JPH05114052A - Method and device for recognizing character - Google Patents

Abstract

PURPOSE:To automatically correct the misrecognition of numerals. CONSTITUTION:Each character written in an image original read out from an image input part 101 is segmented and recognized and plural candidate characters are outputted. While referring to a numeral table 110, a numeral collating part 109 retrieves whether a numeral is included in the candicate characters or not. When a numeral is included in the candidate characters before or after the retrieved numeral, a numeral correcting part 111 corrects the numeral to the 1st candidate character for character recognition. The above processing is repeated up to the finally inputted character.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition method and an apparatus thereof, and more particularly to a character recognition method and an apparatus thereof having an automatic recognition error correction function.

[0002]

2. Description of the Related Art Conventionally, a character recognition device extracts the characteristics of an input character image, calculates the degree of similarity with the characteristics of individual standard pattern characters prepared in advance, and determines the character with the highest degree of similarity as the recognition result. It is outputting. In addition, in order to improve the accuracy of such character recognition, the recognition result is divided into words, collated with a prepared word dictionary, and some characters of the word are replaced with other characters to create a word dictionary. There is a known method for automatically correcting erroneous recognition by replacing the recognition result when it matches a word in.

[0003]

However, in the above-mentioned conventional example, since the automatic correction is performed by collating with the word dictionary, it can be corrected only in units of words composed of two or more characters. Therefore, when you recognize a sentence such as "My date of birth is August 21, 1940",
For example, "My date of birth is August 2l (O of the alphabet) August 2l (L of the alphabet)", even if a number is mistakenly recognized as a similar character, it is not automatically corrected and remains incorrect. There was a flaw.

The present invention has been made in view of the above-mentioned conventional example, and an object thereof is to provide a character recognition method and an apparatus therefor capable of automatically correcting erroneous recognition of numbers.

[0005]

In order to achieve the above object, the character recognition method of the present invention comprises the following steps. That is, a character recognition method for recognizing a character drawn on the image original by inputting an image original on which characters are drawn, the input step of the image original, and the characters drawn on the image original are cut out one by one. Recognition step of performing character recognition by outputting a plurality of recognition candidate characters, a retrieval step of retrieving a number from the characters recognized by the recognition step, and a number when the number is retrieved by the retrieval step. A correction step of checking whether or not a plurality of recognition candidate characters existing before and after include a number, and if the number is included, correcting the number to be the most probable candidate of the plurality of recognition candidate characters. And a character recognition method comprising:

According to another aspect of the invention, there is provided a character recognition device for recognizing a character drawn on the image original by inputting the image original on which the character is drawn, the input means for inputting the image original,
The characters drawn on the image original are cut out one by one to perform character recognition, a recognition means for outputting a plurality of recognition candidate characters, a storage means for storing a plurality of numbers, and a plurality of the plurality of storage means stored in the storage means. With reference to a numeral, a search means for searching a numeral from the characters recognized by the recognition means and a check to see whether or not a plurality of recognition candidate characters existing before and after the numeral include a numeral, When it is included, the character recognition device is provided with a correction means for correcting the numeral so as to be the most probable candidate of the plurality of recognition candidate characters.

[0007]

With the above structure, the present invention searches for a numeral from the characters recognized as characters, checks whether or not a numeral is included in a plurality of recognition candidate characters of the characters before and after the numeral, and recognizes the numeral. If the number is included, the number is corrected so as to be the most promising candidate of the recognition candidate character.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of the present invention will be described in detail below with reference to the accompanying drawings.

FIG. 1 is a block diagram showing the configuration of a character recognition apparatus which is a typical embodiment of the present invention. In FIG. 1, 1 is a CPU, 2 is a ROM for storing programs,
3 is a RAM, 101 is an image input unit such as an image scanner, 102 is a character cutout unit that cuts out a character image for each character from the input image, 103 is a feature extraction unit that extracts features from the cutout character image, and 104 is a A similarity calculation is performed between the extracted features of the input character image and the recognition dictionary 105 in which the features of individual characters are stored, and the recognition result area in which n pieces are arranged in the RAM 3 in descending order of similarity. This is a similarity calculation unit to be stored in 106.

Here, n may be variable for each input character, and the case where the value of n is 1 or 0 (unrecognizable) is allowed. The first candidate character, in descending order of similarity of the n recognition results,
..., Let's call it the nth candidate character. The entire recognition result is called a candidate character.

Reference numeral 107 denotes a word collating unit which collates and corrects the recognition result stored in the recognition result area 106 in units of words by using the word dictionary 108, and 109 is a numeral attached after "pieces""yen" etc. Numerals collating unit for collating with the numerical table 110, 111 is a numeral correcting unit that judges and corrects the erroneous recognition of the numbers before and after the numerals matched by the numeral matching unit 109, and 112 is corrected. It is a display which displays a recognition result. Reference numeral 113 is a number table changing unit that allows addition and deletion of numbers in the number table 110.

According to a known technique, the characters are input and cut out one by one, the characteristics of each character are extracted, the similarity is calculated with reference to the recognition dictionary 105, and candidate characters are obtained. Further, according to a known technique, the word dictionary 1
With reference to 08, correction is made in word units (this is called word matching). The character corrected here is regarded as the first candidate character, and the candidate character that has been stored until then is stored again with the rank shifted.

Assuming that the above processing has been executed, the most characteristic number matching process in this embodiment is executed for the first candidate character at this point. This processing will be described below with reference to the flowchart shown in FIG. 2 and the example of the number table shown in FIG. The numeric collation unit 109 of the character recognition apparatus of this embodiment plays a major role in this processing.

FIG. 3 is a diagram showing an example of information stored in the numeral table 110. In FIG. 3, the number table 1
Numeral 10 includes a numeral such as a word, which is added immediately after the numeral.

First, in step S10, the collation character is changed to the second character.
Set as the initial value for the th input character. The reason why at least one numeric character must always exist before the numeral is the reason for setting the second input character in step S10. In step S15, it is checked whether or not there is a numeral in the numeral table 110, which has the collation character as a head. Here, if the number is in the number table 110, the process proceeds to step S20, and the number table 1
If not present, the process proceeds to step S50 to advance the position of the collation character to the next character.

In step S20, the control of the processing moves to the numeral correction unit 111, and the character immediately before the collation character is set as the target character. Succeedingly, in a step S25, it is checked whether or not a numeral is present in the candidate character of the noted character. Here, if a number is included in the candidate characters of the target character, the process proceeds to step S30, and the character is set as the first candidate. Further, when a plurality of numbers are included in the candidate characters, the one having the higher rank is set as the first candidate. Further, in step S35, the noticed character is moved forward by one, and further, in step S40, it is confirmed that the position of the noticed character is not shifted before the start position of the input character, and then steps S25-
The same process as S35 is performed. On the other hand, if a number is not included in the candidate characters of the target character, or if the target character position has returned to the start position of the input character, the process proceeds to step S45. At this point, control of the process returns from the number correcting unit 111 to the number matching unit 109 again.

In step S45, the position of the collation character is shifted by the longest character number + 1 character among the collated numerals.

Finally, in step S55, it is checked whether or not the last collation of the input character is completed. Here, if it is determined that the collation has not reached by the last input character, the process returns to step S15 to continue the collation process, and if it is determined that the last collation of the input character is completed, the process is terminated. finish.

By executing the above-described number matching process, for example, a character string "My date of birth is August 2l (L) August 194O" is obtained as the first candidate. When the collation character position is "Year", the numerical correction is performed. Here, the character of interest is "O
(O) ”and“ 0 (zero) ”in the candidate characters
If it exists, the first candidate character is changed to “0 (zero)”. Next, the attention character is shifted to "4". In this case, since the first candidate character already has a number, the character of interest moves further forward, and if the number does not exist in the candidate character when "ha" comes, the correction process is performed at that time. Finished,
The matching process is executed again. In the same way, "Month",
The collation process is performed with "day", and particularly when "1" exists in the character candidate when the target character becomes "l", this is changed to the first candidate character. When all the matching is completed, the result is displayed on the display 112.

Therefore, according to the present embodiment, the numeral in the candidate character is searched, and if the candidate character located before the numeral includes a numeral, the numeral is corrected to the first candidate character. This makes it possible to automatically correct misrecognition of numbers.

[0021]

[Other Embodiments] Here, a process of collating a number when only the structure of the number table 110 is different by using a character recognition apparatus common to the above-described embodiments will be described with reference to a flowchart shown in FIG. .. Similar to the above-described embodiment, the input of characters and the acquisition of candidate characters that are the prerequisites for the process of matching the number of words follow the known technique, and the description thereof will be omitted.
In addition, in the number matching process of the present embodiment, the same parts as those in the above-described embodiments are designated by common step reference numerals and description thereof will be omitted.

Numeral table 11 used in this embodiment
0 is a numeric table that collects numerics such as “¥” and “Heisei” that are prepended to numbers. This is the premise of this embodiment. First, in step S100, the collation character is set to the input leading character as an initial value. In the process of step S15, when a number starting with the collation character is searched from the number table 110, the process proceeds to step S120, and the process control proceeds to the number correcting unit 111.

In step S120, the numeric collator 109
The character next to the longest number among the numbers that have been matched with is the character of interest. Thereafter, the same processing as steps S25 to S35 is performed. Here, if there is no number in the character candidates of the target character, or if the target character position has reached the last position of the input character in step S140, the process proceeds to step S145. At this point, control of the process returns from the number correcting unit 111 to the number matching unit 109 again.

In step S145, the numeral correction unit 111
The character of interest at the end of the processing by is the next matching character.

Therefore, according to the present embodiment, the numeral table 11 storing the information of the numeral added before the numeral such as the year number.
By using 0, it becomes possible to automatically correct the misrecognition of the numeral.

In each of the above-mentioned two embodiments, the correction in the erroneous recognition of a number when the number is present before the number and the number is after the number has been described.
The present invention is not limited to these. For example, the present invention can be applied to a combination of the above two embodiments.

Further, if the number of the number which is not included in the number table 110 is known in advance in the sentence to be recognized by the device user, the number table changing unit 113 stores information in the number table 110. Can be added. This allows automatic correction even if there is a unique number in the sentence to be recognized.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of one device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0029]

As described above, according to the present invention, the erroneous recognition of the numbers existing before and after the number can be automatically corrected by using the number in the recognized character as a clue. Not only the recognition accuracy is improved, but also the manual confirmation / correction work of erroneous recognition is reduced, and the time for document input work can be shortened.

[Brief description of drawings]

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

FIG. 2 is a flowchart showing a number matching process.

FIG. 3 is a diagram showing an example of a numeric table 111.

FIG. 4 is a flowchart showing a number matching process according to another embodiment.

[Explanation of symbols]

 1 CPU 2 ROM 3 RAM 101 Image Input Section 102 Character Extraction Section 103 Feature Extraction Section 104 Similarity Calculation Section 105 Recognition Dictionary 106 Recognition Result Storage Area 107 Word Matching Section 108 Word Dictionary 109 Numeral Matching Section 110 Numeral Table 111 Number Modifying Section 112 Display 113 Number table change section

Claims (3)

[Claims] 1. A character recognition method for recognizing a character drawn on an image original by inputting an image original on which a character is drawn, the input step of the image original, and the character drawn on the image original. A recognition step of slicing out one character at a time to perform character recognition and outputting a plurality of recognition candidate characters, a search step of searching for a number from the characters recognized by the recognition step, and a case where the number is searched by the search step , Checking whether or not a plurality of recognition candidate characters existing before and after the number includes a number, and if the number is included, corrects the number to be the most probable candidate of the plurality of recognition candidate characters A character recognition method, comprising: 2. A character recognition apparatus for recognizing a character drawn on an image original by inputting an image original on which a character is drawn, the input means for inputting the image original, and the character drawn on the image original. Character recognition is performed by cutting out one character at a time, and a recognition unit that outputs a plurality of recognition candidate characters, a storage unit that stores a plurality of numbers, and a plurality of numbers stored in the storage unit are referred to,
Retrieval means for retrieving a number from among the characters recognized by the recognition means, and a check is performed to see if a plurality of recognition candidate characters existing before and after the number includes a numeral, and if a numeral is included. A character recognizing device for correcting the numeral so as to be a most probable candidate of the plurality of recognition candidate characters. 3. The character recognition device according to claim 2, further comprising changing means capable of changing the contents of the plurality of numbers stored in the storage means.