JPH02300876A - Character recognizing device - Google Patents

Abstract

PURPOSE:To improve the correct answer rate in the recognition of characters by performing a learning operation in order to obtain a character pattern candidate ensuring the maximum degree of original resemblance at every group as the final character recognition result and deciding a subneural net including a character pattern having the highest assurance. CONSTITUTION:The subneural nets 13 - 32 sort the character pattern candidates for each group of similar characters like 'E, KO, YU' and 'HA, HE, MU', etc., which easily cause the misreading and performs a learning operation in order to obtain a character pattern candidate having the maximum degree of original resemblance as the character recognition result at every group based on each degree of the resemblance between an input character pattern and each character pattern candidate. Based on the learning result, the character pattern candidate having the highest assurance is decided at every group. Then a control neural net 12 decides one of those nets 13 - 32 that includes the character pattern candidate having the highest assurance. Consequently, the character recognizing efficiency is improved.

Description

Detailed Description of the Invention [Object of the Invention] (Field of Industrial Application) The present invention relates to a character recognition device using a neural network.

(Prior Art) In recent years, systems that use neural networks to perform character recognition, voice recognition, image information processing, etc. have been actively developed.

Generally known character recognition processing calculates the degree of similarity between an input character pattern and multiple character pattern candidates by pattern matching, and selects the character pattern candidate with the highest degree of similarity as the final character recognition result. Character recognition is performed by determining and outputting the result as follows. On the other hand, in a method that combines the conventional character recognition method with Bi-ralNet, if an error occurs in the recognition result of a certain input character pattern, the correct character pattern candidate will be used in the next character recognition for that human character pattern. Each arc load value in the neural network is corrected in a direction that yields.

Therefore, by employing this neural network (and character recognition processing), it is possible to stably perform good character recognition in response to changes in the characteristics of input character patterns.

(Problem solved by the invention; problem to be solved by the invention) However, although character recognition processing that employs such a neural network achieves a good recognition rate when recognizing handwritten digits, it is difficult to apply this to handwritten kana characters, etc. ．． There were a number of obstacles.

For example, in the case of handwritten digits, the input layer of the neural network consists of 10 units, each corresponding to the degree of similarity between the input character pattern and each digit from "0" to "9", and a total of 11 units, including dummies [units]. Only one unit is required. On the other hand, in the case of handwritten kana characters, the input layer has a total of 47 units, 46 units corresponding to the similarity between the input character pattern and each kana character from "A" to "N", and one dummy. Is required. Along with this, the number of intermediate layers used to be about 50 for numbers, but at least 100 or more were required for kana characters.

For the output layer, 10 are required for numbers, and 46 are required for kana characters.

Therefore, the scale is large, and in the case of handwritten kana characters, it takes a long time to learn because there are many similar characters such as "ko, ko, yu", and the recognition rate is not satisfactory.

The problem was that I couldn't get it.

The present invention is intended to solve these problems, and is a character recognition system that can perform character recognition with a high accuracy rate even for character groups that have many types of characters and many similar patterns, such as handwritten kana characters. The purpose is to provide equipment.

[Structure of the Invention] (Means for Solving the Problems) The character recognition device of the first invention is based on the degree of similarity between a human character pattern and a plurality of character pattern candidates. In a character recognition device that determines a character pattern candidate with the highest degree of certainty as a result of recognizing an input character pattern, each character pattern candidate is classified into multiple groups of similar characters, and based on each degree of similarity, each group is identified as originally similar. Learning is performed so that the appropriate character pattern candidate with the maximum degree is obtained as the final character recognition result.
Judgment results of multiple sub-neural networks each judge the character pattern candidate with the highest degree of certainty for each group, and a sub-neural network that includes character pattern candidates with the highest degree of similarity based on each similarity. It is equipped with a neural network that performs learning to obtain the final character recognition result and determines the sub-neural network that includes the character pattern candidate with the highest degree of certainty.

Further, in the above-described invention, a second invention is provided with a comparison means for comparing each degree of similarity obtained between the input character pattern and the plurality of character pattern candidates with a predetermined threshold, and this comparison means If the first output signal is output, the character pattern candidate with the maximum similarity is determined as the character recognition result, and if the comparison means outputs the second output signal,
This system is equipped with a means for passing the final character recognition result judgment process to the sub-neural network and the neural network.

(Function) In the character recognition device of the first invention, when each similarity between an input character pattern and a plurality of character pattern candidates is input to each subneural network and each neural network, each subneural large , within each group, determine the character pattern candidate with the highest degree of certainty.

On the other hand, the neural network determines the sub-neural network containing the character pattern candidate with the highest degree of certainty among the determination results output from each sub-neural network (2), thereby determining the final character recognition result.

In addition, during learning, each subneural network selects the appropriate character pattern candidate with the maximum similarity based on the degree of similarity between the input character pattern and multiple character pattern candidates within each group as the final character. Learning is performed to obtain the recognition result, and the neural network determines the final character recognition result as the judgment result of the sub-neural network that contains the character pattern candidate with the maximum similarity from each similarity. Learn how to do this.

Therefore, according to the present invention, it is possible to perform character recognition with a high accuracy rate even for a group of characters such as handwritten kana characters, which have many types of characters and have many similar patterns (-).

Further, in the second invention, each degree of similarity obtained between the input character pattern and the plurality of character pattern candidates is compared with a predetermined threshold value in the comparison means, and the first output is outputted from the comparison means. When a signal is output, the character pattern candidate with the maximum degree of similarity is determined as the character recognition result. Further, when the second output signal is output from the comparing means, the process of determining the final character recognition result is passed to the sub-neural network and the neural network.

According to this, since it is possible to narrow down and learn characters that require character recognition result determination processing using the sub-neural network and the neural network, it is possible to further improve the recognition rate.

(Example) Hereinafter, an example of the present invention will be described in detail based on the drawings.

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

In the figure, 1 is a scanner unit that optically reads kana characters etc. handwritten on a form to obtain the character pattern;
2 is a pattern candidate storage unit in which a group of character pattern candidates (in this embodiment, kana characters from “a” to “n” are essential) to be matched with the input character pattern read by the scanner unit 1; Reference numeral 3 denotes a pattern matching unit which performs predetermined pattern matching between the input character pattern from the scanner unit 1 and the character pattern candidates stored in the pattern candidate storage unit 2 to obtain a degree of similarity of 9, and 4 a pattern matching unit. A character determination unit determines the most probable character pattern candidate as the final character recognition result from the combination of similarities for each character pattern candidate calculated in the matching unit 3; 5 is the character recognition determined in the character determination unit 4; This is a data storage unit that stores results.

Now, in the character recognition device having such a configuration, the character determination section 4 described above employs a neural network and is configured as follows.

FIG. 2 shows the configuration of a character determination section constructed using this neural network.

As shown in the figure, this character determination section 4 includes a determination section 11,
It is configured to include a control neural network 12, a plurality of sub-neural networks 13 to 32 (19 in this embodiment), a final determination section 33, and a learning database 34.

The determining unit 11 determines whether the maximum similarity value for each character pattern candidate calculated by pattern matching exceeds a preset threshold j7, and if it exceeds the maximum similarity value. The obtained character pattern candidate is output as the final character recognition result, and if the result is not exceeded, the character recognition result determination processing is passed to the control neural network 12 and the sub-neural networks 13 to 32.

The sub-neural nets 13 to 32 classify each character pattern candidate into groups of similar characters such as "ko, ko, yu" and "ha, he, mu" that are likely to lead to misreading, and compare the input character pattern with each character. Based on each similarity with pattern candidates, learning is performed for each group so that the appropriate character pattern candidate with the maximum similarity is obtained as a character recognition result, and as a result of that learning, the most confident character pattern candidate for each group This is to judge character pattern candidates with a high degree of accuracy.

In addition, the groups of similar characters mentioned above include "mi, yo, wo", "so, tsu, su", "shi, so, n"
"S-L-S""T-T-U""S-T-Na"
"R, Fu, Wa""K, Two, Yu""Ts, Ri, U""R, Yu, Fu""Ke, Chi, Na""Ki, Ma, Ya""O, Ne, Hoj" Examples include "Se, Hi, Mo,""A, Su, Nu,""To, No, Me," and "I, Power, Mouth."

The control neural network 12 determines, based on the degree of similarity between the human-powered character pattern and each character pattern candidate, that the determination results of the sub-neural networks 13 to 32 containing the character pattern candidate with the maximum degree of similarity are the final ones. Learning is performed so that the final determination unit 33 selects the character pattern candidate as the recognition result, and based on the learning result, the sub-neural nets 13 to 32 containing the character pattern candidate with the highest degree of certainty are determined.

The overall learning database 34 contains data used for learning in the control neural network 12 and each sub-neural network 13 to 32, that is, the similarity between the input character pattern and each character pattern candidate and the corresponding correct answer data. The set is stored. In addition, the learning data stored in this overall learning database 34 will be transferred to the learning databases (described later) of each neural net 12.13 to 32 when learning is executed. , the corresponding learning data is sorted through the data sorter 35 and then transferred.

By the way, each neural network 12.13-3 mentioned above
In No. 2, the learning function is realized using a learning algorithm such as a pack propagation algorithm.

FIG. 3 shows the basic configuration of each neural network.

As shown in the figure, the neural network 12 of this embodiment
．． The configurations 13 to 32 include an input layer consisting of a plurality of input nodes A, into which the degree of similarity for each pattern candidate calculated by pattern matching is input, and each human node A is connected in a network via an arc B. The network is connected to an intermediate layer consisting of a plurality of intermediate nodes C, and each intermediate node C is connected to the network via A to D, and outputs a value indicating the final correctness certainty of each pattern candidate based on the similarity to each pattern candidate. and an output layer consisting of an output node E.

Furthermore, a load value indicating the strength of the connection between each node is set for each of the arcs B and D described above, and these load values are modified by the following network management unit.

The configuration of this network management section will be explained using FIG. 4.

In the same figure, 4] is input node A and intermediate node C
42 is a load value table that holds the load values set for each arc B and D, and 43 is the output from each manual node A and intermediate node C. An integrator 44 multiplies each data by a corresponding load value on the load value table 42 and performs integration, and 44 is an adder that adds the integration results to each intermediate node C or output node E. Further, 45 is a determination device that determines the character pattern candidate that has obtained the maximum value from among the final integrated values (similarities) for each character pattern candidate output in normal character recognition as the final character recognition result; 46 imports and compares the final integrated value for each character pattern candidate output during learning with the data of the correct answer with the highest degree of similarity, and evaluates whether or not they match. When the evaluator 46 determines that each data does not match, the evaluator 47 calculates the value on the load value cheaple 42 with a direction to obtain a correct answer in the next literary magazine 1 processing for the input character pattern. A load value corrector 48 corrects each load value, a learning database 48 stores learning data from the overall learning database 33, and a learning manager 49 manages registered contents of the learning database 48.

Next, the basic operation of learning by this network management section will be explained using the flowchart shown in FIG.

First, from the learning database 48, extract combinations of similarities and corresponding correct data sets one by one. In step a), the output values of each human node of the control neural network 12 and sub-neural networks 13 to 32 are extracted. Set each on the table 41 and reviewer 46 (step b)
.

Now, when the similarity is set in the input node output value table 41, each similarity is set for arc B and arc D from input node A to output node E via intermediate node C. The weight values are integrated and further added, and each value finally sent to each output node E becomes a value (certainty) indicating the probability for each character pattern candidate (step C).

Thereafter, the values for each character pattern candidate output from each output node E are output to the evaluator 46. Comparator 46
In step d), the data output from each output node E and the correct data are compared and evaluated. If the two do not match as a result of the comparison evaluation, the load value corrector 47 is activated and the load value table is changed. Each load value in step 42 is corrected in a direction that will allow correct data to be obtained in subsequent processing (step e). Learning manager 4 at the same time
9 stores the combination of input similarity and correct data in the current character recognition process in the learning database 48 again as learning data.

When the L-th learning for all character pattern samples is completed, the learning manager 49 again retrieves one set of each similarity combination and correct data from the learning database 48 in order to start the second learning. These are extracted and set in each of the neural networks 12, 13 to 32, respectively.

The above learning operation is repeated endlessly, and all learning is terminated when correct data can be obtained for all character pattern samples, or when the load value has been corrected a predetermined number of times (step f).

Next, the operation of this character recognition device will be explained.

When pattern matching is performed on a certain handwritten kana character, and the combination of similarities for each character pattern candidate obtained as a result is input to the character determination unit 4, the determination unit 11 performs the following:
First, it is determined whether the maximum similarity among them exceeds a preset threshold. If the maximum similarity exceeds the threshold, the character pattern candidate with the maximum similarity is output as the final character recognition result. If the maximum similarity does not exceed the threshold, the final judgment process of the character recognition result is passed to the control neural network 2 and the sub-neural networks 13 to 32.

When the control neural network 12 inherits this process from the determination unit 11, it determines 17 which sub-neural net 13 to 32 has the highest degree of certainty in the determination result based on the combination of the above-mentioned respective degrees of similarity, The number is sent to the final determination section 33.

On the other hand, each of the sub-neural networks 13 to 32 determines character pattern candidates with the highest degree of certainty as a recognition result of the input character pattern from the combinations of the respective degrees of similarity between characters belonging to their own groups, and is sent to the final determination section 33.

As a result, the final determination unit 33 outputs the determination result of the sub-neural network determined by the control 221 loop 12 as the final character recognition result.

Next, a specific example of learning using this character recognition device will be explained.

Regarding the input character pattern whose correct answer is "Yu", the degree of similarity with each character pattern candidate was calculated in the subneural networks 12 to 32 that include this character. Suppose that the following results are obtained: ......0.8 Yu...0,7.

In this case, the character pattern candidate with the highest degree of similarity is not "ko" but "yu", so
Through learning, the same neural net! - Correct the load value of each arc in 12-32.

In addition, based on the combination of the above-mentioned similarities, the load value of each arc in the control neural network 12 is determined so that the determination results of the sub-neural nets 12 to 32 are determined as having the highest reliability. Correct.

Thus, according to the character recognition device of this embodiment, each of the sub-neural networks 13 to 32 performs learning between character pattern candidates classified into similar patterns and determines the character recognition results, and the control neural network 1
In 2, we learned and determined which sub-neural network]-3 to 32 contains the character pattern candidate with the highest degree of certainty. Even if a large number of characters are included, character recognition can be performed with a high accuracy rate.

In the above embodiment, character recognition by the neural network was performed only when the maximum similarity obtained as a result of matching the input character pattern and the character pattern candidate did not exceed the threshold value 2. As a method, the value obtained by subtracting the second highest similarity from the maximum similarity may be compared with two thresholds, and the next action may be determined from the result.

Further, although this embodiment describes the recognition of handwritten kana characters, the present invention is not limited thereto, and can also be applied to the recognition of Roman characters, kanji, etc., for example.

[Effects of the Invention] As explained above, according to the character recognition device of the present invention, character recognition is performed with a high accuracy rate even for character groups such as handwritten kana characters, which have many types of characters and have many similar patterns. be able to.

[Brief explanation of drawings]

FIG. 1 is a block diagram for explaining the overall configuration of a character recognition device according to an embodiment of the present invention, and FIG. 2 is a block diagram for explaining the configuration of a character determination section using a neural network in FIG. 1. , Fig. 3 is a diagram showing the basic configuration of the neural network, Fig. 4 is a block diagram for explaining the configuration of the network management section in each neural network, and Fig. 5 is a diagram showing the basic configuration of the neural network.
This figure is a flowchart showing the basic operation of learning by the network management section of FIG. 4. 1]... Judgment unit, 12... Control neural network, 13-32... Sub neural network, 33...
Final judgment part. Applicant Toshiba Corporation Patent Attorney Sasa Suyama - Figure 1 ζ Figure 2 Figure 3

Claims (2)

[Claims] (1) In a character recognition device that determines a character pattern candidate with the highest degree of certainty as a recognition result of the input character pattern from each degree of similarity between an input character pattern and a plurality of character pattern candidates, each of the character pattern candidates is converted into a similar character. A plurality of classifications are performed for each group, and learning is performed so that character pattern candidates with the maximum similarity for each group are obtained as the final character recognition results based on the respective degrees of similarity. A plurality of sub-neural networks each determines a character pattern candidate with the highest degree of certainty for each case, and a determination result of the sub-neural network that includes a character pattern candidate with the highest degree of similarity based on the respective degrees of similarity is determined. A character recognition device comprising: a neural network that performs learning to obtain the character recognition result and determines a sub-neural network that includes a character pattern candidate with the highest degree of certainty. (2) The character recognition device according to claim 1, further comprising a comparison means for comparing each degree of similarity obtained between the input character pattern and the plurality of character pattern candidates with a predetermined threshold; When the first output signal is output, the character pattern candidate with the maximum similarity is determined as the character recognition result, and when the comparison means outputs the second output signal, the sub-neural network and the neural network are finalized. A character recognition device characterized by comprising means for passing a determination process of a character recognition result.