JP2000293635A - Character recognizing device and its method and character recognition program recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To shorten a time required for the calculation of cumulative similarity, and to quickly obtain a recognized result without deteriorating recognizing precision in a character recognizing device which recognizes character pattern data in which data for each character of one word data are subdivided as data for one word. SOLUTION: In this character recognizing method, the information of a word dictionary storing part 13 and a character pattern storing part 12 is successively read, and when the compared result of the number of valid word characters stored in the storing part 12 with the number of word characters successively read from the word dictionary is not within the number of character range, the next word is read. Then, cumulative similarity indicting whether or not the word is similar to each word of the word dictionary is calculated from the number of each character connection of each character pattern and the similarity to each character of the character pattern stored in the storing part 12 and the compared result of the comparing part 15. This cumulative similarity is compared with the previously calculated one, and whether or not this should be stored as a word candidate is judged, and then stored. The stored cumulative similarity is rearranged in order.

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, a character recognition method, and a character recognition program recording medium.
The present invention relates to a character recognizing device and a character recognizing method for recognizing character pattern data in which each character of word data of one word is subdivided as one word, and a character recognition program recording medium.

[0002]

2. Description of the Related Art Conventionally developed character recognition apparatuses recognize one character by using various recognition means from character pattern data in which each character of one word data is subdivided, and further recognize one character. , It is determined which word in the word dictionary is most similar, and the result is output. In this case, there is a problem that the processing time is increased by performing complicated recognition processing to improve the recognition accuracy, and the recognition processing speed is reduced.

[0003] As one of conventional character recognition devices for solving such a problem, there is a character recognition device as described in, for example, Japanese Patent Application Laid-Open No. 7-239917. The character recognition device described in the publication will be described with reference to FIG.

In FIG. 1, a conventional character recognition apparatus includes a character extracting unit 1 for extracting a character image for each character from a character image, a character image comparing unit 2 for comparing character images, a character for performing character recognition. It comprises a recognition unit 3, a character image storage unit 4 for storing character images, a control unit 6 for controlling each unit, and a character attribute comparison unit 7 for comparing character attributes and calculating similarity. I have. It is assumed that the storage unit 5 in the character image storage unit 4 includes character attributes, information on character images, character codes, and frequency information.

In such a configuration, the character extracting section 1
A character image for each character is cut out from the character image and passed to the character image comparing unit 2, the character recognizing unit 3, and the character image storing unit 4, and the character attributes of the cut out character image are compared with the character attribute comparing unit 7.
And to the character image storage unit 4.

The character attribute comparison unit 7 compares the character attribute passed from the character cutout unit 1 with the character attribute stored in the character image storage unit 4 read by the control unit 6, and calculates the similarity. . If the similarity is larger than the predetermined value, the control unit 6 is notified to continue the calculation by the character image comparison unit 2 and the character recognition unit 3. If the similarity is smaller than the predetermined value, the character image comparison unit 2
The control unit 6 is instructed to stop the calculation of the similarity with the information on the character image, which is performed in (1), and the operation is stopped until a new character image is output from the character cutout unit 1.

[0007]

In the above-described configuration, after the similarity is calculated, the magnitude of the similarity is determined again.
Since it is determined whether to continue or stop the calculation, there is a disadvantage that the processing time required to move to the next word increases.

In the above-described configuration, it is assumed that one character has already been recognized. However, in the case of character pattern data in which one character is subdivided, it cannot be determined. In such a case, conventionally, the number of combinations that become one character has been derived by calculating the cumulative similarity of adjacent character data for a predetermined connection combination. This conventional configuration has a drawback that the calculation of the cumulative similarity of the extra connection combinations is sometimes forced, which leads to an increase in the processing time.

Furthermore, in the above-described configuration, a recognition result for one character is output each time, and there is a drawback in that the most similar result must be processed again. .

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned drawbacks of the prior art, and has as its object to shorten the time for calculating the cumulative similarity and at the same time to reduce the word recognition result without reducing the recognition accuracy. Is to provide a character recognition device that can obtain

[0011]

According to the character recognition apparatus of the present invention, words are stored in a character recognition apparatus for recognizing cut-out character pattern data obtained by subdividing each character of one word data into one word. Word dictionary, character pattern storage means for storing the number of characters extracted, character similarity, each character connection number, effective word character number of the extracted character pattern data, and information of the word dictionary and the character pattern storage means Address control means for sequentially reading the word pattern, and if the result of comparison between the number of effective word characters stored in the character pattern storage means and the number of word characters sequentially read from the word dictionary is not within the character number range, the result is stored in the word dictionary. Character number comparing means for controlling the address control means so as to read the next word, and each character stored in the character pattern storing means. Based on each character connection number of the turn data and each character similarity that is the similarity to each character of the character pattern and the comparison result of the character number comparison means, how much of the extracted character pattern data matches each word of the word dictionary. Cumulative similarity calculating means for calculating similarity or cumulative similarity; word candidate storage means for storing cumulative similarity; and cumulative similarity calculated by the cumulative similarity calculating means and stored in the word candidate storage means. A calculation result comparison means for comparing the cumulative similarity stored in the word candidate storage means and a ranking means for sequentially rearranging the cumulative similarity stored in the word candidate storage means when comparing the accumulated similarity with the word candidate. It is characterized by having. Then, the cumulative similarity calculation means connects each character data adjacent to each other as one character according to the number of character connections of each character pattern data obtained by dividing one character stored in the character pattern storage means, It is characterized in that a calculation process is performed based on each character similarity stored in the character pattern storage means to determine how similar each word is in the word dictionary. Further, the character number comparing means includes:
The number of effective word characters stored in the character pattern storage means is compared with the number of word characters sequentially read from the word dictionary. If the result is not within the character number range, the cumulative similarity calculation means does not perform the calculation processing. The address control means is controlled to read the next word stored in the word dictionary. Further, the ranking unit is characterized in that only a predetermined number of the most probable candidates for the cumulative similarity as a result of the word candidate storage unit are rearranged in order.

A character recognition method according to the present invention provides a word dictionary in which words are stored, and a character pattern storing the number of characters extracted, character similarity, number of character connections, and number of effective word characters of the extracted character pattern data. A character recognition method for a character recognition apparatus that recognizes cut-out character pattern data in which each character of one word data is subdivided as one word, including a storage unit, wherein the character pattern data is stored in the character pattern storage unit. If the comparison result between the number of valid word characters and the number of word characters sequentially read from the word dictionary is not within the character number range, a character number comparison step of reading the next word stored in the word dictionary; Each character connection number of each stored character pattern data, each character similarity which is the similarity to each character of the character pattern, and the character number comparison A cumulative similarity calculating step for calculating how much the extracted character pattern data is similar to each word in the word dictionary based on the comparison result of the steps; a cumulative similarity calculating step for calculating the cumulative similarity and the word candidate; When comparing the cumulative similarity stored in the storage unit as a word candidate, the calculation result comparison step to be stored in the word candidate storage unit and the cumulative similarity stored in the word candidate storage unit are rearranged in order. And a ranking step.

The recording medium storing the character recognition program according to the present invention includes a word dictionary in which words are stored, the number of characters to be extracted from the extracted character pattern data, the degree of similarity of each character, the number of connected characters, and the number of effective word characters. And a character pattern storage unit for storing a character recognition program for controlling a character recognition device for recognizing cut-out character pattern data in which each character of one word data is subdivided as one word. Medium, the character recognition program stores the effective word character number stored in the character pattern storage unit and the word character number sequentially read from the word dictionary if the comparison result is not within the character number range. A character number comparing step of reading out the stored next word; and a character connection of each character pattern data stored in the character pattern storage unit. Based on each character similarity, which is the similarity of each character of the number and the character pattern, and the comparison result of the number of characters comparing step, how much the extracted character pattern data is similar to each word of the word dictionary is cumulatively similar. A cumulative similarity calculating step of calculating the degree of similarity, and comparing the calculated cumulative similarity with the cumulative similarity stored in the word candidate storage to store the same in the word candidate storage when the word candidate is used as a word candidate. The method further includes a calculation result comparison step, and a ranking step of sequentially rearranging the cumulative similarities stored in the word candidate storage unit.

In short, in the character recognition apparatus, the processing time is shortened by comparing the number of characters in the character pattern information of the word to be recognized before the cumulative similarity calculation requiring a long processing time in the character recognition processing is performed. Further, in the character pattern data in which one character is subdivided, by referring to each character connection number of the character pattern information of the word to be recognized,
The processing time for deriving the number of combinations that make up one character can be reduced. Also, from a word dictionary containing a large number of words, it is possible to rearrange upper candidates having a predetermined number of words and output the result.

[0015]

Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a character recognition device according to the present invention.

Referring to FIG. 1, the apparatus includes an address control unit 10 for providing a control signal for reading data to a character pattern storage unit 11 and a word dictionary storage unit 13, a number of cutout characters, each character similarity, and each character. A character pattern storage unit 11 that stores the number of connections and the number of valid word characters as a character pattern storage unit 12, and a word dictionary storage unit 13 that stores a dictionary word number, word character information, and the number of word characters for each word.
A number-of-characters comparing unit 15 for comparing and determining whether the word is a word to be calculated based on the number of effective word characters in the character pattern storage unit 12 and the number of word characters in the word storage unit 14; The cumulative similarity calculator 16 receives the number of word characters and calculates the cumulative similarity based on the number of cut-out characters, each character similarity, and each character connection number of the character pattern storage unit 12, and the cumulative similarity calculator 16 calculates the cumulative similarity. A calculation result comparison unit 17 that compares the calculation result with a calculation result stored in the word candidate storage unit 18, a word candidate storage unit 18 that stores word candidates, and a word candidate stored in the word candidate storage unit 18. And a result ranking unit 19 for ranking in descending order of cumulative similarity.

The address control unit 10 supplies a control signal for reading data to the character pattern storage unit 11 and the word dictionary storage unit 13 by a control signal 100. As a result, the number of valid words in the character pattern storage unit 12 is transmitted from the character pattern storage unit 11 to the character comparison unit 6 via the signal line 101 and simultaneously transmitted to the cumulative similarity calculation unit 16 via the signal line 105 from the character pattern storage unit 11. The number of cut-out characters, the degree of similarity of each character, and the number of connected characters in the pattern storage unit 12 are sent.

Further, the word line storage 13 stores the signal line 10
2, the number of word characters in the word storage unit 14 is sent to the character comparison unit 6, and at the same time, the signal line 10
6 to the word dictionary storage unit 13
The dictionary word number and word character information of the word storage unit 14 are sent.

The number-of-characters comparing section 15 determines whether the word is a word to be calculated based on the number of valid words in the character pattern storage unit 12 sent by the signal line 101 and the number of words in the word storage unit 14 sent by the signal line 102. Make a comparison judgment. Further, the address control unit 10 controls the control signal 10
3, the control signal 100 is controlled so that the word dictionary storage unit 13 reads the information of the next word.

The character pattern storage unit 11 stores, as character pattern information of a word to be recognized, the number of cut-out characters, each character similarity, each character connection number, and the number of valid word characters as a character pattern storage unit 12.

The number of extracted characters in the character pattern storage unit 12 indicates the number of character patterns of the word to be recognized as a numerical value. Some of the extracted character patterns can be recognized as one character by one character pattern, and others can be recognized as one character by several character patterns. That is, the “number of extracted characters” refers to the number of characters extracted in the character recognition processing. For example,
When recognizing the character "bum", the character pattern data input to the apparatus is "bum" or "burn".
Could be In this case, the number of extracted characters is “3
Character "or" 4 characters ".

Each character similarity of the character pattern storage unit 12 is represented by a numerical value indicating the similarity with which the character pattern described above can be recognized as one character. The similarity is
Numerical values indicate whether a character pattern is recognized as one character or several adjacent character patterns are recognized as one character. In this example, it is assumed that the larger the character similarity is, the higher the similarity is, and the smaller the character similarity is, the smaller the similarity is.

That is, the "character similarity" is a numerical value representing what each of the extracted characters is similar to. For example, the possibility that the character “a” is the first character is expressed as a numerical value. Also, the possibility of the second character is represented as a numerical value, and similarly, the possibility of the nth character is represented as a numerical value. This “n” changes depending on the value of the number of valid word characters.

The number of each character connection in the character pattern storage unit 12 is a characteristic feature of the present invention, and a numerical value indicates how many connection ranges the character pattern described above is recognized as one character. Show.

Conventionally, the cumulative similarity has been calculated a predetermined number of times, and the number of connected characters of the character pattern having the highest similarity has been selected from among the calculations. On the other hand, the number of character connections in the character pattern storage unit 12 is provided for each character pattern of the word to be recognized, and the maximum similarity can be obtained by performing calculations for the number of character connections. Become.

That is, the “number of connected characters” is a number indicating how many of the extracted characters are likely to be connected from adjacent characters. For example, in the case of the character "burn", when focusing on the character in order from the back, there is a possibility that the fourth character "n" is connected to the third character "r" and becomes "m". Second, "r" has no possibility to connect with "u". Also, “u” has no possibility of being connected to “b”. Therefore, in the case of this "burn", the number of character connections of "b", "r" and "u" is "0", and the number of character connections of "n" is "1".

The number of valid word characters in the character pattern storage unit 12 is a characteristic feature of the present invention, and indicates the number of characters that can constitute a word to be recognized and a numerical value indicating the effective range of the number of characters. The number of valid word characters indicates the number of characters of a word to be recognized by a minimum value and a maximum value. For example, if the minimum value is 3 and the maximum value is 6, the words to be recognized are:
A numerical value indicates a possibility that the word can be a word within the range of three to six characters.

That is, the "number of valid word characters" is data required during the character recognition process, and is the number of characters that is set from the minimum value to the maximum value of how many words are likely to be a word. When calculating which word is similar to one of the words registered in the word dictionary, the number of words indicating the possibility is unknown because the character pattern data input first does not tell how many words the word is. Is the number of valid word characters.

The word dictionary storage unit 13 stores a dictionary word number, word character information, and the number of word characters as word storage units 14 as information of each word prepared in advance. The word dictionary storage unit 13 stores word storage units 14 for the number of words.

The dictionary word number of the word storage unit 14 is a number sequentially assigned to each word in advance. Word storage unit 14
The word character information is information indicating what kind of character the word is composed of. For example, information indicating which character of the alphabet is the first character and which character is the second character is stored.

The number of words in the word storage unit 14 indicates the number of characters in each word by a numerical value.

The number-of-characters comparing section 15 is not included in the conventional constituent elements. The number-of-characters comparing section 15 transmits the number of effective word characters of the character pattern storage unit 12 and the signal line 102 transmitted by the signal line 101.
A comparison is made based on the number of word characters in the word storage unit 14 sent from the computer to determine whether the word is a calculation target word. If the word is not a calculation target word, the next word in the word dictionary storage unit 13 is read out to the address control unit 10. Control signal 103 is output as described above.

For example, if the number of valid word characters in the character pattern storage unit 12 is 3 to 6 characters and the number of word characters in the currently read word storage unit 14 is 8, the address control unit 10 stores the word dictionary storage unit. 13 is read out, and when the next new word character number is read from the word dictionary storage unit 13, it is again determined whether or not the word is a calculation target word. The process is repeated until the data is read out or compared with all the words in the word dictionary storage unit 13.

The cumulative similarity calculator 16 controls the control signal 104
Receiving the instruction of the calculation process and the number of word characters from the character comparison unit 6 sent by the address control unit 10, the number of cut-out characters of the character pattern storage unit 12 read by the address control unit 10,
The character similarity and the number of connected characters are transmitted by the signal line 105, and the dictionary number of the word storage unit 14 and the word character information are stored in a known manner based on the data transmitted by the signal line 106. calculate.

When calculating the cumulative similarity, the "character similarity" is added according to the "number of connected characters" and the "word dictionary". For example, when recognizing a character "bum", the character data pattern is "bum" or "bur".
Since the number may be "n", the calculation is performed assuming that the number is three or four. Character similarity as the first character of "b" + character similarity as the second character of "u" + "m"
And the cumulative similarity of the character similarity as the third character of "b"
Of character similarity as first character + character similarity as second character of "u" + character similarity as third character of "r" + character similarity as fourth character of "n" The similarity is compared, and the one with the larger cumulative similarity (the most similar one) is set as a candidate.

The calculation result comparison unit 17 compares the calculation result sent from the control line 107 by the cumulative similarity calculation unit 16 with the calculation result sent from the control line 108 and stored in the word candidate storage unit 18. Then, only a predetermined number is stored in the word candidate storage unit 18 via the control line 109. For example, if ten word candidates are stored in advance, the word candidate storage unit 18 stores the ten cumulative similarities and the dictionary word numbers from the one with the largest cumulative similarity as a calculation result.

The word candidate storage section 18 is a section for storing the calculation results of the cumulative similarity calculation and dictionary word numbers for a predetermined number.

The result ranking unit 19 stores the word candidate storage unit 1
8 is read through the control line 110,
The ranking is made in order from the one with the largest cumulative similarity. This ranking result is sent to the word candidate storage unit 18 via the control line 111, and the word candidate information is stored again.

[0039]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A more detailed embodiment of the present invention will be described below with reference to FIG.

Here, for the sake of simplicity, it is assumed that the number of effective word characters of the character pattern storage unit 12 transmitted by the signal line 101 is 3 to 6 characters. It is also assumed that the word storage unit 14 sent by the signal line 102 has eight word characters. Then, since the currently read dictionary word is not the number of characters to be calculated, the control signal 103 instructs the address control unit 10 to read the next word. As a result, the information of the next word is newly sent from the word dictionary storage unit 13 to the character comparison unit 6 via the signal line 102 via the signal line 102, and at the same time the cumulative similarity is sent from the word dictionary storage unit 13 via the signal line 106. Degree calculator 16
The dictionary word number and word character information of the word storage unit 14 of the word dictionary storage unit 13 are sent to the server. Then, the character number comparison unit 1
5 again determines whether or not the word is a calculation target word, and reads out a word having a word number of 3 to 6 characters, which is the number of words to be calculated, or checks all words in the word dictionary storage unit 13 Repeat until compared.

If the number of word characters is three to six, the control signal 104 sends an instruction for calculation processing and the number of word characters to the cumulative similarity calculator 16.

The cumulative similarity calculator 16 controls the control signal 104
Receiving the instruction of the calculation process and the number of word characters from the character comparison unit 6 sent by the address control unit 10,
Each character similarity and each character connection number are transmitted by the signal line 105, and the cumulative similarity is calculated based on the dictionary number of the word storage unit 14 and the word character information transmitted by the signal signal 106. The accumulative similarity calculation unit 16 calculates the signal line 10
5 is a circuit for calculating the accumulated similarity in a well-known manner based on the information transmitted by the signal line 5 and the information transmitted by the signal line 106. The calculated cumulative similarity is represented by the signal line 106
Along with the dictionary number of the word storage unit 14 sent to the calculation result comparison unit 17 via the signal line 107.

The calculation result comparing section 17 compares the cumulative similarity sent from the signal line 107 with the calculation result read out via the control line 108 and stored in the word candidate storage section 18. As the calculation results of the word candidate storage unit 18, ten calculation results are stored. The calculation result comparison unit 17 searches the calculation results of the word candidate storage unit 18 for the smallest calculation result (the least similar one) for each word, and calculates the calculated value and the calculation value transmitted from the signal line 107. The resulting cumulative similarity and the dictionary word number are replaced and stored via the control line signal line 109.

The result ranking unit 19 stores the word candidate storage unit 1
The 10 word candidates stored in 8 are read by the control line 110, and are ranked in descending order of the cumulative similarity.
This is a portion where the word candidate information is stored again in the word candidate storage unit 18 by the control line 111. The word candidate storage unit 18 stores ten cumulative similarities and dictionary word numbers in order from the one with the largest cumulative similarity.

The ranking by the result ranking unit 19 is performed after all the words in the word dictionary have been collated. In this example, ten candidate words will eventually remain in the word candidate storage unit 18, but initially the word candidate storage unit 1
Nothing is stored in 8. That is, initially, the lowest value is stored in the word candidate storage unit 18 as an initial value. After the “minimum calculation result” is searched by the calculation result comparison unit 17 and the value is replaced, the result ranking unit 19 ranks the last.

For example, when the lowest value is "-9999", the initial value is: Candidate 0: -9999 Candidate 1: -9999 Candidate 3: -9999 Candidate 4: -9999 Candidate 5: -9999 Candidate 6: -9999 Candidate 7: -9999 Candidate 8: -9999 Candidate 9: -9999 The result of the calculation by the cumulative similarity calculation unit 16 is as follows: Candidate 0: +100 Candidate 1: +4000 Candidate 2: +800 Candidate 3: −Candidate 4: +6000 Candidate 5: +200 Candidate 6: −50 Candidate 7: +5500 Candidate 8 : +3800 Candidate 9: +4500

The result ranking unit 19 calculates the calculation result.
In order of the following, the candidate 0: +6000 candidate 1: +5500 candidate 2: +4500 candidate 3: +4000 candidate 4: +3800 candidate 5: +800 candidate 6: +200 candidate 7: +100 candidate 8: -20 candidate 9: -50 Become.

Although one embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various additions and changes are possible. For example, the character pattern storage unit 12
Although the number of valid word characters is changed from 3 to 6 or the number of word candidates stored in the word candidate storage unit 18 is set to 10, the number can be increased or decreased. Further, these need not be fixed values. For example, it is also possible to add storage means such as a register and a memory to change the numerical values. Further, in the result ranking unit 19, sorting is performed in order from the cumulative similarity of the maximum value, but it is also possible to sort from the minimum value. In addition, the present invention is not limited to the above embodiments, and within the scope of the technical idea of the present invention,
Obviously, each embodiment can be appropriately changed.

In the character recognition apparatus, the word dictionary storing words and the characters storing the number of characters extracted from the extracted character pattern data, each character similarity, each character connection number, and the number of valid word characters are stored. A character recognition method in a character recognition device including a pattern storage unit and recognizing cut-out character pattern data in which each character of one word data is subdivided as one word is realized. This character recognition method will be described with reference to FIG.

In the figure, first, when the comparison result between the number of effective word characters stored in the character pattern storage unit and the number of word characters sequentially read from the word dictionary is not within the range of the number of characters, it is stored in the word dictionary. The next word that has been read is read (step S21). Next, based on the number of character connections of each character pattern data stored in the character pattern storage unit 12 and each character similarity, which is the similarity of each character of the character pattern, and the comparison result in step S21, the clipping is performed. A cumulative similarity calculation is performed to determine how similar the character pattern data is to each word in the word dictionary (step S22). Further, the calculated cumulative similarity is compared with the content stored in the word candidate storage unit 18 to determine whether or not to be stored as a word candidate (step S23). In this case, if the previously calculated cumulative similarity is stored in the word candidate storage unit 18, it is compared with the stored cumulative similarity. However, when the cumulative similarity is not stored in the word candidate storage unit 18 and an initial value such as “−9999” is stored, it is compared with the initial value.

Then, the cumulative similarity determined in step S23 is stored (step S24). Finally, the stored cumulative similarities are rearranged in order (step S25).

Here, in step S22, adjacent character data are connected as one character according to the number of character connections of each character pattern data in which one character stored in the character pattern storage unit 12 is subdivided. A calculation process is performed to determine how similar each word in the word dictionary is from each character similarity stored in the character pattern storage unit 12.

In step S21, if the result of comparing the number of effective word characters stored in the character pattern storage section with the number of word characters sequentially read from the word dictionary is not within the character number range, the calculation process in step S22 is performed. Instead, the next word stored in the word dictionary is controlled to be read.

Further, in step S25,
Only a predetermined number of the most probable candidates for the cumulative similarity stored in step S24 are rearranged in order.

It is to be noted that the same operation as described above can be performed by preparing a recording medium on which a program for realizing the processing of FIG. 2 described above is recorded and controlling each unit of FIG. 1 using the recording medium. Is obvious. This recording medium contains
Various recording media can be used in addition to the semiconductor memory and the magnetic disk device not shown in FIG.

If the computer is controlled by the program recorded on the recording medium, it is obvious that the character recognition operation can be performed in the same manner as described above. As this recording medium, various recording media other than the semiconductor memory and the magnetic disk device can be used.

[0057]

As described above, the present invention compares the number of characters in the character pattern information of the word to be recognized before the cumulative similarity calculation which takes a long processing time in the character recognition processing. In the character pattern data in which one character is subdivided, the processing time for deriving the number of combinations that become one character is reduced by referring to the number of each character connection of the character pattern information of the word to be recognized. There is an effect that it can be shortened. In addition, there is an effect that it is possible to rearrange upper candidates having a predetermined number of words from a word dictionary storing a large number of words and output the result.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing one embodiment of a character recognition device of the present invention.

FIG. 2 is a flowchart illustrating a character recognition method realized by the apparatus of FIG. 1;

FIG. 3 is a configuration diagram illustrating an example of a conventional character recognition device.

[Explanation of symbols]

 Reference Signs List 10 Address control unit 11 Character pattern storage unit 12 Character pattern storage unit 13 Word dictionary storage unit 14 Word storage unit 15 Character number comparison unit 16 Cumulative similarity calculation unit 17 Calculation result comparison unit 18 Word candidate storage unit 19 Result ranking unit

Claims (12)

[Claims] 1. A character recognition apparatus for recognizing cut-out character pattern data in which each character of one word data is subdivided as one word, comprising: a word dictionary in which words are stored; Character pattern storage means for storing the number of extracted characters of data, each character similarity, each character connection number, and the number of valid word characters; address control means for sequentially reading information from the word dictionary and the character pattern storage means;
If the comparison result between the number of effective word characters stored in the character pattern storage means and the number of word characters sequentially read from the word dictionary is not within the character number range, the next word stored in the word dictionary is read. A number-of-characters comparison unit that controls the address control unit; a character-similarity that is a similarity to each character of each character-pattern data stored in the character-pattern storage unit; Cumulative similarity calculating means for calculating how similar the extracted character pattern data is to each word in the word dictionary based on the comparison result of the means, and word candidate storing means for storing the cumulative similarity And comparing the cumulative similarity calculated by the cumulative similarity calculating means with the cumulative similarity stored in the word candidate storage means, Character recognition apparatus characterized by having a successively permuting ranking means and calculation results comparing means to be stored in the word candidate storage unit, the cumulative similarity stored in the word candidate storage means if that. 2. The method according to claim 1, wherein the accumulative similarity calculating means connects each character data adjacent to each other as one character according to the number of character connections of each character pattern data obtained by dividing one character stored in the character pattern storage means. 2. The character recognition device according to claim 1, wherein the character recognition device calculates a degree of similarity with each word in the word dictionary from each character similarity stored in the character pattern storage means. 3. The character number comparing means compares the number of effective word characters stored in the character pattern storage means with the number of word characters sequentially read from a word dictionary. 2. A character recognition apparatus according to claim 1, wherein said address control means controls said address control means to read the next word stored in said word dictionary without performing the calculation processing of said degree calculation means. 4. The character recognition device according to claim 1, wherein the ranking unit sequentially rearranges only a predetermined number of the most probable candidates for the cumulative similarity as a result of the word candidate storage unit. apparatus. 5. A word dictionary in which words are stored, and a character pattern storage unit that stores the number of extracted characters of the extracted character pattern data, each character similarity, each character connection number, and the number of valid word characters, What is claimed is: 1. A character recognition method for a character recognition apparatus for recognizing cut-out character pattern data in which each character of one word data is subdivided as one word, the number of effective word characters stored in said character pattern storage unit and the number of words If the result of comparison with the number of words and characters sequentially read from the dictionary is not within the number of characters range, a character number comparing step of reading the next word stored in the word dictionary; and each character stored in the character pattern storage unit. Based on each character similarity which is a similarity to each character of the character data and each character of the pattern data, and a comparison result of the character number comparing step A cumulative similarity calculating step of calculating how much the extracted character pattern data is similar to each word in the word dictionary, and storing the calculated cumulative similarity and the word candidate storage unit. When a word candidate is compared with the cumulative similarity, a calculation result comparison step of storing the cumulative similarity in the word candidate storage unit and a ranking step of sequentially rearranging the cumulative similarity stored in the word candidate storage unit are included. A character recognition method characterized in that: 6. In the step of calculating the cumulative similarity, each character data which is stored in the character pattern storage unit is defined as one character according to the number of character connections of each character pattern data obtained by subdividing one character. 6. The character recognition method according to claim 5, wherein a connection process is performed to calculate how similar each word in the word dictionary is from each character similarity stored in the character pattern storage unit. 7. In the character number comparing step, if the result of comparing the number of effective word characters stored in the character pattern storage unit with the number of word characters sequentially read from the word dictionary is not within the character number range, the cumulative similarity is calculated. 7. The character recognition method according to claim 5, wherein control is performed such that the next word stored in the word dictionary is read out without performing the calculation processing in the degree calculation step. 8. The method according to claim 5, wherein in the ranking step, only a predetermined number of the most probable candidates for the cumulative similarity stored in the word candidate storage step are rearranged in order. The character recognition method described in any of the above. 9. A word dictionary in which words are stored, and a character pattern storage unit that stores the number of cut-out characters of the cut-out character pattern data, each character similarity, each character connection number, and the number of valid word characters, A recording medium storing a character recognition program for controlling a character recognition device that recognizes cut-out character pattern data in which each character of one word data is subdivided as one word, the character recognition program comprising: If the comparison result between the number of valid word characters stored in the character pattern storage unit and the number of word characters sequentially read from the word dictionary is not within the range of the number of characters, a character number comparison for reading the next word stored in the word dictionary Steps and each sentence indicating the number of character connections of each character pattern data stored in the character pattern storage unit and the degree of similarity of each character in the character pattern A cumulative similarity calculating step for calculating, based on the character similarity and the comparison result of the character number comparing step, how much the extracted character pattern data is similar to each word in the word dictionary; Comparing the calculated cumulative similarity with the cumulative similarity stored in the word candidate storage unit as a word candidate, a calculation result comparison step of storing the result in the word candidate storage unit; and storing the calculated result in the word candidate storage unit. A ranking step of sequentially rearranging the obtained cumulative similarities. 10. In the step of calculating the cumulative similarity, each character data which is stored in the character pattern storage unit is set as one character according to the number of character connections of each character pattern data obtained by subdividing one character. 10. The recording medium according to claim 9, wherein the recording medium is connected to calculate how much the word is similar to each word in the word dictionary from each character similarity stored in the character pattern storage unit. 11. In the character number comparing step,
If the result of comparing the number of effective word characters stored in the character pattern storage unit with the number of word characters sequentially read from the word dictionary is not within the range of the number of characters, the word is not processed in the cumulative similarity calculation step and the word is not processed. 11. The recording medium according to claim 9, wherein control is performed to read a next word stored in the dictionary. 12. The method according to claim 9, wherein in the ranking step, only a predetermined number of the most probable candidates for the cumulative similarity stored in the word candidate storage step are rearranged in order. The recording medium according to any one of the above.