JPH0512493A - Character recognizing device - Google Patents

Abstract

PURPOSE:To simultaneously reduce the number of misreading and the number of abandonment by using plural character classifying means at the same time. CONSTITUTION:Codes Ci (d1(i) and d2(i)) are generated from a distance value to the first candidate character code Ci of a first classifying means 14 and the distance value of a second character classifying means 15 to the code Ci, and codes Cj (d1(j), d2(j)) are generated from the distance value of the means 15 to the first candidate character code Cj and the distance value of the means 14 to the code Cj. Then, evaluation values Ei and Ej to the candidates Ci and Cj are generated from a first main axis, which is calculated from the distribution of vectors to the learning pattern of each candidate according to a main component analyzing method, and the codes Ci (di(i), d2(i)) and Cj (d1(j), d2(j)), and a pair judging means 16 judges which candidate is adopted and abandonned according to those evaluation values Ei and Ej. In this case, even when the first candidate character codes of the both means are equal, the abandonment is judged by using the evaluation values calculated by the method at the pair judging means 16.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device which uses two or more kinds of character classification means in combination.

[0002]

2. Description of the Related Art In a conventional character recognition device using a pattern matching method, the similarity between an unknown character and a standard character stored in a recognition dictionary is calculated, and the maximum similarity is set in advance for that standard character. If the difference is greater than or equal to the threshold and the difference between the maximum similarity and the next largest similarity is equal to or greater than another preset threshold, the unknown character outputs the recognition result that it is the same character type as the standard character showing the maximum similarity. To do. By appropriately setting these two thresholds for each character type, it is possible to suppress the number of unknown characters that are misread and reduce the number of unknown characters that cannot be recognized. (Japanese Patent Publication No. 58-178484).

There is also known a character recognition device which uses a plurality of character classification means in combination and uses a threshold value as described above (Japanese Patent Application No. 3-526). FIG. 7 shows the outline of the determination algorithm in this device. According to this, when the first candidates of the classification results output by the plurality of types of classification means are different, the evaluation value is calculated for the first candidates of the classification results output by the respective classification means, which It is determined whether or not the first candidate of the classification result output by the classification unit is to be the final recognition result. At this time, a threshold value is set for the evaluation value for the first candidate of the classification result output by each classification means, and the first candidate for the classification result exceeding this is rejected.

[0004]

However, in the device for preventing the deterioration of the recognition rate by using the above-mentioned plurality of character classification means together, the first candidates of the classification results output by the respective classification means are the same. If there is, the decision is not made and the candidate is immediately set as the final recognition result.
In this case, misreading is likely to occur. Therefore, even if the first candidates of the classification results output by the respective classification means are the same, it is necessary to improve the recognition accuracy by determining whether to use the first candidate as the final recognition result or to reject it.

An object of the present invention is to provide a character recognition device capable of reducing the number of unknown characters that are misread or rejected.

[0006]

The character recognition device of the first invention has two types of character classification means, and outputs the final recognition result from the character code of the first candidate selected by each classification means. In the character recognition device, (1) the first candidate character code Ci selected by the first classification means and its distance value d1 (i), and the distance value d2 with respect to the character code Ci in the second classification means.
From (i), the two-dimensional vector Ci (d1 (i), d2
(I)) generating means, (2) the first candidate character code Cj selected by the second classifying means and its distance value d2 (j), and the character code Cj in the first classifying means. Distance value d1 to
From (j), the two-dimensional vector Cj (d1 (j), d2
(J)) generating means, and (3) the two-dimensional vector Ci (d1 (i), d2
(I)) and Cj (d1 (j), d2 (j)), the means for calculating the evaluation value Ei for the character code Ci and the evaluation value Ej for the character code Cj are the character types to be classified. Learning character pattern Dk belonging to K
Two-dimensional vector Ck (d1 on a plane having coordinate values of d1 (k) and d2 (k), which are distance values to
The main axis of the distribution of (k) and d2 (k) is Vk (p1
(K), p2 (k))) is obtained from the maximum eigenvalue of a two-dimensional variance-covariance matrix obtained by using the principal component analysis method with the elements d1 (k) and d2 (k) of each vector as variables. The evaluation values Ei and Ej are determined by determining the evaluation value Ek for the character code Ck from Vk (p1 (k), p2 (k)) and the two-dimensional vector Ck (d1 (k), d2 (k)). And (4) when the character code Ci and the character code Cj are the same, the character code Ci is calculated using the evaluation value Ei.
Is output as the final recognition result or is rejected, or the character code Ci and the character code C are determined.
When j is different from the above, it is characterized by comprising means for making a pair decision by using the evaluation value Ej which character code is to be output as the final recognition result or both are to be rejected.

The character recognition device of the second invention is a character recognition device which has two types of character classification means and outputs a final recognition result from the character code of the first candidate selected by each classification means, 1) The first candidate character code Ci selected by the first classifying means and its distance value d1 (i), and the distance value d2 with respect to the character code Ci in the second classifying means.
From (i), the two-dimensional vector Ci (d1 (i), d2
(I)) generating means, (2) the first candidate character code Cj selected by the second classifying means and its distance value d2 (j), and the character code Cj in the first classifying means. Distance value d1 to
(J) and the two-dimensional vector Cj (d1 (j), d2
(J)) generating means, and (3) the two-dimensional vector Ci (d1 (i), d2
(I)) and Cj (d1 (j), d2 (j)), the evaluation value Ei for the character code Ci and the character code C
In the means for calculating the evaluation value Ej for j, the distance values d1 (k) and d2 (k), which are distance values for the learning character pattern Ck belonging to the character type k to be classified, are coordinate axes. Two-dimensional vector Ck (d1
The main axis of the distribution of (k) and d2 (k) is Vk (p1
(K), p2 (k))) is calculated from the maximum eigenvalue of a two-dimensional correlation matrix obtained by using the principal component analysis method with the elements d1 (k) and d2 (k) of each vector as a variate.
(P1 (k), p2 (k)) and the two-dimensional vector Ck
Means for calculating the evaluation values Ei and Ej by determining the evaluation value Ek for the character code Ck from (d1 (k), d2 (k)), and (4) the character code Ci and the character code Cj. The same character code Ci using the evaluation value Ei
Is output as the final recognition result or is rejected, or the character code Ci and the character code C are determined.
When j is different, the evaluation value Ei and the evaluation value E
It is characterized by comprising a means for making a pair determination by using j and which character code is to be output as the final recognition result or which is to be rejected.

The character recognition device of the third invention has three or more kinds of all n kinds of character classification means, and outputs the final recognition result from the first candidate character code selected by each classification means. In the apparatus, (1) the first candidate character code Ch selected by the m-th classification means for m not exceeding n and its distance value d
An n-dimensional vector Ch (d1) in which m (h) and distance values with respect to the character code Ch in the n-1 number of classification means other than the m-th one are arranged in the order of the classification means
(H), d2 (h) ,. ． ． , Dm (h) ,. ． ． , D
(2) the n-dimensional vector Ch (d1 (h), d2) for generating n (h)) for the first candidate of all classification means;
(H) ,. ． ． , Dm (h) ,. ． ． , Dn (h)), the means for calculating an evaluation value for the character code of the first candidate selected by each of the n total classification means,
A learning character pattern C belonging to the character type k to be classified
n-dimensional vector Ck (d1 (k), d2 obtained from h
(K) ,. ． ． , Dm (k) ,. ． ． , Dn (k)) is the n-dimensional vector C in the space with the coordinate axes.
k (d1 (k), d2 (k), ..., dm
(K) ,. ． ． , Dn (k)) main axis Vk (p1
(K), p2 (k) ,. ． ． , Pm (k) ,. ． ． , P
n (k)) is the element d of each vector using the principal component analysis method
Vk (p1 (k), p2 is obtained from the maximum eigenvalue of the n-dimensional variance-covariance matrix obtained by using m (k) as a variable
(K) ,. ． ． , Pm (k) ,. ． ． , Pn (k)) and the n-dimensional vector Ck (d1 (k), d2
(K) ,. ． ． , Dm (k) ,. ． ． , Dn (k)) and means for calculating the evaluation value Ek for the character code Ck, and (3) first candidate character selected by all classification means up to the n-th When the codes are the same, it is determined whether the candidate character code is selected as the final recognition result or rejected using the evaluation value for the first candidate character code, or the classification means up to the n-th When the character code of the first candidate of the classification result is different, it is determined which candidate is to be output as the final recognition result or both are rejected from the respective evaluation values for all n character codes of the first candidate. And means.

A character recognition apparatus according to a fourth aspect of the present invention has three or more kinds of all n kinds of character classifying means, and outputs a final recognition result from the first candidate character code selected by each classifying means. In the apparatus, (1) the first candidate character code Ch selected by the m-th classification means for m not exceeding n and its distance value d
An n-dimensional vector Ch (d1) in which m (h) and distance values with respect to the character code Ch in n-1 classification means other than the m-th one are arranged in the order of the classification means.
(H), d2 (h) ,. ． ． , Dm (h) ,. ． ． , D
(2) the n-dimensional vector Ch (d1 (h), d2) for generating n (h)) for the first candidate of all classification means;
(H) ,. ． ． , Dm (h) ,. ． ． , Dn (h)), the means for calculating an evaluation value for the character code of the first candidate selected by each of the n total classification means,
A learning character pattern C belonging to the character type k to be classified
n-dimensional vector Ck (d1 (k), d2 obtained from k
(K) ,. ． ． , Dm (k) ,. ． ． , Dn (k)) is the n-dimensional vector C in the space with the coordinate axes.
k (d1 (k), d2 (k), ..., dm
(K) ,. ． ． , Dn (k)) main axis Vk (p1
(K), p2 (k) ,. ． ． , Pm (k) ,. ． ． , P
n (k)) is the element d of each vector using the principal component analysis method
Vk (p1 (k), p2 is obtained from the maximum eigenvalue of the n-dimensional correlation matrix obtained by using m (k) as a variable.
(K) ,. ． ． , Pm (k) ,. ． ． , Pn (k)) and the n-dimensional vector Ck (d1 (k), d2
(K) ,. ． ． , Dm (k) ,. ． ． , Dn (k)) and means for calculating the evaluation value Ek for the character code Ck, and (3) first candidate character selected by all classification means up to the n-th When the codes are the same, it is determined whether the candidate character code is selected as the final recognition result or rejected using the evaluation value for the first candidate character code, or the classification means up to the n-th When the character code of the first candidate in the classification result is different, it is determined which candidate is to be output as the final recognition result or both are rejected from the respective evaluation values for all n character codes of the first candidate. And means.

[0010]

The operation of the present invention will be described by taking as an example the case where two kinds of classification means shown in FIG. 3 are used together.

The evaluation value Ei for the first candidate character code Ci of the classification result output by the first classifying means and the evaluation value Ej for the first symptom character code Ci of the classification result output by the second classifying means are calculated. In the means for learning, for each character type k to be classified, a learning character pattern Ck belonging to the character type k is placed on a plane whose axes are distance values of the first classifying means and the second classifying means. When the obtained distance values d1 (k) and d2 (k) of the first classifying means and the second classifying means are plotted, they are distributed like the inside of the closed curve S in the figure. The principal axis Vk (p1 (k), which represents the slope of this distribution,
p2 (k)) is as shown in the same figure, and is obtained by the principal component analysis method.

A straight line which is orthogonal to the main axis Vk and which is in contact with the above-described distribution S is set with respect to the main axis Vk. Two-dimensional vector Ck of candidate character code Ck
When (d1 (k), d2 (k)) is closer to the origin than this, it is assumed that Ck belongs to the character type k. Seed k
Is determined to belong to. Also, the two-dimensional vector C'k (d1 'for the candidate character code C'k indicated by the triangle mark in the figure
Since (k) and d2 ′ (k) are outside the shaded area, they are not judged to belong to the character type k and are discarded. Also E
In this example, kmax is a value on the main axis Vk with respect to the intersection of the above-described straight line L and the main axis Vk previously obtained, and whether or not it can be considered that the candidate character code Ck belongs to the character type k. Use as a judgment threshold.

Consider the case where the threshold values are set individually for the first classifying means and the second classifying means, and the threshold values for the first classifying means and the second classifying means at that time are t1 and
Let t2. If t1 and t2 are the maximum distance values for the learning pattern, the area that does not exceed either of them is a rectangular area surrounded by the vertical axis, horizontal axis and one-dot chain line, two-dot chain line in the figure, Since the rejection area of is set, the recognition rate decreases. As is clear from this comparison, according to this method, the area that is not rejected can be wide,
If the first candidate character code of one of the classification methods outputs a distance value that is very close, it is less than or equal to the threshold,
It is possible to determine that character type.

The first candidate character code C of the first classification means
i and the first candidate character code Cj of the second classifying means, in the means for determining which should be output or rejected, the first candidate character code Ci, Cj of the respective classifying means. Evaluation values Ei, E calculated for
If both the threshold values Eimax and Ejmax are exceeded for j, the first candidate character codes Ci and Cj output by the first classifying unit and the second classifying unit are rejected. In other cases, the smaller evaluation value is closer to the origin, so when the smaller evaluation value does not exceed the judgment threshold for that character type, the character code of the first candidate for that evaluation value is set. As the final recognition result, if not, the first candidate character codes Ci and Cj output by the first classifying means and the second classifying means are rejected. When the character code Ci of the first candidate of the first classifying means and the character code Cj of the first candidate of the second classifying means are the same, an evaluation value is calculated for the corresponding character code and If the judgment threshold for the character type is not exceeded, the character code is used as the final recognition result, and conversely, if the judgment threshold is exceeded, the character code is rejected. As a result, rejection judgment can be performed and misreading can be reduced even when the classification methods used together output the same candidate.

The same effect is obtained when the three kinds of classification means are used together.

[0016]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention.

The character pattern input means 1 is a means for taking in a character image on a form, and may be an ordinary image scanner. The captured character images are classified by the first classifying means 14 and the second classifying means 15. The pair determination means 16 determines which of the first candidate character codes output by the first classification means 14 and the second classification means 15 is to be adopted as the final recognition result, or which is rejected. Then, the determination result is sent to the final recognition result output means 13. These character pattern input means 1 and first classification means 1
4, the control means 17 controls the second classification means 15, the pair determination means 16, and the final recognition result output means 13.

As the internal elements related to the classifying means 14 and 15, for example, the configurations shown in FIGS. 4 and 5 can be considered. As shown in FIG. 4, the first classification means includes the first character recognition means 2, the first classification dictionary memory 3 referred to by the first character recognition means 2, and the classification output by the first character recognition means 2. It is composed of a first classification result storage memory 4 for storing the results. Further, as shown in FIG. 5, the second classification means includes a second character recognition means 5, a second classification dictionary memory 6 referred to by the second character recognition means 5, and a second character recognition means 5. It is composed of a second classification result storage memory 7 for storing the output classification result. The first classification dictionary memory 3 and the second classification dictionary memory 6 store the feature amounts referred to by the first character recognition means 2 and the second character recognition means 5, respectively. Note that this feature amount is created in advance from a learning pattern. The first classifying means 14 and the second classifying means 15 calculate the distance between the standard and unknown character patterns for the classification target character type k stored in the respective classification dictionary memories, and calculate the distance in ascending order of distance. It is stored in the classification result storage memory 4 and the classification result storage memory 7 together with the corresponding character code.

As an internal element related to the pair determination means 16, for example, a configuration as shown in FIG. 6 can be considered.

The classification result integrating means 8 is the first classification means 1
When the processing of 4 and the second classification means 15 is completed, it is started by the control means 17.

The activated classification result integrating means 8 passes the route r1 from the classification result storage memory 4 to the first candidate character code Ci and the distance value d1 in the first classification means 14.
Extract (i). Further, the second classification result storage memory 7
The distance value d2 (i) corresponding to the character code Ci is extracted from. Then, Ci, d1 (i), and d2 (i) are converted into a two-dimensional vector format and stored in the integrated result storage register 10.

Similarly, the classification result integrating means 8 uses the route 2
The character code Cj and the distance value d2 (j) are extracted from the classification result storage memory 7 via. Further, the distance value d1 corresponding to the character code Cj from the first classification result storage memory 4
Extract (j). And Cj, d2 (j), d1
(J) is converted into a two-dimensional vector format and stored in the integrated result storage register 10.

The evaluation value calculation means 9 is the classification result integration means 8
After completion of the above process, the control means 17 activates the integrated result storage register 10 and the evaluation value calculation dictionary memory 11.
With reference to, the evaluation values Ei and Ej for the candidate character codes output by the first classifying unit 14 and the second classifying unit 15 are calculated by the following equations (1) and (2), for example.

Ei = d1 (i) p1 (i) + d2 (i) p2 (i) Ej = d1 (j) p1 (j) + d2 (j) p2 (j) The first classification means 14 and the second classification. When the first candidates of the classification result of the means 15 are the same, the evaluation value is calculated by using either the formula (1) or the formula (2).

In the evaluation value calculation dictionary memory 11, the principal axis Vk (p1 (k), obtained from the above-mentioned variance-covariance matrix,
p2 (K)) and the threshold value Ekmax for determination are stored for each character type k. Here, the calculation of the main axis Vk for each character type k uses a two-dimensional variance-covariance matrix when the variables d1 (k) and d2 (k) are used as variables. If the scales of the distance values output by the classifying unit 15 are significantly different, the correlation matrix may be used instead. As a result, it is possible to eliminate the effect of weighting the evaluation value of the dispersion means having the larger dispersion of the distance values when calculating the evaluation value, and the distance values output by the first classification means 14 and the second classification means 15 can be eliminated. Can be equally evaluated by the threshold determination means 12.

When the processing of the evaluation value calculation means 9 is completed, the threshold value judgment means 12 is started by the control means 17 and the first classification means 14 calculated by the evaluation value calculation means 9,
The comparison of the evaluation values Ei and Ej with respect to the respective candidate character codes obtained from the expressions (1) and (2) output from the second classification means 15 is stored in the evaluation value calculation dictionary memory 11. By comparing with a threshold value, it is determined which classification result is adopted or which classification result is rejected. An example of the algorithm will be described with reference to FIG. When the character codes of the candidates output by the first classifying unit 14 and the second classifying unit 15 are the same, the determination d
In 1, the evaluation value is calculated using one of the formula (1) or the formula (2). If they are not the same, the evaluation value is calculated for each first candidate using both Expression (1) and Expression (2). The evaluation value calculation means 9 performs these calculations. The threshold determination means 12 is
When the character codes of the first candidates output by the first classifying means 14 and the second classifying means 15 are not the same, the judgment d2
The following processing is performed. The evaluation values for the first candidate character codes output by the first classifying unit 14 and the second classifying unit 15 obtained by the evaluation value calculating unit 9 both exceed the threshold value of the evaluation value calculating dictionary memory 11. If so, the character code of either candidate is rejected. If not, the candidate character code that does not exceed the threshold is selected, and the determination d
According to 3, if the evaluation value is less than or equal to the other evaluation value,
The character code of the corresponding candidate is adopted as the final recognition result, and the character code is sent to the final recognition result output means 13. Otherwise, the character code of the corresponding candidate is rejected, and a preset rejection corresponding to the rejection is rejected. The code is sent to the final recognition result output means 13. If the candidate character codes output by the first classifying unit 14 and the second classifying unit 15 are the same, the process of determination d4 and below is performed. When the evaluation value obtained by the evaluation value calculation means 9 exceeds the threshold value of the evaluation value calculation dictionary memory 11, the candidate character code is rejected, and a preset rejection code corresponding to the rejection is set. It is sent to the final recognition result output means 13. If the threshold is not exceeded, the character code of the corresponding candidate is adopted as the final recognition result, and the character code is sent to the final recognition result output means 13.

The final recognition result output means 13 is the control means 1
The final recognition result output by the pair determination means 16 may be output as a character code in accordance with the command of 7.

The embodiment in which two types of classifying means according to the present invention are used together has been described above. The distance measure calculated by the first classifying means 14 and the second classifying means 15 is used in character recognition. It is possible to correspond to any distance measure or similarity measure, not limited to those commonly used,
It is also possible to mix them. Further, even when three or more kinds of classification means are used together, the device shown in this embodiment can be easily constructed. Each of these functions may be realized by means of either hardware or software.

[0029]

As described above, according to the present invention,
By using two or three or more types of character classification means together, the number of misread characters is reduced, and at the same time, the evaluation values calculated for the candidate character codes output by each classification means are compared with each other. By performing the rejection determination in the threshold determination process of, the number of rejected characters can be reduced compared to the case where the thresholds for rejection determination are individually set for the two types of character classification means. It is possible to improve. At this time, the number of erroneous readings can be further reduced as compared with the “character recognition method and apparatus” (Japanese Patent Application No. 3-526) described in the related art.

[Brief description of drawings]

FIG. 1 is a block diagram for explaining an embodiment of a character recognition device according to the present invention.

FIG. 2 is a flowchart showing the processing example of FIG.

FIG. 3 is a diagram for explaining the operation of this method.

FIG. 4 is a configuration example of a first character classification means 14.

5 is an example of a configuration of a second character classification means 15. FIG.

FIG. 6 is a configuration example of pair determination means 16.

FIG. 7 is a flowchart of a conventional method.

[Explanation of symbols]

1-character pattern input means 2, 5 character recognition means 3, 6 classification dictionary memory 4, 7 Classification result storage memory 8 Classification result integration means 9 Evaluation value calculation means 10 Integrated result storage register 11 Evaluation value dictionary memory 12 Threshold judgment means 13 Final recognition result output means 14, 15 character classification means 16 Pair judging means 17 Control means

Claims (4)

[Claims] 1. A character recognition device which has two types of character classification means and outputs a final recognition result from the character code of the first candidate selected by each of the classification means, (1) The first classification means The selected first candidate character code Ci and its distance value d1 (i), and the distance value d2 with respect to the character code Ci in the second classification means.
From (i), the two-dimensional vector Ci (d1 (i), d2
(I)) generating means, (2) the first candidate character code Cj selected by the second classifying means and its distance value d2 (j), and the character code Cj in the first classifying means. Distance value d1 to
From (j), the two-dimensional vector Cj (d1 (j), d2
(J)) generating means, and (3) the two-dimensional vector Ci (d1 (i), d2
(I)) and Cj (d1 (j), d2 (j)), the evaluation value Ei for the character code Ci and the character code C
In the means for calculating the evaluation value Ej for j, the learning character pattern C belonging to the character type K to be classified
A two-dimensional vector Ck (d1 on a plane having coordinate axes of d1 (k) and d2 (k), which are distance values with respect to k.
The main axis of the distribution of (k) and d2 (k) is Vk (p1
(K), p2 (k))) is obtained from the maximum eigenvalue of the two-dimensional variance-covariance matrix obtained by using the principal component analysis method with the elements d1 (k) and d2 (k) of each vector as variables. V
k (p1 (k), p2 (k)) and the two-dimensional vector C
The character code C from k (d1 (k), d2 (k))
means for calculating the evaluation values Ei and Ej by determining the evaluation value Ek for k; and (4) the character code using the evaluation value Ei when the character code Ci and the character code Cj are the same. Ci
Is output as the final recognition result or is rejected, or when the character code Ci and the character code Cj are different, which character code is finally recognized using the evaluation value Ej. A character recognition device comprising: a means for determining whether to output as a result or to reject either of them. 2. A character recognition device which has two types of character classification means and outputs a final recognition result from the character code of the first candidate selected by each of the classification means, (1) By the first classification means The selected first candidate character code Ci and its distance value d1 (i), and the distance value d2 with respect to the character code Ci in the second classification means.
From (i), the two-dimensional vector Ci (d1 (i), d2
(I)) generating means, (2) the first candidate character code Cj selected by the second classifying means and its distance value d2 (j), and the character code Cj in the first classifying means. Distance value d1 to
From (j), the two-dimensional vector Cj (d1 (j), d2
(J)) generating means, and (3) the two-dimensional vector Ci (d1 (i), d2
(I)) and Cj (d1 (j), d2 (j)), the evaluation value Ei for the character code Ci and the character code C
In the means for calculating the evaluation value Ej for j, the learning character pattern C belonging to the character type K to be classified
A two-dimensional vector Ck (d1 on a plane having coordinate axes of d1 (k) and d2 (k), which are distance values with respect to k.
(K), d2 (k)) distribution main axis Vk (p1 (k),
p2 (k))) is calculated from the maximum eigenvalue of a two-dimensional correlation matrix obtained by using the principal component analysis method as the elements d1 (k) and d2 (k) of each vector, and Vk (p1 (k),
p2 (k)) and the two-dimensional vector Ck (d1 (k),
d2 (k)) and the evaluation value Ek for the character code Ck
Means for calculating the evaluation values Ei and Ej by determining (4) the character code Ci using the evaluation value Ei when the character code Ci and the character code Cj are the same.
Is output as the final recognition result or is rejected, or when the character code Ci and the character code Cj are different, which character code is finally recognized using the evaluation value Ej. A character recognition device comprising: a means for determining whether to output as a result or to reject either of them. 3. A character recognition device which has three or more kinds of all n kinds of character classification means and outputs a final recognition result from the character code of the first candidate selected by each classification means, wherein (1) n The character code Ch of the first candidate selected by the m-th classifying means for m that does not exceed and the distance value d thereof
An n-dimensional vector Ch (d1) in which m (h) and distance values with respect to the character code Ch in n-1 classification means other than the m-th one are arranged in the order of the classification means.
(H), d2 (h) ,. ． ． , Dm (h) ,. ． ． , D
(2) the n-dimensional vector Ch (d1 (h), d2) for generating n (h)) for the first candidate of all classification means;
(H) ,. ． ． , Dm (h) ,. ． ． , Dn (h)), the means for calculating an evaluation value for the character code of the first candidate selected by each of the n total classification means,
A learning character pattern C belonging to the character type k to be classified
n-dimensional vector Ck (d1 (k), d2 obtained from h
(K) ,. ． ． , Dm (k) ,. ． ． , Dn (k)) is the n-dimensional vector C in the space with the coordinate axes.
k (d1 (k), d2 (k), ..., dm
(K) ,. ． ． , Dn (k)) main axis Vk (p1
(K), p2 (k) ,. ． ． , Pm (k) ,. ． ． , P
n (k)) is the element dm of each vector using the principal component analysis method
Vk (p1 (k), p2 is calculated from the maximum eigenvalue of the n-dimensional variance-covariance matrix obtained by using (k) as a variable.
(K) ,. ． ． , Pm (k) ,. ． ． , Pn (k)) and the n-dimensional vector Ck (d1 (k), d2
(K) ,. ． ． , Dm (k) ,. ． ． , Dn (k)) and means for calculating the evaluation value Ek for the character code Ck, and (3) first candidate character selected by all classification means up to the n-th If the codes are the same, it is determined whether to select the candidate character code as the final recognition result using the evaluation value for the first candidate character code or to reject it,
Or, when the character codes of the first candidates of the classification results for the n-th classification means are different, which candidate is output as the final recognition result from the respective evaluation values for the character codes of all the n first candidates, Alternatively, a character recognition device comprising a means for determining whether to reject either. 4. A character recognition device which has three or more kinds of all n kinds of character classification means and which outputs a final recognition result from the character code of the first candidate selected by each classification means, wherein (1) n The character code Ch of the first candidate selected by the m-th classifying means for m that does not exceed and the distance value d thereof
An n-dimensional vector Ch (d1) in which m (h) and distance values with respect to the character code Ch in the n−1 classification means other than the m-th one are arranged in the order of the classification means.
(H), d2 (h) ,. ． ． , Dm (h) ,. ． ． , D
(2) the n-dimensional vector Ch (d1 (h), d2) for generating n (h)) for the first candidate of all classification means;
(H) ,. ． ． , Dm (h) ,. ． ． , Dn (h)), the means for calculating an evaluation value for the character code of the first candidate selected by each of the n total classification means,
A learning character pattern C belonging to the character type k to be classified
n-dimensional vector Ck (d1 (k), d2 obtained from k
(K) ,. ． ． , Dm (k) ,. ． ． , Dn (k)) is the n-dimensional vector C in the space with the coordinate axes.
k (d1 (k), d2 (k), ..., dm
(K) ,. ． ． , Dn (k)) main axis Vk (p1
(K), p2 (k) ,. ． ． , Pm (k) ,. ． ． , P
n (k)) is the element d of each vector using the principal component analysis method
Vk (p1 (k), p2 is obtained from the maximum eigenvalue of the n-dimensional correlation matrix obtained by using m (k) as a variable.
(K) ,. ． ． , Pm (k) ,. ． ． , Pn (k)) and the n-dimensional vector Ck (d1 (k), d2
(K) ,. ． ． , Dm (k) ,. ． ． , Dn (k)) to calculate the evaluation value Ek for the character code Ck, and (3) a first candidate character selected by all classification means up to the n-th When the codes are the same, it is determined whether the candidate character code is selected as the final recognition result or rejected using the evaluation value for the first candidate character code, or the classification means up to the n-th When the character code of the first candidate in the classification result is different, it is determined which candidate is to be output as the final recognition result or both are rejected from the respective evaluation values for all n character codes of the first candidate. And a character recognition device.