JPS60579A - Character recognizing method - Google Patents

Abstract

PURPOSE:To improve the rate of character recognition by recognizing a character by correlating the read character to a character that corresponds to the width of the letter. CONSTITUTION:A microprocessing unit MPU11 sends a card in the lateral direction at a constant speed when it senses a card detection signal by a sensor 16. A magnetic head for reading 18 acans the card relatively and reads magnetic ink. The reading signal from the head 18 passes through an amplifier 19 and an LPF20 and converted to a digital signal by an A/D converter 21 and stored once in an RAM22. Then, the MPU11 detects a starting point of the character from read data stored in the RAM22, and at the same time, detects end point of the letter referring to the starting point, and detects the width of the character. The MPU11 compares the width with stored reference code, and when they coincide, stores it in the RAM22. Thus, by recognizing a character to be read by correlating it only with a character corresponding to the width of the character, erroneous reading can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character recognition method for reading and recognizing characters.

The character recognition method is to read the characters, pay attention to their changing points, create a character code, and sequentially compare this character code with all the fixed codes on the fixed code table to recognize the read character file. There is. However, in this method, a read character is recognized by comparing the character code with all fixed codes, so misreading may occur and the character recognition rate is low.

SUMMARY OF THE INVENTION An object of the present invention is to provide a character recognition method in which a read character is recognized by comparing it only with characters corresponding to the character width in order to reduce misreading.

The present invention will be described in detail below with reference to the drawings.

To explain the present invention with reference to FIG. 1, a character whose character width is determined to be one of a plurality of specified widths is read, the starting point of the character is detected from the output, and the ending point of the character is detected based on the starting point. is detected to detect the character width of the character, and the read character is recognized by comparing it with a character corresponding to the detected character width.

FIG. 2 shows an example of equipment used in practicing the invention.

In this example, magnetic ink characters (E13B) are read and recognized from a medium such as a card such as a check on which magnetic ink characters (E13B) are recorded at regular intervals in the horizontal direction, and the micro processing unit (hereinafter referred to as MPU) 1
When a read command is input from an external terminal device or the like via an interface 12 and a peripheral interface adapter (hereinafter referred to as PIA) 13, the read command is displayed on the display circuit 15 via the PIA 14 as a ready state.

When a card having magnetic ink characters is inserted into the entrance in this state, this card is detected by the sensor 16. When the MPU II senses a card detection signal from the sensor 16 via the PIA 14, it operates the motor control circuit 17 via the PIA 14 to rotate the motor at a constant speed, thereby stabilizing the card in the horizontal direction with the motor. Rapid feed is made and the display circuit 15 is made to display that it is in operation. A reading magnetic head 18 scans the card at a constant speed relative to the fixed position to read the magnetic ink characters. The read signal from the magnetic head 18 becomes a differential waveform representing the amount of magnetic change and is amplified by an amplifier 19, noise is removed through a low-pass filter 20, and converted into a digital signal by an analog/digital converter 21. MPUII
takes in the digital signal from the analog/digital converter 21 via the PIA 14 at predetermined time intervals (time intervals corresponding to dividing one character into eight blocks in the card feeding direction and sampling one block eight times). De-H random access memory (hereinafter referred to as RAM) 2
2, perform the processing described below to recognize the read characters, and send the recognized characters to the PIA 13 and the interface 1.
2 to an external terminal device or the like for display or predetermined processing.

The MPU 11 performs predetermined processing according to a program in a read-only memory (hereinafter referred to as ROM) 23.
The main processing is specifically performed as shown in FIGS. 3 to 7. That is, when the card is running, as shown in FIG. 3, digital signals from the analog/digital converter 21 are read at the above-mentioned time intervals and are sequentially stored in consecutive addresses of the RBUF in the RAM 22 as read data. After reading all the read data for one card and storing it in the RAM 22, character recognition starts. First, in order to detect the starting point of one character from the read data, the beginning of the character is detected and the recognition of that character is performed. start.

A positive peak value and its address, and a negative peak value and its address are detected from the data of one character in the read data (data for one character from the first character detected above) and stored separately in the RAM 22. Plus table (PTB) and minus table (MTB)
make. Here, the RAM 22 has a read buffer (RBUF) in which read data is stored, as shown in FIG.
B.

MTB, plus peak table (PPTB), minus peak table (MPTB), etc. are assigned. Next, the contents of PTB and MTB are rearranged in descending order of peak value and stored in PPTB and MPTB. In this example, the first plus peak value of PTB and its address are first stored in PPTB without being rearranged.)
B. 2 from the contents of MPTB in order of the largest peak value
One or all three of those addresses (PPTB, M
A character code is created using the PTB address (RBUF address) and stored in the code table of the RAM 22. In this case, the plus peak address and the minus peak address are the plus peak and minus peak positions of the read waveform obtained from the magnetic head 18 when the magnetic ink characters are scanned by moving the card and read by the magnetic head 18. Since they are compatible, character codes can be created by looking at how far apart the first plus peak address is from other plus peak addresses and minus peak addresses, and capturing the characteristics of the characters, and storing them in the code table. Store.

However, the magnetic ink characters are constructed so that the first plus peak appears in the first block of every character. Next, the last negative peak is detected to detect the character width. Here, the first positive peak of the read data is caused by the character start point, and the last negative peak is caused by the character end point, but the magnetic ink character E13B has four types of character widths and four types of positions of the last negative peak. become. ROM2
A plurality of fixed codes are stored in advance in the fixed code table No. 3, and the above character code corresponds to the character width detected above among the fixed codes on the fixed code table (those having the character width detected above). If it matches any fixed code, that code is determined to be the code of the read character and is set in the RAM 22, and if there is no match, it is determined that it is unrecognizable. The code of RAM2
2 cents, and in the same manner, each character is sequentially recognized from the read data.

FIG. 4 is a flowchart specifically showing the program for character order detection and peak detection in FIG. RBU
Read the read data sequentially from F and detect the starting point of the data, i.e. the point where the read data first exceeds a predetermined level for each character or 25% of the maximum peak value of the previous character. It is detected as the starting point, and the address of the starting end of this data is set as the character first address (MHAD).
) and set MHAD + 70 (hexadecimal) as the character end address (MEAD). and R.B.U.
MI of F (Data from AD to MEAD is taken as one character and its peak value is detected. However, as a noise countermeasure, data below a certain level is not subject to peak value detection. The detected peak value is a plus peak. If it is a value, set this plus peak value and its address in the PTB,
If the detected peak value is a negative crack value, this negative peak value and its address are sent to the MTB.

FIG. 5 is a flowchart specifically showing the part of FIG. 3 in which the contents of PTB and MTB are rearranged in descending order of peak values. First, the contents of the first address of PTB (
The first plus peak value and its address) are set in the PPTB, and all the contents of other addresses in the PTB are transferred to the peak data buffer (PDBF). and P.D.B.
Find the one with the thickest peak value from the contents of F, erase this maximum peak value and its address from PDBF, and set it in PPTB. By repeating this operation until PDBF is empty, the second largest peak value in PPTB is found. The beam values and their addresses are set in subsequent addresses in order of increasing peak value.

In addition, after all the contents of the MTB are transferred to the PDBF, they are similarly sent in the order of the highest peak value.

FIG. 6 is a flowchart showing the part of FIG. 3 in which a character code is created. One character is divided into 8 blocks and each block is read by 8 samplings, but the first half of the first block (4 samplings) and the second half (4 samplings) of the last block are always blank, so one character is essentially It is 7 blocks long. Furthermore, the read data is such that a peak appears only once in one block for any character, and each character always has two or three normal plus peaks and minus peaks. Furthermore, the magnetic ink characters are constructed so that the first plus peak of any character appears at the center of the first block. So first, set data < 5ETD) = 80 (binary 1.0000
000), set the plus code (PCODE) to 80, and enter the address PPTB(1) of the first plus peak value stored first and second in PPTB.
The difference between the address PPTB(2) and the address PPTB(2) having the largest plus peak value among the other plus peak values is calculated, 4 (half block) is subtracted from the difference, and the result is divided by 8. The result of this calculation, Z, when rounded down to the decimal point, indicates how many blocks apart the first plus peak is from the largest glass peak among the other plus peaks.

So, if we set 5ETD to 40 and move it to the right by Z, 5ETD becomes data that shows how many blocks away the largest positive peak among the other positive peaks is from the first positive peak, and we OR it with PCODE. and set it as the first plus code in the code table. Address P of the third plus peak value in PPTB
Similarly, if PTB(3) exists, P P T B(1) and P
Calculate the difference with PTB (3), subtract 4 from the difference, and get 8
Divide by and cut off the decimal point. Then, if we cent 5ETD to 40 (01000000 in binary) and shift it to the right by the calculation result Z, we can find out how many blocks apart the second largest positive peak among the other positive peaks is from the first positive peak. The data shown is obtained. This data is ORed with PCODE (first plus code) and set as a second plus code in the code table. Every character has two or three regular plus peaks, so if you don't have PPTB(3), set the first plus code in the code table and finish creating the plus code. (4,) Since the following is unnecessary, the second plus code is placed in the code table and the creation of the plus code is completed.

Following the creation of the neck plus code, P P T B (1)
Create a negative code in the same way based on , that is, take two or three negative peak values in descending order of negative peak values and set their addresses MP T B (1), MP T B (2).
), create a minus code by looking at how well MP TB (3) and P P T B (1) are done, and
Cent in the code table as shown. This plus code and minus code constitute a character code as a pair.

FIG. 7 shows a portion of FIG. 3 in which the last negative peak is detected, the character width is detected, and the character code is compared with a fixed code corresponding to the character width to recognize the read character.

The magnetic ink characters E13B are 0, 1.2. -9. A,
It consists of 14 types of characters B, C, and D, and the character width corresponds to the width at any of the 5th to 8th blocks from the first plotter to any of the 8 blocks.

That is, the last positive peak is 0.8. A, B, C,
For each letter D, it becomes the 8th block, and 4. ．． Each character of 6.9 becomes the 7th block, each character of 3, 5.7 becomes the 6th block, and each character of ■, 2 becomes the 5th block.

Then, use the minus code (MCOD) (if there are two, use the second minus code) OL 02 (binary 10)
.

The character width is detected by performing an AND with 04 (binary number 100) and 08 (binary number 1.000) and detecting which of the 5th to 8th blocks the last negative peak is located. When it is detected that the last negative peak is in the 8th block, each character code consisting of each combination of a positive code and a negative code on the code table is changed to 0.8 on the fixed code table. A, B,
Compare it with the fixed codes of C and D, and if it matches any fixed code, set that code in the RAM 22, and if there is no match? is set in RAM22. When it is detected that the last negative peak is in the 7th block, the character code consisting of a combination of a plus code and a minus code on the code table is compared with the fixed code of 4.6.9 on the fixed code table. If it matches any fixed code, set that code to RA and M22, and if there is no match? is set in RAM 22.

When it is detected that the last negative peak is in the 6th block, compare the character code consisting of the combination of the positive code table and the negative code on the code table with the fixed code in 3.5.7 of the fixed code table. If it matches any fixed code, send that code to RA.
What if I set it to M22 and it doesn't match? RAM22
Set to . When it is detected that the last negative peak is in the 5th block, the character code consisting of the combination of the plus code and the minus code on the code table is compared with the fixed code of 1°2 on the fixed code table, and one of the characters is selected. If it matches the fixed code of
What if I set it to M22 and it doesn't match? RAM22
Set to .

As described above, according to the present invention, since a read character is recognized by comparing it only with characters corresponding to the character width, it is possible to reduce misreading and improve the character recognition rate. Furthermore, since the character width is detected from the read output, the character width can be detected accurately and recognition accuracy can be increased.

[Brief explanation of drawings]

FIG. 1 is a flowchart showing the present invention, FIG. 2 is a block diagram showing an example of a device used to implement the present invention, and FIGS. 3 to 7 are flowcharts showing the program of the device.
FIG. 8 is a diagram showing a memory map of the RAM in the same device, and FIG. 9 is a diagram showing a code table of the RAM. 11-・MPU, 18 magnetic head, 22RAM,...
23 ・ROM0! shi' 弗b Figure (-) Procedural amendment (direction) October 4, 1982 Patent Application No. 108955 2 Title of invention character recognition method 3 Related case to be amended Name of patent applicant (223 ) Sankyosei R Seisakusho Co., Ltd. (1 person) 4 Representative Director Address 4-5-4-6 Kyodo, Seiftani-ku, Tokyo Subject of amendment

Claims (1)

[Claims] A character recognition method for reading and recognizing characters whose character width is determined to be one of a plurality of specified widths, comprising a storage means for storing a standard code of each character corresponding to the character width,
The start point of a character is detected from the read output, the end point of the character is detected based on the start point, the character width of the character is detected, and the read character is stored in the storage means corresponding to the detected character width. A character recognition method that recognizes characters by comparing them with a standard code.