognition candidate. The category is displayed on the display unit 41 as a recognition result.

The operator looks at the recognition result displayed on the display means 41 and checks whether or not the characters on the form are correctly recognized. Then, in the case of misreading, the fact that the misreading is performed is input from the correction information input means 42 such as a keyboard. In response to this, the result correcting means 40 displays on the display means 41 the recognition candidate categories after the second candidate outputted from the candidate adding means 50.

Therefore, for example, although the character written on the form is "mi", the recognition candidates determined by the recognition means 22 are "width", "ke", and "sa" in order from the top. ，
If the candidate addition means 50 does not add a new candidate, the width of the first candidate is displayed on the display means 41 as a recognition result, and the operator misreads. , The remaining candidates “ke”, “sa”,
“H” and “ya” are displayed on the display means 41.

If the displayed candidate includes a correct answer, the operator selects the correct answer to cause the result correcting means 40 to change the recognition result to the selected candidate. As described above, when the correct answer is not in the candidates, the correct answer "mi" is input from the correction information input means 42 by, for example, kana-kanji conversion, and the correction is instructed. When the correct answer is input, the result correcting means 40 performs a process of changing the recognition result to the correct answer, and at the same time, the misreading category (the category of “width” which was the first candidate) and the correct answer category (input by the operator). "Mi" category)
And are sent to the table updating means 52. Table updating means 5
2 additionally registers the set of the misread category and the correct answer category in the category table 51 if it is not registered.

[0009] By the way, when such a category table 51 is registered, after that, a similar typeface "mi" is again displayed.
The recognition means 22 outputs the recognition candidates of “width”, “ke”, “sa”, “H”, and “ya” in order from the higher order and outputs them to the candidate addition means 50 as in the above. Then, "mi" is added to the new candidate by the candidate addition processing by the candidate addition means 50. That is, the candidate adding means 50 searches the category table 51 for a misreading category “width” that is the same category as the category “width” of the first candidate sent from the recognizing means 22, and if found, the misreading category “width”. The correct answer category "mi" paired with is added to the candidates as the second candidate. Therefore, in the above case, the new recognition candidate is
“Width”, “mi”, “ke”, ..., The recognition result first displayed on the display means 41 is “width”, which is the same as the previous time, but is not displayed last time when the operator gives an error reading. The "mi" will be displayed on the display means 41 as a recognition candidate, and the correct answer can be corrected simply by selecting the displayed "mi".

A technique similar to the conventional technique shown in FIG. 14 is also described in Japanese Patent Application Laid-Open No. 1-100686 (hereinafter referred to as Document 2).

On the other hand, JP-A-59-27375 (hereinafter referred to as Document 3), JP-A-63-208180 (hereinafter referred to as Document 4) and JP-A-5-73709 (hereinafter referred to as Document 3). 5)), when the recognition result of a certain character is corrected in the series of character strings displayed as the recognition result, there is a place where the same correction should be performed in the subsequent recognition results. There is described a technology that automatically detects whether or not to do so and can collectively correct a plurality of misread portions of the same type by a single correction operation by an operator.

[0012]

As described above, in the conventional character recognition device, various techniques have been proposed to assist the operator in correcting the recognition result, but there are the following problems. .

According to the techniques disclosed in Document 1 and Document 2,
Since the correct answer category registered in the category table 51 is only added as a recognition candidate having a fixed rank (second rank in the above example), the recognition result displayed first is always the recognition means being the first rank candidate. Limited to what you have done. That is, the recognition result correction information by the operator does not affect the first-ranked candidate at all. Therefore, even if the correct answer content rate in which the correct answer is included in the candidates is improved, the recognition rate in which the first-ranked candidate is the correct answer cannot be improved. Therefore, it is difficult to reduce the number of corrections performed by the operator.

Further, the techniques disclosed in Documents 3 to 5 are
Although the efficiency of the correction work for a series of recognition results is realized, the recognition result correction information input by the operator in the correction work is used only for the correction of the recognition result this time. The recognition of the recognition target character string produces the same result again. That is, the recognition rate cannot be improved by utilizing the recognition result correction information of the operator.

The present invention has solved such a conventional problem, and its purpose is to utilize the recognition result correction information of the operator for the subsequent recognition processing, and to determine the correct answer content rate in which the correct answer is included in the candidates. In other words, it is to improve the recognition rate in which the first candidate is the correct answer.

[0016]

According to the present invention, a recognition feature is extracted from recognition target data such as a character image image and compared with a standard feature for each category registered in a recognition dictionary in advance to recognize the recognition feature. A plurality of recognition candidate categories having a small distance value which is a difference between the feature and the standard feature, a recognition means for obtaining the distance values thereof, and a correction table for storing a set of misreading category, correct category and correction information. And a set including a misreading category that matches the first-ranked recognition candidate category in the recognition candidate category group obtained by the recognition means is stored in the correction table, it matches the correct category in the set. Candidate correction means for reducing the distance value of the recognition candidate category by a value according to the correction information in the set, and the first candidate in the recognition candidate category group after processing by the candidate correction means. Position recognition candidate category is output as a recognition result, and the result correction means corrects the recognition result in accordance with the correction information inputted from the correction information input means, and a misread category, a correct answer category and a correction category are corrected at the time of correction by the result correction means. Table updating means for adding a combination with the usage information to the correction table.

Further, when the recognition candidate category matching the correct answer category does not exist in the recognition candidate category group obtained by the recognizing means, the candidate correcting means sets a predetermined distance value that matches the correct answer category. The distance value is corrected after adding the given recognition candidate category to the recognition candidate category group.

The correction information in the set added to the correction table by the table updating means is, for example, (1) a distance difference which is a difference between the distance value of the misread category and the distance value of the correct category. (2) A distance difference, which is the difference between the distance value of the misread category and the distance value of the correct category, and the recognition feature extracted from the recognition target data. Is used.

When the correction information as in the above (1) is used, the candidate correction means corrects the distance value by the following equation, for example.

D ′ = D−w · s where D ′ is the corrected distance value, D is the uncorrected distance value,
w is a weighting coefficient, and s is a distance difference which is correction information.

Further, when the correction information as in the above (2) is used, the candidate correction means corrects the distance value by the following equation, for example.

D ′ = D−w · s · a / d where D ′ is the corrected distance value, D is the uncorrected distance value,
w is a weighting coefficient, s is a distance difference in the correction information, d is a distance value between the recognition feature in the correction information and the recognition feature of the current recognition target data, and a is a constant for normalizing the distance value d. It is a certain reference distance value. Further, the candidate correction means is a new first-ranked candidate category when the first-ranked recognition candidate category in the recognition candidate category group obtained by the recognition means is changed by the candidate correction using the correction table. Has a configuration in which the candidate is corrected again by using.

In the present invention configured as described above, when the operator corrects the correct category at the time of misreading, the correct category, the misreading category (the category output by the device as the first candidate) and the correction information are provided. Is registered in the correction table. After that, when the same category as the misreading category is obtained as the first recognition candidate category, the distance value of the recognition candidate category of the same category as the correct answer category corresponding to the misreading category is reduced. Therefore, depending on the size of the value to be reduced, it may jump to the first place candidate. This is a big difference from the techniques described in Documents 1 and 2, and the recognition result displayed first is not limited to the one in which the recognition means is always the first candidate, and is the category previously input as the correct category. May be output as a recognition result. With such an action, it is possible to improve the recognition rate in which the first candidate is the correct answer.

[0024]

BEST MODE FOR CARRYING OUT THE INVENTION Next, an example of an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of a character recognition device to which the present invention is applied. As shown in the figure, the character recognition device of this embodiment includes a data input means 1 and a recognition means 2.
The recognition dictionary 3, the candidate correction means 4, the correction table 5, the result correction means 6, the display means 7, the correction information input means 8 and the table update means 9 are provided.

FIG. 2 is an operation flowchart of the embodiment shown in FIG. Hereinafter, the function and operation of each unit of this embodiment will be described with reference to FIGS. 1 and 2.

The data input means 1 is composed of, for example, a scanner, and scans, for example, handwritten characters written on a form (not shown) to generate a character image image for each character and outputs it to the recognition means 2.

The recognition dictionary 3 stores in advance the recognition characteristics (standard characteristics) of the standard character pattern for each character category. The recognition unit 2 extracts the recognition feature from the character image image input from the data input unit 1, and the recognition dictionary 3
In comparison with the standard features stored in advance in each character category, a plurality of recognition candidate categories with a small distance value (for example, five up to the fifth place) and their respective distance values are obtained (S1). ). The obtained recognition candidate category group and each distance value are output to the candidate correction means 4. Here, the distance value is the difference between the recognition feature and the standard feature, and the smaller the similarity, the smaller the value. Various methods have been conventionally proposed as a method for performing character recognition by obtaining the distance value between the recognition feature of the recognition target character and the standard feature in the recognition dictionary, and the present invention may use any of the techniques. it can. For example,
As a document describing such a method, "Improvement of Directional Pattern Matching Method and Application to Handwritten Kanji Recognition" (199)
Jun. 0, IEICE Technical Report, PRU90-
There is 20).

The correction table 5 is a table for storing the set 50 of the misread category 51, the correct answer category 52, and the distance difference 53, and has a capacity capable of storing, for example, 200 sets. In this embodiment, the distance difference 53, which is the difference between the distance value of the misread category 51 and the distance value of the correct category 52, is used as the correction information. The initial state of the correction table 5 is an empty state, that is, a state in which no set is stored. The correction table 5 can be exchanged for each user or for each type of recognition target (type of form, etc.).

The candidate correction means 4 refers to the correction table 5 and newly recognizes the recognition candidate category group (five recognition candidate categories from the first rank to the fifth rank) input from the recognition means 2. It is a means for adding a candidate category or reducing the distance value of a predetermined recognition candidate category and then sorting by the distance value to generate a new recognition candidate category group, and executes the processing of steps S2 to S10 of FIG. To do.

First, the candidate correction means 4 searches for the misreading categories 51 in the set 50 in the correction table 5 one by one, for example, in order from the set closest to the beginning of the table, and the misreading categories are input from the recognition means 2. It is checked whether the category is the same as the category of the first candidate in the recognition candidate category group (S2, S3, S4). Then, every time a misreading category of the same category as the first candidate is found (YE in S4
S), the following processing is executed.

First, it is determined whether or not the recognition candidate category of the same category as the correct category 52 in the set 50 including the found misreading category is included in the recognition candidate category group input from the recognition means 2. One by one searching the recognition candidate categories from the group (S5, S
6, S7).

Then, when there is a recognition candidate category of the same category as the correct answer category 52 (YE in S7).
S), the distance value of the recognition candidate category is corrected by the following equation (S9).

D ′ = D−w · s Here, D ′ is the corrected distance value, D is the uncorrected distance value,
w is a weighting coefficient, and s is a distance difference 53 which is correction information in the set 50 including the correct answer category 52. Here, the weighting coefficient w may be fixed to, for example, 1 or may be set to a value smaller than 1 for an old storage of the set 50 in the correction table 5.

On the other hand, when there is no recognition candidate category of the same category as the correct answer category 52 (YE in S6)
S), a recognition candidate category of the same category as the correct answer category 52 is added to the recognition candidate category group (S8). At this time, as the distance value of the added recognition candidate category, for example, a predetermined constant α (for example, 200) is added to the distance value of the lowest recognition candidate category in the recognition candidate category group input from the recognition unit 2. Value. And
Proceeding to step S9, the distance value of the added recognition candidate category is corrected by the above equation.

After the above processing, the candidate correction means 4
The recognition candidate categories in the recognition candidate category group are sorted in ascending order by their distance values (S10). At this time, if the distance value is corrected, the candidate rank may be different from the time when the candidate rank is output. Among the newly ordered recognition candidate category groups, a plurality of high-ranking recognition candidate categories (for example, 5 ranks up to the fifth rank) and their distance values are output to the result correcting means 6.

The result correcting means 6 displays the first recognition candidate category in the recognition candidate category group input from the candidate correcting means 4 as the recognition result on the display means 7 composed of, for example, a CRT (S11). . In addition, the second and subsequent recognition candidate categories are displayed in another display area of the display means 7 for correction in the case of misreading. The display of the second and lower recognition candidate categories may be performed at the time when the operator gives an instruction to misread.

The operator checks the recognition result displayed on the display means 7, and if it is misread, inputs the correction information from the correction information input means 8 such as a keyboard to make a correction. At this time, when the correct answer category is present in the second or lower recognition candidate categories displayed on the display means 7, the information for selecting the correct answer category is given from the correction information input means 8 (YES in S12). The result correction means 6 corrects the recognition result into the selected correct answer category (S13). When the correct answer category does not exist in the second or lower recognition candidate categories, the operator inputs the correct answer category by the kana-kanji conversion or the like through the correction information input means 8 and corrects it. At this time, the result correction means 6 corrects the recognition result with the input correction category (S13).

Then, when the recognition result is corrected, the result correcting means 6 updates the table (correct reading category) and its distance value before correction, and the corrected category (correct category) and its distance value to the table updating means 9 To update the correction table (S14). When the correct answer category does not exist in the recognition candidate category group, for example, a value obtained by adding a predetermined constant α (for example, 200) to the distance value of the lowest recognition candidate category in the recognition candidate category group is set to the correct answer category. The distance value.

The table updating means 9 first subtracts the distance value of the correct answer category from the distance value of the misread category notified from the result correcting means 6 to obtain the distance difference. And
A set 50 including the distance difference, the correct answer category, and the misread category is additionally registered in the correction table 5.

At this time, the table updating means 9 registers a new set 50 in order, for example, from the beginning of the correction table 5, and when the set 50 is registered up to the end of the table and is full (that is, (In the previous example, when 200 sets have been registered), the registration is returned to the top of the table again and the oldest registered set 50 is overwritten for additional registration. By limiting the number of groups to be registered in this way to, for example, about 200 and cyclically using the area of the table 5 after registering 200, it is possible to prevent an increase in the memory required for the correction table 5.

Further, in the present embodiment, the table updating means 9 determines that a set including the same misreading category and correct answer category as that of the set registered this time is already registered in the correction table 5. , The current distance difference is added to the registered distance difference.

Next, this embodiment will be described in more detail with reference to specific examples.

In the initial state in which the contents of the correction table 5 are empty as shown in FIG. 3A, a certain character handwritten on a form or the like is recognized, and the recognition means 2 detects the character.
It is assumed that the recognition candidate category group and the distance value as shown in (b) are output. That is, the first candidate is “width”, the second candidate is “yu”, the third candidate is “ke”, the fourth candidate is “ya”, and the fifth candidate is “sa”. Distance values of 1300, 1500, 1600, 1700, 1800
Assume that At this time, in the candidate correction means 4, no set is registered in the correction table 5, so that FIG.
The distance value is not corrected because the search is immediately determined to be completed in step S3. Therefore, the new recognition candidate category group sorted by the distance value in step S10 and the distance value are the same as those in FIG. 3B as shown in FIG. Therefore, the "width" of the first rank candidate is displayed on the display means 7 as the recognition result, and the candidates of the second rank and below are displayed (S11). When the operator judges that the original input character is "Yu" because it is an erroneous reading and selects the second-ranked candidate "Yu" displayed on the display unit 7, the recognition result is " Corrected from "width" to "yu" (S
13). Also, the misread category "width" and its distance value 13
00, the correct category “Yu” and the distance value 1500 thereof are notified to the table updating means 9. In the table updating means 9, 200, which is obtained by subtracting the distance value 1300 of the misreading category “width” from the distance value 1500 of the correct answer category “yu”, is set as the distance difference 53, and the misreading category “width” and the correct answer category “yu” are added. The set 50 is registered in the correction table 5 (S14). As a result, the contents of the correction table 5 are shown in FIG.
(D).

Next, the contents of the correction table 5 are shown in FIG.
When in the state shown in (a) (same as in FIG. 3D),
Recognition of another certain character is performed, and the recognition means 2 to FIG.
It is assumed that the recognition candidate category group and the distance value as shown in (b) are output. That is, the first candidate is “width”, the second candidate is “ke”, the third candidate is “sa”, the fourth candidate is “H”, and the fifth candidate is “yu”. Distance values of 1100, 1150, 1200, 1400, 1400
Assume that At this time, the candidate correction means 4 uses the first
There is one set in the correction table 5 that includes the misreading category “width” of the same category as the rank candidate “width”. Therefore, according to the correct answer category “yu” and the distance difference 200 of that set, the recognition candidate “yu” is selected. Change the distance value from 1400 to 1200.
Therefore, the new recognition candidate category group and distance values sorted by distance value in step S10 are as shown in FIG. 4 (c). Next, the display means 7 displays the "width" of the first rank candidate as a recognition result, and also displays the second rank and lower rank candidates (S11). Here, since the original input character is "mi", the operator judges that it is a misreading, and inputs "mi" from the correction information input means 8, the result correcting means 6 recognizes the recognition result from "width" to "mi". (S13). Further, the table updating means 9 is notified of the misreading category “width” and its distance value 1100, and the correct answer category “mi” and its distance value. Here, since the correct answer category “mi” does not exist in the recognition candidate category group, the distance value 1400 of the lowest candidate “yu” is calculated.
Is added to 200 (= α) to obtain 1600 as the distance value of the correct answer category “mi”. The table updating means 9 subtracts the distance value 1100 of the misreading category “width” from the distance value 1600 of the correct answer category “Mi” to obtain a distance difference 53 of 500, and adds the misreading category “width” and the correct answer category “Mi”. The set 50 is registered in the correction table 5 (S1
4). As a result, the contents of the correction table 5 are shown in FIG.
become that way.

Next, the contents of the correction table 5 are shown in FIG.
When in the state shown in (a) (the same as in FIG. 4D),
Another certain character (here, "Yu", which is also the previous "Yu")
It is assumed that the handwriting is more likely to be “smooth” as compared with the above), and the recognition unit 2 outputs the recognition candidate category group and the distance value as shown in FIG. 5B. That is, the first candidate is “width”, the second candidate is “yu”, the third candidate is “ke”, the fourth candidate is “sa”, and the fifth candidate is “ya”. Distance value of 120
It is assumed that they are 0, 1350, 1500, 1800, and 1900. At this time, the candidate correction means 4 corrects the distance value because there are two sets in the correction table 5 including the misreading category “width” of the same category as the first-ranked candidate “width”. First, one set of correct category "Yu" and distance difference 2
00, the distance value of the recognition candidate “Yu” is changed from 1350 to 1150. Next, the recognition candidate “mi” is added according to the remaining one set of correct category “mi” and the distance difference 500, and the distance value thereof is set to 1600 (= 1900 +).
200-500). Therefore, the new recognition candidate category group and distance values sorted by distance value in step S10 are as shown in FIG. 5 (c). Next, the display means 7 displays the first candidate "Yu" as the recognition result.

As described above, when the misreading is performed under the condition that the distance difference is as small as 200, and the pair "width, yu" of the same category occurs again under the condition that the distance difference is as small as 150, it is possible to use The category "Yu", which was corrected as the correct answer, is ranked higher. In other words, it indicates that a character that is more similar to a character that is a recognition target at the time of registering a set can be a first-ranked candidate.

Next, the contents of the correction table 5 are shown in FIG.
In the state shown in (a) (same as in FIG. 4 (d) and FIG. 5 (a)), another certain character (here, "width" is assumed) is recognized, It is assumed that the recognition means 2 outputs the recognition candidate category group and the distance value as shown in FIG. 6B. That is, the first candidate is “width”, the second candidate is “ke”, the third candidate is “sa”, the fourth candidate is “yu”, and the fifth candidate is “ya”. Distance value of 120
It is assumed that they are 0, 1350, 1400, 1650 and 1800. At this time, the candidate correction means 4 corrects the distance value because there are two sets in the correction table 5 including the misreading category “width” of the same category as the first-ranked candidate “width”. First, one set of correct category "Yu" and distance difference 2
00, the distance value of the recognition candidate “Yu” is changed from 1650 to 1450. Next, a recognition candidate “mi” is added according to the remaining one set of correct answer category “mi” and the distance value 500, and the distance value is set to 1500 (= 1800 +).
200-500). Therefore, the new recognition candidate category group and distance values sorted by distance value in step S10 are as shown in FIG. 6 (c). Next, the display means 7 displays the “width” of the first candidate as the recognition result.

In this way, when the misreading is performed under the condition that the distance difference is as small as 200, and the pair "width, yu" of the same category occurs again under the condition that the distance difference is as large as 450, the recognizing means 2 determines. The recognition result is maintained.

Next, the contents of the correction table 5 are shown in FIG.
When in the state shown in (a) (the same as in FIG. 4D),
Another certain character (here, "mi", which is also the previous "mi"
It is assumed that the handwriting is somewhat “like” compared to the above) and the recognition means 2 outputs the recognition candidate category group and the distance value as shown in FIG. 7B. That is, the first candidate is “width”, the second candidate is “ke”, the third candidate is “sa”, the fourth candidate is “H”, and the fifth candidate is “ya”. Distance value of 120
It is assumed that they are 0, 1250, 1300, 1400, and 1450. At this time, the candidate correction means 4 corrects the distance value because there are two sets in the correction table 5 including the misreading category “width” of the same category as the first-ranked candidate “width”. First, one set of correct category "Yu" and distance difference 2
00, the recognition candidate “Yu” is added, and the distance value is set to 1450 (= 1450 + 200−200). Next, the remaining one set of correct category “M” and distance value 50
In accordance with 0, the recognition candidate “mi” is added, and the distance value is set to 1150 (= 1450 + 200−500). Therefore, the new recognition candidate category group and distance value sorted by distance value in step S10 are as shown in FIG. 7 (c). Next, the display means 7 displays the first candidate "mi" as a recognition result.

As described above, when the misreading is performed under the condition that the distance difference is large, such as 500, and the group "width, mi" of the same category appears again, even if the distance difference is large, the lower candidates come to the upper position. It is possible.

Next, the contents of the correction table 5 are shown in FIG.
When in the state shown in (a) (the same as in FIG. 4D),
Another certain character (here, "yu", which is also assumed to be handwritten that does not look like "yu" more than "yu" at the time of correction) is recognized, and the recognition means 2 From Figure 8
It is assumed that the recognition candidate category group and the distance value as shown in (b) are output. That is, the first candidate is “width”, the second candidate is “yu”, the third candidate is “ke”, the fourth candidate is “ya”, and the fifth candidate is “sa”. Distance values of 1300, 1550, 1600, 1700, 1800
Assume that At this time, the candidate correction means 4 uses the first
Since there are two sets in the correction table 5 including the misreading category “width” of the same category as the position candidate “width”, the distance value is corrected. First, the distance value of the recognition candidate “Yu” is set to 15 according to one set of correct category “Yu” and the distance difference 200.
Change from 50 to 1350. Next, a recognition candidate "mi" is added according to the remaining one set of correct category "mi" and the distance value 500, and the distance value is set to 1500 (= 1800).
+ 200-500). Therefore, the new recognition candidate category group and distance values sorted by distance value in step S10 are as shown in FIG. 8 (c). Next, the display means 7
Displays the "width" of the first candidate as the recognition result. When the operator judges that the original character is "Yu" because the original character is "Yu" and the second candidate "Yu" displayed on the display unit 7 is selected, the recognition result is "width" by the result correction unit 6. Is corrected to "Yu" (S13). Also, the misread category "width" and its distance value 1300, and the correct category "yu"
And its distance value 1350 are notified to the table updating means 9. In the table updating means 9, the correct answer category "Yu"
From the distance value 1350 of the misreading category "width" to the distance value 13
50 obtained by subtracting 00 is set as the distance difference 53, and the set 50 in which the misreading category “width” and the correct answer category “Yu” are added is registered in the correction table 5 (S14). At this time, since the registered table {width, yu, 200} having the same misread category “width” and the same correct answer category “yu” exists in the correction table 5, the distance difference 200 is the distance difference 50 of this time. Is added. As a result, the contents of the correction table 5 are shown in FIG.
become that way. The registered set having the same misread category and correct answer category may not be deleted, but the current one may be registered separately. That is, in the case of the above example, the set {width, yu, 200} is left while the set {width, yu, 200} is left.
50} may be additionally registered.

As described above, when the misreading is performed under the condition that the distance difference is as small as 200, and the group "width, yu" of the same category appears with a larger distance difference, the category "corrected" is corrected. "Yu" cannot necessarily be the first candidate. However, since the distance difference in the correction table 5 is changed by correcting again, the distance difference that can be absorbed can be gradually increased. Therefore, as shown in the following example, when reading "yu" which is more "yu" than "yu", "yu" is output again as a recognition result.

That is, the contents of the correction table 5 are shown in FIG.
When in the state shown in FIG. 8A (same as FIG. 8D),
"Yu" that is more or less "Yu" is recognized as compared with "Yu" assumed in FIG. 8, and the recognition means 2 outputs the recognition candidate category group and the distance value as shown in FIG. 9B. Suppose That is, the first candidate is “width”, the second candidate is “yu”, the third candidate is “ke”, the fourth candidate is “ya”, and the fifth candidate is “sa”. Has a distance value of 130
It is assumed that they are 0, 1500, 1600, 1700 and 1800. At this time, the candidate correction means 4 corrects the distance value because there are two sets in the correction table 5 including the misreading category “width” of the same category as the first-ranked candidate “width”. First, a set of correct category "Yu" and distance value 2
50, the distance value of the recognition candidate “Yu” is changed from 1500 to 1250. Next, a recognition candidate “mi” is added according to the remaining one set of correct answer category “mi” and the distance value 500, and the distance value is set to 1500 (= 1800 +).
200-500). Therefore, the new recognition candidate category group and distance values sorted by distance value in step S10 are as shown in FIG. 9C, and the first candidate "Yu"
Is displayed as the recognition result.

FIG. 10 is another operation flowchart of the embodiment shown in FIG. This operation flowchart differs from that of FIG. 2 in that the first-ranked candidate category is held in step S15, and after the processing of steps S2 to S10 (step S16) in FIG. 2, the first-ranked candidate category is changed to step S15. It is checked whether it is different from the candidate category held in step S17 (step S17). If it is different (YES in step S17)
S) is that the processes of steps S2 to S10 are performed again (step S18). The candidate correction means 4 shown in FIG.
The processing of steps S15 to S18 of 0 is executed.

Next, a specific example will be described in more detail.

When the content of the correction table 5 is in the state shown in FIG. 11A (same as FIG. 4A), another certain character (here, "width" is assumed) is displayed. It is assumed that the recognition is performed and the recognition means 2 outputs the recognition candidate category group and the distance value as shown in FIG. That is, the first candidate is “width”, the second candidate is “yu”, the third candidate is “sa”, the fourth candidate is “H”, and the fifth candidate is “ya”. Distance values of 1100, 1150, 120
It is assumed that they are 0, 1400, and 1400. First, the candidate correction means 4 stores the first-ranked candidate "width" (step S15). Next, a set including the misreading category “width” in the same category as the first-ranked candidate “width” is set to 1 in the correction table 5.
Therefore, the distance value of the recognition candidate “Yu” is set to 1150 according to the correct category “Yu” and the distance difference 200 of the set.
Change from 950 to 950. Therefore, the new recognition candidate category group and the distance value are as shown in FIG. 11C (step S16). Next, it is checked whether the new first-ranked candidate "Yu" is different from the stored candidate (in this case, "width") (step S17). In this case, since it is different, a set including a misreading category of the same category as the first-ranked candidate “Yu” is searched for in the correction table 5, but it is not corrected because it does not exist (step S18). Next, the display means 7 displays the first-ranked candidate "Yu" as the recognition result, and also displays the second-ranked candidates and below (S11). Here the original input character is "width"
Therefore, when the operator judges that the reading is erroneous and inputs "width" from the correction information input means 8, the result correction means 6 corrects the recognition result from "yu" to "width" (S13). The table updating means 9 registers the misreading category "Yu", the correct answer category "Width", and the distance difference 150 (= 1100-950) in the correction table 5 (S14). As a result, the contents of the correction table 5 are as shown in FIG.

Next, the contents of the correction table 5 are shown in FIG.
In the state shown in FIG. 11A (same as FIG. 11D), another certain character (here, "width" is assumed).
Is recognized, and the recognition means 2 outputs a recognition candidate category group and a distance value as shown in FIG. That is, the first candidate is "width", the second candidate is "yu",
The third candidate is “ke”, the fourth candidate is “sa”, the fifth candidate is “ya”, and the distance values are 1200 and 135, respectively.
It is assumed that they are 0, 1500, 1800 and 1900. First, the candidate correction means 4 stores the first-ranked candidate "width" (step S15). Next, since there is one set in the correction table 5 that includes the misreading category “width” of the same category as the first-ranked candidate “width”, the distance value of the recognition candidate “yu” is changed from 1350 to 1150 according to the distance difference 200. Change to. Therefore, the new recognition candidate category group and the distance value are as shown in FIG. 12C (step S16). Next, it is checked whether the new first-ranked candidate "Yu" is different from the stored candidate (in this case, "width") (step S17).
In this case, since it is different, a set including the misread category of the same category as the first-ranked candidate “Yu” is searched for in the correction table 5 and 1
Since there are two, there is a recognition candidate “width” according to the distance difference 150.
Change the distance value of from 1200 to 1050. Therefore,
The new recognition candidate category group and distance value are shown in FIG.
(Step S18).

As described above, the group "Yu" and "Width" which are easily misread
If both are recognized many times, it is easy for the correction process to be misread by the opposite characters, but if the correction process is performed twice, the result of the first recognition process is emphasized and the recognition rate is reversed. It is possible to prevent the phenomenon that it lowers.

FIG. 13 is a block diagram of another embodiment of the character recognition device to which the present invention is applied. The character recognition device of this embodiment is different from that of the embodiment of FIG. 1 in that a distance difference 53 and a recognition feature 54 are used as correction information. Due to this difference, the recognition means 2 ', the candidate correction means 4', the correction table 5 ', and the table update means 9'are changed from the corresponding means shown in FIG. 1 as follows.

The correction table 5'has a misreading category 51.
Has a capacity capable of storing a plurality of sets 50 ′ including the correct answer category 52, the distance difference 53, and the recognition feature 54.

The recognizing means 2'extracts the recognizing features from the character image image inputted from the data inputting means 1, compares the recognizing features with the standard features stored in the recognizing dictionary 3 in advance for each character category, and calculates the distance value. Multiple higher ranks (eg, 5th
The recognition candidate categories of up to five) and their respective distance values are obtained and output to the candidate correction means 4 ', and the recognition features extracted from the character image image are output to the table update means 9'.

The candidate correction means 4 ', instead of the above equation,
Correct the distance value using the following formula.

D ′ = D−w · s · a / d ... where D ′ is the corrected distance value, D is the uncorrected distance value,
w is a weighting coefficient, s is a distance difference in the correction information, d is a distance value between the recognition feature in the correction information and the recognition feature extracted from the current character image, and a normalizes the distance value d. It is a reference distance value that is a constant. The calculation of the distance value d between the recognition features may be the same as the calculation of the distance between the standard feature of each category of the recognition dictionary 3 in the recognition means 2 ′ and the recognition feature obtained from the input character image image, for example, the Euclidean distance. Or city block distance can be used.

The table updating means 9'is the result correcting means 6
When the misreading category and its distance value and the correct answer category and its distance value are notified from, the distance difference 53 is calculated in the same manner as the table updating means 9 in FIG.
, The recognition feature input from the recognition means 2 ′ is added to create correction information, and the misread category and the correct answer category are further added to this, and the set 50 to be registered in the correction table 5 ′ is added. 'Is generated.

The recognition feature registered in the correction table 5'may be a compression feature obtained by compressing the recognition feature. By doing so, the required capacity of the correction table 5'can be reduced.

Other functions and operations of the character recognition device of this embodiment are the same as those of the character recognition device of FIG. In the present embodiment, when the recognition feature 54 of the correction table 5 ′ and the recognition feature obtained from the input character are close to each other, and as a result, the distance value d is small, the distance value D is greatly reduced, and the corresponding candidate is ranked higher. Easier to come to. On the contrary, when the recognition feature 54 of the correction table 5 ′ and the recognition feature obtained from the input character are distant and the distance value d is large as a result, the distance value D does not decrease so much, and therefore the rank of the corresponding candidate is reduced. Indicates a characteristic that does not rise so much.

The embodiments of the present invention have been described above.
The present invention is not limited to the above embodiments, and various other additions and changes are possible. For example, although the input data obtained by the data input means is a character image image input by a scanner or the like, it may be online character data input by a tablet or voice data input by a microphone. Further, the misread category stored in the correction table may be the template number of the recognition dictionary. If the template number is used, the templates can be distinguished when the recognition dictionary holds a plurality of templates in the same category.

[0069]

As described above, according to the present invention, the recognition result correction information of the operator is used for the correction of the distance value in the subsequent recognition processing, and the category that has been misread once becomes the output of the recognition means. If it appears as a first-ranked candidate, the distance value of the recognition candidate category that matches the correct category registered in the correction table is appropriately reduced, leaving the possibility of jumping to the first-ranked candidate. It is possible to improve the recognition rate in which the candidate is the correct answer as compared with the conventional one.

When the first-ranked candidate is corrected by the correction process, the correction process is performed again, so that the original recognition result is returned to the character set in which the misread category and the correct answer category are mutually registered. Since the possibility of a decrease in the recognition rate is emphasized and reduced, the recognition rate can be improved as compared with the conventional case.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of a character recognition device to which the present invention is applied.

FIG. 2 is a flowchart showing an operation procedure of the character recognition device of the embodiment of FIG.

FIG. 3 is an operation explanatory diagram of the embodiment of FIG. 1, showing a situation in which a first set including correction information is registered in a correction table in an initial state due to correction of misreading. It is a figure.

FIG. 4 is a diagram for explaining the operation of the embodiment of FIG. 1, showing a situation in which a second set including correction information is registered in the correction table due to correction of misreading again. is there.

5 is an operation explanatory diagram of the embodiment in FIG. 1, and is a diagram showing an example in which a correct category is derived by a set including correction information registered in a correction table.

6 is an operation explanatory diagram of the embodiment in FIG. 1, and is a diagram showing an example in which a correct category is derived without being affected by a set including correction information registered in a correction table.

FIG. 7 is an operation explanatory diagram of the embodiment of FIG. 1, and is a diagram showing another example in which a correct category is derived by a set including correction information registered in a correction table.

FIG. 8 is an explanatory diagram of the operation of the embodiment of FIG. 1, in which the correct category cannot be derived even by the set including the correction information registered in the correction table, and a correction is made again for that. It is a figure which shows the condition which registered the group containing various corrective information in the corrective table.

FIG. 9 is a diagram for explaining the operation of the embodiment of FIG. 1, and is a diagram showing an example in which a correct answer category has been derived by another correction.

10 is another flowchart showing an operation procedure of the character recognition device in the embodiment of FIG.

FIG. 11 is an operation explanatory diagram according to the flow of FIG. 10 of the embodiment of FIG. 1, and is a diagram showing a situation in which a set in which the misread category and the correct answer category are reversed is registered in the correction table.

12 is an explanatory diagram of an operation according to the flow of FIG. 10 of the embodiment of FIG. 1, and is a diagram showing an example in which a correct category is derived by two correction processes.

FIG. 13 is a block diagram of another embodiment of the character recognition device to which the present invention is applied.

FIG. 14 is a block diagram showing a configuration example of a conventional character recognition device.

[Explanation of symbols]

 1 data input means 2, 2'recognition means 3 recognition dictionary 4, 4 'candidate correction means 5, 5' correction table 50, 50 'set 51 misreading category 52 correct category 53 distance difference 54 recognition feature 6 result correction means 7 Display means 8 Correction information input means 9'Table updating means

Claims (7)

[Claims] 1. A feature for recognition is extracted from data to be recognized,
Recognition that obtains a plurality of high-ranking recognition candidate categories having small distance values, which are the differences between the recognition characteristics and the standard characteristics, and the distance values thereof, by comparing with the standard characteristics of each category registered in advance in the recognition dictionary Means, a correction table for storing a set of misreading category, correct answer category, and correction information, and a set including a misreading category that matches the first-ranked recognition candidate category in the recognition candidate category group obtained by the recognition means. Is stored in the correction table, candidate correction means for reducing the distance value of the recognition candidate category matching the correct category in the set by a value according to the correction information in the set, and the candidate correction means. The first-ranked recognition candidate category in the recognition candidate category group after processing by the means is output as a recognition result, and according to the correction information input from the correction information input means. And a table updating unit for adding a set of a misread category, a correct category, and correction information to the correction table when the result correction unit corrects the recognition result. Recognizing device candidate correction method. 2. The candidate correcting means sets a predetermined distance value that matches the correct answer category when a recognition candidate category matching the correct answer category does not exist in the recognition candidate category group obtained by the recognizing means. 2. The configuration is such that the distance value is corrected after adding the given recognition candidate category to the recognition candidate category group.
Candidate correction method for the described recognition device. 3. The table updating means uses, as the correction information, a distance difference which is a difference between a distance value of a misreading category and a distance value of a correct answer category.
Candidate correction method for the described recognition device. 4. The candidate correcting means sets the corrected distance value D ′ as D ′ = D, where D is a distance value before correction, w is a weighting coefficient, and s is a distance difference which is correction information. 4. The candidate correction method for a recognition device according to claim 3, wherein the candidate correction method is obtained by −w · s. 5. The table updating means uses, as the correction information, a distance difference which is a difference between a distance value of a misreading category and a distance value of a correct answer category, and a recognition feature extracted from recognition target data. The candidate correction method for a recognition device according to claim 2, wherein: 6. The candidate correction means recognizes a distance value before correction, D, a weighting coefficient, w, a distance difference in the correction information, s, a recognition feature in the correction information, and recognition target data at this time. If the distance value to the feature for use is d, and the reference distance value that is a constant for normalizing the distance value d is a, the corrected distance value D ′ is given by D ′ = D−w · s · a / d The candidate correction method of the recognition device according to claim 5, wherein the candidate correction method is obtained. 7. The candidate correction means, when the first recognition candidate category in the recognition candidate category group obtained by the recognition means is changed by the candidate correction using the correction table, a new first correction candidate 7. The candidate correction method for a recognition apparatus according to claim 1, wherein the candidate correction method is configured to perform the candidate correction again by using the position candidate category.