JPH05258100A - Character recognition device - Google Patents

Abstract

PURPOSE:To improve the processing efficiency of a word retrieval by selecting a recognized character for each digit constituting a candidate word based on a candidate character group by a candidate character selecting part. CONSTITUTION:This device is equipped with recognition processing parts 2 and 4 which recognizes character information 42 read by plural reading parts 101 and 301 according to the attribute of the ink of character on a slip 44, and prepares plural ranked candidate character groups 21 and 41, and a word preparing part 55 which judges whether or not the same character is present in the both candidate character groups 21 and 41, and prepares the most appropriate candidate word 51. Then, the character information 42 is read by an OCR reading processing part 11 and an MICR reading processing part 31, and the candidate character groups 21 and 41 prepared by the recognition processing parts 2 and 4 are outputted to a candidate character selecting part 5. The candidate character selecting part 5 prepares extracted characters 45 and 25 based on the candidate character groups 21 and 41, outputs them to the word preparing part 55, and the word preparing part 55 prepares the candidate word 51 constituted of the combination of the extracted character 45 and 25.

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 for reading character information printed with magnetic ink by optical means and MICR.

The read character is collated with a predetermined character pattern for each character, predetermined candidate characters are determined in descending order of similarity to the character pattern, and are generated from a combination of the candidate characters. Character recognition is performed by searching a predetermined word dictionary with the candidate word as a key and comparing and checking. As the number of candidate characters increases in order to improve the accuracy of character recognition, the number of candidate words increases exponentially, and it takes a lot of time to search a dictionary file etc. Selection is requested.

[0003]

2. Description of the Related Art FIG. 7 shows an explanatory view of conventional character recognition.
In FIG. 7, FIG. 7A shows an example of a form used for tax payment, etc., but the character data of this form, for example, character information consisting of a plurality of characters in the name field is read by an optical character reading device. Then, each character is recognized, and a plurality of candidate characters (for example, three characters) are selected for each character in order of high similarity for each digit where the character appears.

A case in which character data of three characters as shown in FIG. 7B is recognized will be described below. That is, when the character data is read by the optical character reading device and the character is recognized, three candidate characters are selected for each character in order from the higher order of recognition as shown in FIG. 7C. It

1 out of each of the first, second and third candidate characters
Characters are extracted to generate candidate words that form a name.
Using the candidate word as a key, a knowledge dictionary (file in which name information is filed) is searched, and it is compared and checked whether or not there is a word that matches the candidate word. The name existing in the knowledge dictionary is determined as the final recognized word of the character data.

[0006]

[Problems to be Solved by the Invention] As described above, 3
Words composed of letters exist only in combinations (3 to the third power) that take out one character from each of the first, second, and third candidate characters, and knowledge increases as the number of words and candidate characters increases. There was a problem that it took a lot of time to search the dictionary.

An object of the present invention is to improve the processing efficiency of character recognition.

[0008]

FIG. 1 is a block diagram showing the principle of the present invention. A character recognition device that has a plurality of reading units (1,3) according to the attributes of the ink that formed the characters on the form, and recognized and ranked the character information read by each reading unit 1,3. The recognition processing units 2 and 4 for respectively generating a plurality of candidate character groups 21 and 41 are compared with the candidate character group characters 21 and 41 to determine whether or not the same character exists in both candidate character groups. , A word generator 55 for generating the most probable candidate word 51.

[0009]

The character information is read by the OCR reading processing section 10 and the MICR reading processing section 30, and the candidate character group 21, 41 generated by the recognition processing section 2, 4 is output to the candidate character selecting section 5, and the candidate character group is selected. The selection unit 5 generates the extracted characters 45,25 based on the candidate characters 21,41 and outputs the extracted characters 45,25 to the word generation unit 55, and the word generation unit 55 generates the candidate word 51 composed of the combination of the extracted characters 45,25. To do.

[0010]

EXAMPLE FIG. 2 shows a system configuration diagram of an example. Figure 2
In the character recognition device, an optical character reading device 101 including an OCR unit 11 and a recognition processing unit 2 and an MICR unit 31 are used.
A magnetic ink character reading device 301 including a recognition processing section 4, a processing device 50, a knowledge file 57, and a display device 58.

The processing device 50 includes a character reading processing unit 13, a candidate character storage unit 52, an extracted character storage unit 53, a candidate character selection unit 5, a word generation unit 55, a candidate word storage unit 56, and knowledge. The knowledge file 57 includes a name dictionary used for character recognition and the like.

In FIG. 2, the character information 42 printed on the form 44 with magnetic ink characters is read by the optical character reading device 10 and the magnetic ink character reading device 30. The details of the form 44 are shown in FIG. 3A of the character information and candidate character generation explanatory diagram shown in FIG. It is assumed that the character information 42 is printed with magnetic ink characters, and the first step of the recognition processing of the character information will be described in accordance with the candidate character generation processing procedure of FIG.

In step 80 of FIG. 4, the character reading processing unit 13 commands the optical character reading device 101 to read the form 44. In processing step 81, the OCR unit 11 reads the character information 42, and the recognition processing unit 2 performs character recognition by referring to a recognition dictionary (not shown) for each digit of the character. In the character recognition result processing step 82, a predetermined number of characters that are ranked in order of increasing similarity to the standard character pattern for each digit,
For example, a candidate character group 21 of 4 characters is generated. The generated candidate character for each character is shown in FIG. That is, character information
For the character "Ku" in the first digit of 42, four characters "Ku, Wa, Ta, Fu" are generated as the candidate character group 21.

In processing step 83, the character reading processing unit 13 inputs the candidate character group 21 from the optical character reading device 101 and stores it in the candidate character storage unit 52. Next, in step 84, the character reading processing unit 13 commands the magnetic ink character reading device 301 to read the form 44.

At the processing step 85, the MICR section 31 determines the character information.
42 is read, and the recognition processing unit 4 performs character recognition by referring to a recognition dictionary (not shown) for each character digit. In the character recognition result processing step 86, a candidate character group 41 of a predetermined number, for example, four characters, which is ranked in order of increasing similarity to the standard character pattern for each character, is generated.

FIG. 3C shows the generated candidate characters for each character.
Shown in. That is, the character "Ku" in the first digit of the character information 42 is
Four characters of “ku, ta, ku, fu” are generated as the candidate character group 41.
In processing step 87, the character reading processing unit 13 determines that the candidate character group 41
Is input from the magnetic ink character reading unit 301 and stored in the candidate character storage unit 52.

Candidate character groups resulting from character recognition by the optical character reader 101 and the magnetic ink character reader 301
When 21,41 are stored in the candidate character storage unit 52, the second step of character recognition is started according to the word generation processing procedure of FIG.

That is, in FIG. 5, the character reading processing unit 13 notifies the candidate character selection unit 5 in a processing step 90 when the storage of the candidate character groups 21, 41 in the candidate character storage unit 52 is completed. In processing step 91, the candidate character selection unit 5 selects the character having the first rank (represented by C1 and C2, respectively) from among the candidate characters in the first digit of the candidate character groups 21 and 41, respectively. , (C), the characters are both "ku" and "ku").

Candidate character group extracted in processing step 92
Search the candidate character group 41 using the 21 candidate characters (C1) as a key,
The same character as the character is extracted, and the extracted character storage unit 53 as the extracted character 45 together with the rank information of the extracted character.
Store to.

Similarly, in processing step 93, the candidate character group 21 is searched by using the extracted character (C2) of the candidate character group 41 as a key, the same character as the character is extracted, and the extracted character has Extracted character storage as extracted character 25 with link information
Store in 53.

Although the generation of the first-digit extracted character is completed, the second- and third-digit candidate character groups 2 are generated.
The extracted characters are similarly generated for 1 and 41, and are stored in the extracted character storage unit 53 in processing step 94.

That is, as shown in the character extraction from the candidate characters and the candidate words in FIG. 6, the extracted character 45 (indicated by * 1 mark in FIG. 6A) and the extracted character 25 (see FIG. ) Is shown as an extraction character.

In processing step 95, the candidate character selection section 5 notifies the word generation 55 of the word generation. The word generation unit 55 reads out the extracted characters 45 and 25 for each digit, and if they are the same character,
If the characters are not the same character, the rank values of the respective characters are compared. If the characters have the same rank, the two characters are selected. Then, the candidate words 51-1 and 2 (FIG. 6B) composed of the character strings of the combinations are generated and stored in the candidate word storage unit 56.

In processing step 97, the knowledge processing unit 59 reads out the candidate words 51-1, 2 from the candidate word storage unit 56, and the candidate word 51
By searching the knowledge dictionary file 57 with -1, 2 as a key, the registered name is matched, and the matched name is displayed on the display device 58, for example.

[0025]

The character information read and recognized by the OCR reading processing unit 10 and the MICR reading processing unit 30 is the candidate character group 2
It is output to the candidate character selection unit 5 as 1,41, and the recognition character selection unit 5 selects at most two digit-by-digit recognized characters constituting the candidate word 51 based on the candidate character group 21,41. , It is possible to greatly improve the processing efficiency of word search.

[Brief description of drawings]

FIG. 1 is a block diagram of the principle of the present invention.

FIG. 2 is a system configuration diagram of an embodiment.

FIG. 3 is an explanatory diagram of generation of character information and candidate characters.

FIG. 4 Generation procedure of candidate characters

FIG. 5: Word generation processing procedure

FIG. 6 Character extraction from candidate characters and candidate words

FIG. 7 is an explanatory diagram of conventional character recognition.

[Explanation of symbols]

 1, 3 is a reading unit, 11 is an OCR unit, 31 is a MICR unit, 10 is an OCR reading processing unit, 13 is a character reading processing unit, 101 is an optical character reading device, 2 and 4 are recognition processing units, and 21,41 are Candidate character group, 25 and 45 are extracted characters, 30 is MICR reading processing unit, 42 is character information, 44 is form, 5 is candidate character selecting unit, 51, 51-1, 51-2 are candidate words, 52 is candidate The character storage unit 53 is an extracted character storage unit, 55 is a word generation unit, 56 is a candidate word storage unit, 57 is a knowledge file, 58 is a display device, 59 is a knowledge processing unit, C1 and C2 are candidate characters, and 80 to 87 and 90 to 97 represent processing steps.

Claims (2)

[Claims] 1. A character recognition device having a plurality of reading units (1,3) according to the attributes of ink forming characters on a form, wherein the character information read by each reading unit (1,3). Recognize
A recognition processing unit (2,4) that generates a plurality of ranked candidate character groups (21,41) is compared with the candidate character group characters (21,41), and the same character is used for both candidate character groups. A character recognition device, comprising: a word generation unit (55) for determining whether or not a word exists and generating a most probable candidate word (51). 2. The character recognition device according to claim 1, further comprising a sensor for detecting a predetermined attribute of ink, and when the sensor detects the ink attribute, the character is read by the attribute. Recognition device.