JPH0317150B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention particularly relates to an optical character reading device having a function of reducing the occurrence of misreading.

[Technical background of the invention and its problems]

In optical character reading devices (hereinafter referred to as OCR), improving the reading rate and preventing misreading are extremely important in determining the performance. Conventionally, as a method for preventing misreading, a method has been developed in which the same character pattern is read twice.

In the above-mentioned OCR, multiple types (for example, two types) of slice level values are usually set as standards for determining white and black in the photoelectric conversion unit, and based on these multiple types of slice level values, the same character is A plurality of types of binarization patterns are created. and,
Recognition processing is performed for each binarization pattern,
Comparison processing is performed between the respective recognition results. Based on the comparison result, if they match, the recognition result is output as an answer, and if they do not match, it is rejected.

However, with conventional OCR as mentioned above,
In order to create a plurality of types of binarization patterns and store each binarization pattern, a plurality of binarization pattern storage memories are required. This has resulted in drawbacks such as the overall configuration of the OCR becoming complex and the cost increasing due to the large number of parts such as memory.

[Purpose of the invention]

The present invention was made in view of the above circumstances, and
The purpose is to provide an optical character reading device that has a simple configuration and can significantly reduce misrecognition of characters to be read.

[Summary of the invention]

In the present invention, a subset table memory is provided which stores in advance combination information of recognition subsets corresponding to a plurality of recognition operations. The recognition control means controls the recognition means to perform matching processing multiple times on the same character pattern based on the combination information of the recognition subsets. Further, a plurality of recognition results for the same character pattern outputted from the recognition means are compared by a comparison means.
The comparison means outputs the recognition result as an answer when the comparison results match, and outputs a reject signal when they do not match. Thereby, the character pattern is recognized multiple times based on a predetermined combination of recognition subsets.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a partial configuration of an OCR according to an embodiment. In FIG. 1, a recognition unit control circuit (hereinafter simply referred to as a control circuit) 10 is a circuit that controls the entire recognition unit of OCR. The character cutout pattern memory 11 stores a character pattern P for one character in response to the control signal 2 of the control circuit 10. This character pattern P is sent from memory 11 to matching circuit 12. A matching circuit 12 matches a character pattern P and a standard pattern memory (dictionary memory) 1 according to a control signal m from a control circuit 10.
Matching processing with standard patterns FP1 to FP4 from 3 is performed. The standard pattern memory 13 stores in advance standard patterns FP1 to FP4 each consisting of a subset of numbers (0 to 9), alphabetic characters (A to Z), kana characters, special symbols, etc., and a dictionary from the control circuit 10. Standard patterns FP1 to FP are sent to the matching circuit 12 according to the load signals _b1 to _b4.
Outputs 4. The control circuit 10 outputs the dictionary load signals b ₁ to b ₄ to the standard pattern memory 13 based on the recognition subset combination information stored in the subset table memory 14 in advance. In this case, the control circuit 10 outputs dictionary load signals b ₁ to b ₄ in response to an instruction signal CO from the matching circuit 12 instructing the number of recognition processing.
On the other hand, the comparator 15 compares a plurality of recognition results R1 and R2 for the same character pattern output from the matching circuit 12, and if the comparison results match, outputs the recognition result (R1 or R2) as an answer. Further, if the comparison result is a non-coincidence, a reject signal is output.

The operation of the OCR configured as described above will be explained. First, when the control circuit 10 receives a drive signal d from the outside (usually a control section for the entire OCR), it outputs a control signal a to the character cutting pattern memory 11. In response to this control signal a, a character pattern P for one character is cut out and stored in the character cutout pattern memory 11. In this case, the character pattern P is cut out character by character from a line buffer that normally stores character patterns for one line. Furthermore, the control circuit 10 is supplied with subset information E (for example, alphanumeric characters, kana characters, etc.) necessary for recognition processing of the character pattern P from the outside. The control circuit 10 includes a memory 14
A subset necessary for the first recognition operation is determined by referring to the subset table in the dictionary, and dictionary load signals b ₁ to b ₄ based on the subset are output to the standard pattern memory 13 .

Here, the subset table memory 14 stores a subset table as shown in FIG. 2, for example, and stores combination information of subsets corresponding to the number of recognition operations. That is, for example, if the numerical subset is N and the alphabetic subset is A, it is assumed that the numerical subset information E of the control circuit 10 is given. Therefore, as is clear from FIG. 2, the control circuit 10 sets a subset N in the first recognition operation and outputs a dictionary load signal _b1 corresponding to the subset N to the standard pattern memory 13. Standard pattern memory 1
3, pre-stored standard patterns FP1 to FP4
From there, a dictionary load signal _b1 , that is, a standard pattern FP1 corresponding to the subset N is set and sent to the matching circuit 12. In the matching circuit 12, matching processing is performed between the character pattern P from the character cutting pattern memory 11 and the standard pattern (consisting of numeric subset N) FP1 in response to the control signal m from the control circuit 10. The first recognition result R1 by this matching process
is given to the comparator 15. When the matching circuit 12 completes the first matching process,
The instruction signal CO is supplied to the control section 14.

The control unit 10 controls the second recognition operation for the same character pattern P in response to receiving the instruction signal CO. That is, the control section 1
0 determines a subset NA that is a combination _of numeric subset _N and alphabetic subset A using the subset table shown in FIG.
Output to 3. As a result, the standard pattern corresponding to the subset NA is stored from the standard pattern memory 13.
FP1 and FP2 are applied to a matching circuit 12. Then, the matching circuit 12 uses the same character pattern P and the standard pattern FP from the standard pattern memory 13 as in the first matching process.
1 and FP2 is performed, and the second recognition result R2 is output to the comparator 15. In the comparator 15, each recognition result R
1 and R2, and if the comparison result is a match, the recognition result R1 or R2 is output as the answer. Further, if the comparison result is a mismatch, a reject signal is output.

As described above, the same character pattern P is recognized twice, for example. In this case, a standard pattern is set for each recognition process by a combination of subsets set in the subset table. Specifically, for example, if the letters "B", "S", and "T" are written in a numeric field on a form, the first recognition process uses numeric subset N (0 to 9). Since matching processing is performed based on
There is a high possibility that it will be recognized as "5" or "7". moreover,
In the second recognition process, the number subset N (0 to 9)
Matching processing is performed based on subset NA, which is a combination of alphabetical character subsets A (A to Z). Therefore, in the second recognition process, the alphabetic characters "B", "S", and "T" are recognized. Then, as described above, a comparison process is performed between the first and second recognition results R1 and R2,
Since the comparison results do not match, a reject signal is output as the final output. Therefore, if an alphabetic character is mistakenly written in place of a number on a form, conventionally it would be read as a mistaken number depending on the numeric subset, but in the present invention, it is rejected as described above, It is possible to prevent a situation where the information is read by mistake. Additionally, if the numbers "1", "2", and "5" are written in non-standard fonts on a form, there is a possibility that they may be misread as "7", "0", "6", etc. in the past. many. Even in such a case, in the second recognition process of the present invention, the alphabetic characters "I", "Q",
Since there is a high possibility that "S" will be recognized, the first and second recognition results are compared for "7" and "I", "0" and "Q", and "6" and "S", respectively. As a result, the comparison results will be inconsistent. Therefore, it is possible to prevent the message from being rejected and mistakenly recognized.

In the above embodiment, the combination of subsets in the subset table memory 14 is a combination of numbers and letters, but the combination is not limited to this and other combinations may be used. For example, as shown in FIG. 2, a combination of a kana character subset K and a special symbol subset S may be used.

〔Effect of the invention〕

As described in detail above, according to the present invention, recognition processing is performed multiple times on the same character pattern based on combinations of recognition subsets with a simple configuration, and characters that are likely to be misread are misrecognized. can be significantly reduced. Therefore, it is possible to provide an optical character reading device that can reduce the occurrence of misreading and realize a reliable reading operation without particularly increasing the number of components such as a memory for storing binarized patterns.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a partial configuration of an optical character reading device according to an embodiment of the present invention, and FIG. 2 is a diagram showing an example of a subset table stored in the subset table memory of FIG. 1. It is. 10... Recognition unit control circuit, 11... Character cutout pattern memory, 12... Matching circuit, 1
3...Standard pattern memory, 14...Subset table memory, 15...Comparator.

Claims (1)

[Claims] 1. A character pattern memory that stores a character pattern for one character, a subset table memory that stores combination information of recognition subsets according to recognition operations performed multiple times in advance, and a character pattern memory that stores character patterns in the character pattern memory and the aforementioned subsets. recognition means for outputting a recognition result of the character pattern through matching processing with a standard pattern selected based on the combination information of the recognition subset in the set table memory;
recognition control means for controlling the recognition means to perform matching processing multiple times based on combination information of the recognition subsets set in advance for the same character pattern; and the same character output from the recognition means. An optical character reader characterized by comprising: comparing means for comparing a plurality of recognition results for a pattern and outputting a recognition result as an answer when the comparison results match, and outputting a reject signal when they do not match. Device.