JP2012174093A - Recording medium processing apparatus, and method and program for controlling recording medium processing apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To suppress erroneous recognition in recognizing magnetic ink characters.SOLUTION: In a check reader 1 equipped with a magnetic head 54 magnetically reading a magnetic ink character recorded on a check 4, a control unit 71 executes: a first recognition phase in which candidates of a character type of the magnetic ink character are selected on the basis of a difference amount calculated in accordance with a difference between a detection waveform obtained by reading the magnetic ink character and a waveform obtained by sliding a reference waveform by a prescribed distance; and a second recognition phase in which candidates of the character type of the magnetic ink character are selected on the basis of a difference amount calculated in accordance with a difference between the detection waveform and a waveform obtained by expanding/contracting the reference waveform by a prescribed distance. Candidates of the character type of the magnetic ink character are determined by comparing the difference amounts of the character types selected as the candidates in the first recognition phase and the difference amounts of the character types selected as the candidates in the second recognition phase.

Description

  The present invention relates to a recording medium processing apparatus, a control method for the recording medium processing apparatus, and a program.

A recording medium processing device (check device reader) that includes a magnetic head for reading magnetic ink characters recorded on a recording medium such as a check, and that reads magnetic ink characters on a recording medium conveyed on a conveyance path and recognizes the magnetic ink characters ) Is known (see, for example, Patent Document 1).
In such a recording medium processing apparatus, a magnetic ink character recognition is performed by comparing a detection waveform obtained by reading a magnetic ink character with a reference waveform of a character type determined by the standard, and the character type of the magnetic ink character. Is determined.

JP 2004-206362 A

However, in magnetic ink character recognition, the position of the detected waveform is deviated from the ideal waveform due to misprinting of the magnetic ink character, the state of conveyance of the recording medium, reading of the magnetic ink by the magnetic head, etc. Or the detected waveform may be stretched or contracted.
The result of magnetic ink character recognition by the recording medium processing apparatus described above is used, for example, when various processes are performed on a check. For this reason, when recognizing magnetic ink characters, there is a need to suppress misrecognition by eliminating the influence of the displacement of the detection waveform position and the expansion and contraction of the detection waveform as much as possible.

  SUMMARY An advantage of some aspects of the invention is to solve at least a part of the problems described above, and the invention can be implemented as the following forms or application examples.

  Application Example 1 A recording medium processing apparatus according to this application example stores a reading unit that magnetically reads magnetic ink characters recorded on a recording medium and a reference waveform of a plurality of character types defined by a standard. And a control unit that controls the reading unit and the storage unit, wherein the control unit is stored in the storage unit and a detection waveform obtained by the reading unit reading the magnetic ink characters. A difference amount corresponding to a difference between a plurality of reference waveforms of a plurality of character types and a waveform obtained by sliding each of the reference waveforms by a predetermined distance is calculated, and character type candidates for the magnetic ink characters are selected based on the calculated difference amount. A difference amount according to a difference between the first recognition phase, the detected waveform, and a waveform obtained by expanding and contracting each of the reference waveforms of the plurality of character types by a predetermined distance is calculated, and based on the calculated difference amount Magnetic ink statement A second recognition phase for selecting the character type candidates of the character type, the difference amount of the character types selected as the candidates in the first recognition phase, and the candidate selected in the second recognition phase. The difference amount of the character type is compared to determine a character type candidate for the magnetic ink character.

According to this configuration, in the first recognition phase, a detection waveform obtained by reading a magnetic ink character is compared with a waveform obtained by sliding each of a plurality of character type reference waveforms defined in the standard by a predetermined distance. Therefore, when the position of the detected waveform is deviated from the ideal waveform, magnetic ink characters can be recognized while suppressing the influence of the deviation. In the second recognition phase, the detected waveform is compared with a waveform obtained by expanding and contracting each of the reference waveforms of a plurality of character types by a predetermined distance. Therefore, when the detected waveform is expanded or contracted, the influence of the expansion or contraction is generated. It is possible to recognize magnetic ink characters while suppressing the above.
Further, the character type candidate selected in the first recognition phase is compared with the character type difference selected as the candidate in the second recognition phase to determine a character type candidate for the magnetic ink character. To do. For this reason, it is possible to select a character type candidate that is considered to be more reliable from the results of magnetic ink character recognition by the two different methods described above, and thus it is possible to suppress erroneous recognition of magnetic ink characters. Become.

  [Application Example 2] The recording medium processing apparatus according to the application example, wherein the control unit includes at least the magnetic field in the first recognition phase and the second recognition phase in ascending order of the calculated difference amount. The first candidate and the second candidate of the character type of the ink character are selected, and the difference amount of each of the character types selected as the first candidate and the second candidate in the first recognition phase; The first candidate and the difference amount of each of the character types selected as the second candidate in the second recognition phase are compared, and the character type of the magnetic ink character is ordered in ascending order of the difference amount. It is preferable to determine the first candidate and the second candidate.

  According to this configuration, the difference amount is further increased from the first candidate and the character type selected as the second candidate in the order of small difference amount in each of the first recognition phase and the second recognition phase having different recognition methods. Since the first candidate and the second candidate are determined in ascending order, erroneous recognition of magnetic ink characters can be effectively suppressed.

  Application Example 3 In the recording medium processing apparatus according to the application example described above, the control unit determines that the difference amount of the character type selected as the first candidate in the first recognition phase is the second recognition phase. If the difference is smaller than the difference amount of the character type selected as the first candidate, the reference waveform of the plurality of character types is slid by the distance slid in the first recognition phase. The second recognition phase is executed, the first candidate in the second recognition phase for the second time, and the difference amount of each character type selected as the second candidate, and the first amount in the first recognition phase. The candidate and the difference amount of each of the character types selected as the second candidate are compared, and the first candidate of the character type of the magnetic ink character and the second are in order of increasing the difference amount. Weather It is preferable to determine.

  According to this configuration, when the difference amount of the character type selected as the first candidate in the first recognition phase is smaller than the difference amount of the character type selected as the first candidate in the second recognition phase, the detected waveform It is considered that there is a high possibility that the position of is shifted. Therefore, by re-running the second recognition phase after sliding for the distance slid in the first recognition phase, the influence of both the displacement and expansion / contraction of the detected waveform is suppressed, and magnetic ink character recognition is performed. It can be performed. Thereby, misrecognition can be suppressed more effectively.

  Application Example 4 In the recording medium processing apparatus according to the application example, in the second recognition phase for the second time, the font of the magnetic ink character is an E-13B font, and the first recognition phase It is preferable to execute the processing except when the character type selected as the first candidate and the character type selected as the second candidate are “2” and “5”.

  According to this configuration, the character types “2” and “5” in the E-13B font are similar to each other on the one end side and the other end side of the detection waveform obtained by reading the magnetic ink characters. If the waveform is slid, “2” and “5” may be erroneously recognized. Therefore, when the character types selected as the first candidate and the second candidate in the first recognition phase are “2” and “5”, the second second recognition is performed after sliding the waveform. By not executing the phase, erroneous recognition of “2” and “5” can be suppressed.

  Application Example 5 A control method for a recording medium processing apparatus according to this application example is a control method for a recording medium processing apparatus including a reading unit that magnetically reads magnetic ink characters recorded on a recording medium, A difference amount corresponding to a difference between a detection waveform obtained by reading the magnetic ink character by the reading unit and a waveform obtained by sliding a reference waveform of a plurality of character types defined in the standard by a predetermined distance is calculated. A first recognition phase for selecting character type candidates of the magnetic ink character based on the calculated difference amount, a waveform obtained by expanding and contracting each of the detection waveform and the reference waveforms of the plurality of character types by a predetermined distance. And a second recognition phase for selecting a character type candidate of the magnetic ink character based on the calculated difference amount, and performing the previous in the first recognition phase. The difference amount of the character type selected as a candidate is compared with the difference amount of the character type selected as the candidate in the second recognition phase to determine a character type candidate of the magnetic ink character. It is characterized by that.

According to this method, in the first recognition phase, a detection waveform obtained by reading a magnetic ink character is compared with a waveform obtained by sliding each of a plurality of character type reference waveforms defined in the standard by a predetermined distance. Therefore, when the position of the detected waveform is deviated from the ideal waveform, magnetic ink characters can be recognized while suppressing the influence of the deviation. In the second recognition phase, the detected waveform is compared with a waveform obtained by expanding and contracting each of the reference waveforms of a plurality of character types by a predetermined distance. Therefore, when the detected waveform is expanded or contracted, the influence of the expansion or contraction is generated. It is possible to recognize magnetic ink characters while suppressing the above.
Further, the character type candidate selected in the first recognition phase is compared with the character type difference selected as the candidate in the second recognition phase to determine a character type candidate for the magnetic ink character. To do. Therefore, since it is possible to select a character type candidate that is considered to be more reliable from the results of magnetic ink character recognition by two different methods, it is possible to suppress erroneous recognition of magnetic ink characters.

  Application Example 6 A program according to this application example is a program executed by a control unit that controls each unit of a recording medium processing apparatus including a reading unit that magnetically reads magnetic ink characters recorded on a recording medium. The control unit determines a difference according to a difference between a detection waveform obtained by reading the magnetic ink character and a waveform obtained by sliding each of a plurality of character type reference waveforms defined in the standard by a predetermined distance. A first recognition phase for calculating a quantity, and selecting a character type candidate of the magnetic ink character based on the calculated difference quantity, the detected waveform, and the reference waveforms of the plurality of character types are each set at a predetermined distance. A second recognition phase for calculating a difference amount according to a difference between the expanded and contracted waveform and selecting a character type candidate of the magnetic ink character based on the calculated difference amount; and The difference amount of the character type selected as the candidate in the first recognition phase is compared with the difference amount of the character type selected as the candidate in the second recognition phase. It is made to function as a means to determine the character type candidate.

According to this configuration, in the first recognition phase, a detection waveform obtained by reading a magnetic ink character is compared with a waveform obtained by sliding each of a plurality of character type reference waveforms defined in the standard by a predetermined distance. Therefore, when the position of the detected waveform is deviated from the ideal waveform, magnetic ink characters can be recognized while suppressing the influence of the deviation. In the second recognition phase, the detected waveform is compared with a waveform obtained by expanding and contracting each of the reference waveforms of a plurality of character types by a predetermined distance. Therefore, when the detected waveform is expanded or contracted, the influence of the expansion or contraction is generated. It is possible to recognize magnetic ink characters while suppressing the above.
Further, the character type candidate selected in the first recognition phase is compared with the character type difference selected as the candidate in the second recognition phase to determine a character type candidate for the magnetic ink character. To do. Therefore, since it is possible to select a character type candidate that is considered to be more reliable from the results of magnetic ink character recognition by two different methods, it is possible to suppress erroneous recognition of magnetic ink characters.

1 is an external perspective view of a check reading apparatus according to the present embodiment. It is a figure which shows the internal structure of a check reading apparatus. It is a block diagram which shows the functional structure of a check reading apparatus. It is a flowchart which shows operation | movement of the character recognition part which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the character recognition part which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the character recognition part which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the character recognition part which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the character recognition part which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the character recognition part which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the character recognition part which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the character recognition part which concerns on 1st Embodiment. It is a figure which shows an example of detection signal waveform data. It is a figure which shows an example of character waveform data. It is a figure for demonstrating a difference amount. It is a figure which shows the data structure of an expansion-contraction table. It is a figure which shows the expanded reference waveform data. (A) shows a contraction table, (B) is a figure which shows an extension table. It is a flowchart which shows operation | movement of the character recognition part which concerns on 2nd Embodiment. It is a flowchart which shows operation | movement of the character recognition part which concerns on 3rd Embodiment. (A) schematically shows the content of reference waveform data corresponding to the character type “2” in the E-13B font, and (B) shows the content of reference waveform data corresponding to the character type “5” in the E-13B font. FIG.

  Hereinafter, a recording medium processing apparatus, a control method for the recording medium processing apparatus, and a program according to an embodiment of the present invention will be described with reference to the drawings. In each drawing referred to, in order to show the configuration in an easy-to-understand manner, the dimensional ratio, angle, and the like of each component may differ.

(First embodiment)
<Recording medium processing apparatus>
A recording medium processing apparatus according to the first embodiment will be described. The recording medium processing apparatus according to this embodiment includes a check reading apparatus 1 and a host computer 70.

First, a schematic configuration of the check reading apparatus 1 according to the first embodiment will be described. FIG. 1 is an external perspective view of a check reading apparatus 1 according to this embodiment.
The check reading device 1 reads a magnetic ink character (MICR character) recorded on the check 4, reads images on both sides of the check 4, and endorses the check 4 with respect to the check 4 that is a sheet-like recording medium. It is an apparatus that performs processing such as recording of a predetermined image.
The check reading device 1 includes a main body case 2 and a lid case 3 placed on the upper side of the main body case 2, and each component is incorporated therein.
The cover case 3 is formed with a conveyance path 5 of a check 4 consisting of a narrow vertical groove having a U shape when viewed from above, and one end of the conveyance path 5 is a check consisting of a wide vertical groove. The other end of the conveyance path 5 branches right and left and is connected to a first check discharge portion 7 and a second check discharge portion 8 each having a wide vertical groove.

On the surface 4a of the check 4, a money amount, a payer, a number, a signature, and the like are written on the background on which a predetermined pattern is formed, and the lower end portion thereof has a long side direction as shown in FIG. An extended magnetic ink character string 4A is recorded. The magnetic ink character string 4A is formed by arranging a plurality of magnetic ink characters in the horizontal direction.
In addition, an endorsement column is formed on the back surface 4 b of the check 4. A predetermined image related to the endorsement is recorded in the endorsement column by a recording device 56 described later.
The check 4 is aligned in the vertical direction so that the upper end 4e is positioned upward and the lower end 4f is positioned downward, and the front and back are aligned so that the surface 4a faces the outside of the U-shaped conveyance path 5 ( In the state shown in FIG. 1, it is inserted into the check supply unit 6. The check 4 inserted into the check supply unit 6 is sent out to the transport path 5 with the rear end 4d as the head.

As the check 4 sent out from the check supply unit 6 is conveyed along the conveyance path 5, the front image that is the image of the front surface 4a and the back image that is the image of the back surface 4b are read, and further, the check 4 is transferred to the front surface 4a. The recorded magnetic ink character string 4A is magnetically read. The check 4 for which the magnetic ink character string 4 </ b> A has been successfully read is discharged to the first check discharge unit 7 after a predetermined image related to the endorsement is recorded.
On the other hand, the check 4 that has failed to be read is discharged to the second check discharge unit 8 without recording a predetermined image related to the endorsement. The check 4 discharged to the second check discharge unit 8 is subjected to processing such as investigation of the cause of reading failure and re-reading.

FIG. 2 is an explanatory diagram showing the internal structure of the check reading device 1.
The check supply unit 6 is provided with a check delivery mechanism 10 for sending the check 4 to the transport path 5. The check delivery mechanism 10 includes a delivery roller 11, a delivery roller 12, a retard roller 13 pressed against the delivery roller 12, a delivery motor 14, and a check pressing hopper 15.
When the feeding motor 14 is driven, the check 4 placed in the check supply unit 6 is pressed against the feeding roller 11 by the hopper 15, and in this state, the feeding roller 11 and the feeding roller 12 rotate synchronously.

The check 4 is fed between the feed roller 12 and the retard roller 13 by the feed roller 11. The retard roller 13 is given a predetermined rotational load, and only one check 4 that is in direct contact with the delivery roller 12 is separated from the other checks 4 and delivered to the transport path 5.
The conveyance path 5 has a U shape as described above, and is connected to the linear upstream conveyance path portion 21 connected to the check supply unit 6, the first check discharge unit 7, and the second check discharge unit 8. In addition, a downstream conveyance path portion 23 extending in a slightly bent state and a curved conveyance path portion 22 that connects the downstream conveyance path portion 22 are provided.

  A check conveyance mechanism 30 that conveys the check 4 sent from the check supply unit 6 to the conveyance path 5 along the conveyance path 5 is pressed by the first conveyance roller 31 to the sixth conveyance roller 36 and the rotation. The first pressing roller 41 to the sixth pressing roller 46 and a transport motor 37 for rotationally driving the first transport roller 31 to the sixth transport roller 36 are provided. The 1st conveyance roller 31-the 6th conveyance roller 36 rotate in synchronization. For example, a stepping motor is used as the transport motor 37. For this reason, the conveyance amount of the check 4 can be known from the number of steps for driving the stepping motor.

  The first conveyance roller 31 to the third conveyance roller 33 are respectively disposed at the upstream end of the upstream conveyance path portion 21, the middle position thereof, and the boundary position with the curved conveyance path portion 22. The fourth transport roller 34 is disposed at a downstream position in the curved transport path portion 22. The fifth transport roller 35 and the sixth transport roller 36 are respectively disposed at the middle position and the downstream end in the downstream transport path portion 23.

  Between the first conveyance roller 31 and the second conveyance roller 32 in the upstream conveyance path portion 21, from the upstream side, a magnet 51 for magnetizing magnetic ink characters, a surface side contact image sensor 52 as an optical reading unit, and A contact image sensor 53 on the back side is arranged. The front-side contact image sensor 52 faces the front surface 4a of the check 4 transported on the transport path 5, and reads a front surface image that is an image of the front surface 4a. The back side contact image sensor 53 faces the back side 4b of the check 4 conveyed on the conveyance path 5, and reads a back side image that is an image of the back side 4b.

Further, a magnetic head 54 as a magnetic reading unit for reading magnetic ink characters is disposed between the second transport roller 32 and the third transport roller 33, and the check 4 is pressed against the magnetic head 54. The pressing roller 55 for this is opposed.
Between the fifth transport roller 35 and the sixth transport roller 36 in the downstream transport path portion 23, a recording device 56 for recording a predetermined image related to the endorsement is disposed. The recording device 56 includes a print head, a stamp, and the like that can record a predetermined image in an appropriate direction at an appropriate position on the back surface 4b of the check 4 conveyed on the conveyance path 5.

Various sensors for check conveyance control are arranged in the conveyance path 5. A paper length detector 61 for detecting the length of the check 4 to be sent out is disposed at a position on the front side of the magnet 51. Between the back side contact image sensor 53 and the second transport roller 32, a double feed detector 62 for detecting that the check 4 is transported in an overlapping state is disposed. A jam detector 63 is disposed at a position on the front side of the fourth transport roller 34. When the check 4 is continuously detected by the jam detector 63 over a predetermined time, the transport path 5 is detected. It can be seen that the check 4 is jammed.
A print detector 64 for detecting the presence or absence of the check 4 endorsed by the recording device 56 is disposed at a position on the front side of the fifth transport roller 35. Furthermore, it is discharged into these positions on the downstream side of the sixth transport roller 36, that is, the position of the branch path 9 that branches from the transport path 5 to the first check discharge section 7 and the second check discharge section 8. A discharge detector 65 for detecting the check 4 is arranged.

  A switching plate 66 that is switched by a drive motor 67 (see FIG. 3) is disposed in the branch path 9. The switching plate 66 is a member for selectively switching the downstream end of the transport path 5 with respect to the first check discharge unit 7 and the second check discharge unit 8 and guiding the check 4 to the selected discharge unit.

FIG. 3 is a block diagram showing a functional configuration of the check reading apparatus 1.
The control unit 71 centrally controls each part of the check reading device 1 under the control of the host-side control unit 73 of the host computer 70 described later, and controls the CPU, ROM, RAM, and other peripheral circuits. I have.
The control unit 71 drives the delivery motor 14 and the conveyance motor 37 under the control of the host-side control unit 73 to feed the checks 4 (see FIG. 1) one by one to the conveyance path 5, and the sent checks 4 is transported along the transport path 5. The conveyance control of the check 4 by the control unit 71 is based on detection signals from the paper length detector 61, the double feed detector 62, the jam detector 63, the print detector 64 and the discharge detector 65 arranged in the conveyance path 5. Based on.

As the check 4 is transported, the front-side contact image sensor 52 and the back-side contact image sensor 53 respectively read the image on the front surface and the back surface of the check 4 transported on the transport path 5, and read the read images. The image data shown is output to the control unit 71. The control unit 71 outputs these image data to the host side control unit 73.
Further, the magnetic head 54 detects an electromotive force generated by a change in the magnetic field formed by the passing magnetic ink character string 4A (see FIG. 1) under the control of the control unit 71, and a signal processing circuit 74 as a detection signal. Output to.

The signal processing circuit 74 includes an amplifier, a noise removal filter circuit, an A / D converter, and the like, amplifies the detection signal input from the magnetic head 54, shapes the waveform, and outputs the amplified signal to the control unit 71 as data. To do. The control unit 71 transmits data indicating the detection signal input from the signal processing circuit 74 to the host side control unit 73.
The operation unit 75 includes various switches such as a power switch and operation switches formed on the main body case 2 (see FIG. 1), detects user operations on these switches, and outputs them to the control unit 71.

A host computer 70 is connected to the check reading device 1 via a communication cable 72.
The host computer 70 includes a host-side control unit 73 that includes a CPU, ROM, RAM, and other peripheral circuits. The host-side control unit 73 includes a character recognition unit 80 described later.
The host-side control unit 73 includes a display 76 that can display various types of information, an operation unit 77 to which input devices such as a keyboard and a mouse are connected, and a storage unit that stores various types of rewritable data such as an EEPROM and a hard disk. 78 is connected.
The storage unit 78 stores the front image of the check 4 input from the check reading device 1, data indicating the back image, and data indicating the detection signal.

  In the present embodiment, the control unit 71 of the check reading device 1 controls each part of the check reading device 1 under the control of the host side control unit 73 of the host computer 70. Specifically, the host-side control unit 73 executes a program stored in the ROM by the CPU to generate control data for controlling the control unit 71, and uses the generated control data for the check reading device 1. By outputting to the control unit 71, each part of the check reading device 1 is controlled. That is, in this embodiment, the host computer 70 and the check reading device 1 cooperate to function as a recording medium processing device that executes various processes on the check 4 as a recording medium.

<Character recognition part>
Next, the character recognition unit 80 included in the host-side control unit 73 will be described.
The character recognition unit 80 performs character recognition for each of the magnetic ink characters constituting the magnetic ink character string 4A. The magnetic ink character is a character magnetically printed on the check 4 in accordance with a predetermined font (for example, E-13B font), and one magnetic ink character has a plurality of predetermined character types. Corresponds to any one character type. The recognition of the magnetic ink character is to determine the character type of each magnetic ink character for each of the read magnetic ink characters or to detect that the character type of the magnetic ink character cannot be determined.

In the E-13B font, the shapes of magnetic ink characters are “0” to “9”, transit symbol (TRANSIT SYMBOL), amount symbol (AMOUNT SYMBOL), on-as symbol (ON-US SYMBOL), and Shapes corresponding to 14 character types of dash symbols (DASH SYMBOL) are prepared.
In addition, each magnetic ink character of the magnetic ink character string 4A may be printed on the check 4 by offset printing or may be printed by laser printing. The shape of the magnetic ink character of E-13B font printed by offset printing is different from that of the magnetic ink character of E-13B font printed by laser printing.

In the present embodiment, if the character type of the magnetic ink character can be determined for all the magnetic ink characters included in the magnetic ink character string 4A, it is determined that the magnetic ink character string 4A has been successfully read. If one of the characters cannot determine the character type of the magnetic ink character, it is determined that the reading of the magnetic ink character string 4A has failed.
As described above, under the control of the host-side control unit 73, the control unit 71 conveys the check 4 to the first check discharge unit 7 for the check 4 for which the magnetic ink character string 4A has been successfully read. For the check 4 for which reading of the magnetic ink character string 4A has failed, the check 4 is conveyed to the second check discharge unit 8 after executing a predetermined process.
Hereinafter, the operation of the character recognition unit 80 will be described in detail.

4 to 11 are flowcharts showing the operation of the character recognition unit 80 according to the first embodiment.
The function of the character recognition unit 80 is realized by the cooperation of hardware and software, such as the CPU of the host-side control unit 73 executing a program stored in the ROM.

  The flowcharts of FIGS. 4 to 11 show the operation of the character recognition unit 80 when executing recognition of a magnetic ink character string 4A recorded on a check 4 for a certain check 4. The magnetic ink character string 4A recorded on the check 4 is read by the magnetic head 54, and the signal processing circuit 74 performs various processing such as amplification processing, filtering processing, and waveform shaping processing. It is assumed that data indicating the detection signal (hereinafter referred to as “detection signal waveform data”) is generated and the detection signal waveform data is output to the host-side control unit 73 by the control unit 71. Further, it is assumed that an image of the surface of the check 4 is read by the front-side contact image sensor 52 and output from the control unit 71 to the host-side control unit 73 as image data.

Referring to FIG. 4, the host-side control unit 73 cuts out characters from one of the magnetic ink characters included in the magnetic ink character string 4 </ b> A from the detection signal waveform data input from the control unit 71. Perform (step SA1). Character segmentation refers to generating character waveform data indicating the waveform of a magnetic ink character for one magnetic ink character included in the magnetic ink character string 4A based on the detection signal waveform data. is there.
The process of step SA1 is executed as follows, for example.

FIG. 12 is a diagram schematically illustrating an example of the contents of the detection signal waveform data (part), and FIG. 13 is a diagram schematically illustrating an example of the contents of the character waveform data generated by the processing in step SA1. It is.
As described above, in the present embodiment, the magnetic ink character string 4 </ b> A recorded on the check 4 is read by the magnetic head 54 while the check 4 is being conveyed in the conveyance path 5. The magnetic ink character string 4A of the check 4 being conveyed is read at a predetermined sampling period from the rear end 4d side to the front end 4g side, thereby reading the magnetic ink character string 4A. In the following description, the interval of the minimum unit of the sampling period is referred to as one sampling unit.

  In FIG. 12, the X axis (horizontal axis) defines the passage of time (elapsed sampling period) (indicating that one sampling unit has sequentially passed from the origin to the right of the X axis), and the Y axis ( The vertical axis) defines the relative value of the amount of change in magnetic charge with time. In FIG. 12, in the magnetic ink character string 4A, how the relative value of the change amount of the magnetic charge changes every predetermined sampling period from the rear end 4d side of the check 4 toward the front end 4g side. The value in the Y-axis direction rises and falls according to the amount of change in magnetic charge of each magnetic ink character included in the magnetic ink character string 4A, and the amount of change in magnetic charge is positive or negative. Depending on whether the value is positive or negative in the Y-axis direction.

  Here, in this embodiment, the range occupied by the waveform corresponding to one magnetic ink character in the X-axis direction of the detection signal waveform data is defined as a predetermined number of sampling units (for example, 70 sampling units). In accordance with this rule, various transport controls are executed and the length of one sampling unit is specified.

Based on this, in step SA1, the character recognition unit 80 is based on the waveform indicated by the detection signal waveform data, and each peak position (the X-axis value corresponding to the maximum value and the minimum value) of the waveform. The character start position is detected for each of the magnetic ink characters included in the magnetic ink character string 4A based on the detected peak position. Then, the character recognition unit 80 cuts out the waveform indicated by the detection signal waveform data from the detected character start position in a predetermined number of sampling units corresponding to the width of one magnetic ink character.
At that time, the character recognition unit 80 cuts out the waveform indicated by the detection signal waveform data so that the first peak position in the cut-out waveform is located at a predetermined sampling unit on the X axis. For example, the character recognition unit 80 determines the waveform of the detection signal waveform data so that the first peak position of the waveform is located at the 11th sampling from the character start position for one magnetic ink character included in the magnetic ink character string 4A. Cut out.

Character waveform data having the contents shown in the example of FIG. 13 based on the actual detection signal of the magnetic ink character for one magnetic ink character included in the magnetic ink character string 4A by cutting out the character in step SA1. Is generated.
The character cut-out in step SA1 is performed for each magnetic ink character included in the magnetic ink character string 4A from the magnetic ink character arranged on the rear end 4d side of the check 4 to the magnetic ink character arranged on the front end 4g side. Steps SA <b> 2 and subsequent processing relating to character recognition are executed for each magnetic ink character cut out in order.
Hereinafter, one magnetic ink character cut out in step SA1 is referred to as a “processing target character”.

  After the character segmentation is performed, the character recognition unit 80 determines that the amplitude level in the Y-axis direction of the waveform indicated by each data of the generated character waveform data of the processing target character is the Y-axis of the reference waveform data for pattern matching described later. Data normalization is performed so as to be equivalent to the amplitude level in the direction (step SA2).

  Referring to FIG. 4, step SA2 includes two processing phases: a first recognition phase (step SA3) and a second recognition phase (step SA7). These two processing phases are processes related to so-called magnetic ink character recognition, which are sequentially executed on the magnetic ink characters (processing target characters) cut out in step SA1, and are thereby included in the magnetic ink character string 4A. Run once for each magnetic ink character. These two processing phases are phases in which magnetic ink character recognition of the character to be processed is performed by different methods, and the character type of the character to be processed is determined or it is detected that it cannot be determined.

FIG. 5 is a flowchart showing the operation of the character recognition unit 80 in the first recognition phase of step SA3.
In the first recognition phase, first, the character recognizing unit 80 executes a first recognition pass for executing a process of determining that the character type of the processing target character is not determined or cannot be determined (step SB1).

  Next, the character recognition unit 80 determines whether or not the character type of the processing target character has been determined in the first recognition pass (step SB2), and when the character type of the processing target character has been determined (step SB2: YES). The first recognition phase is finished. On the other hand, when the character type of the character to be processed cannot be determined (step SB2: NO), the character recognition unit 80 executes the second recognition pass (step SB3).

Next, the character recognition unit 80 determines whether or not the character type of the processing target character has been determined in the second recognition pass (step SB4), and when the character type of the processing target character has been determined (step SB4: YES). The first recognition phase is finished. On the other hand, when the character type of the processing target character cannot be determined (step SB4: NO), the character recognition unit 80 determines that the character type of the processing target character cannot be determined in the first recognition phase (step SB5).
Hereinafter, the processing of the first recognition pass and the second recognition pass will be described in detail.

FIG. 6 is a flowchart showing the operation of the character recognition unit 80 in the first recognition pass of step SB1 of FIG.
In the first recognition pass, the character recognition unit 80 specifies a character type to be compared in the process of step SC2 from the 14 character types defined in the E-13B font (step SC1). Hereinafter, the character type specified in step SC1 is referred to as a “comparison character type”. The process of step SC2 is performed once in order, with each of the 14 character types being compared.

  Next, the character recognition unit 80 calculates a difference amount between the waveform of the character waveform data of the processing target character cut out in step SA1 and the waveform indicated by each of the reference waveform data corresponding to the comparison target character type (step). SC2).

  The processing in step SC2 will be described in detail. The reference waveform data corresponding to one character type is an ideal waveform of a detection signal when a magnetic ink character corresponding to the one character type is read by the magnetic head 54. This is data corresponding to template data provided in pattern matching processing in so-called magnetic ink character recognition.

Here, as described above, the magnetic ink characters printed by offset printing and the magnetic ink characters printed by laser printing have different shapes. Therefore, the reference waveform data includes two types of data: reference waveform data for offset printing and reference waveform data for laser printing. In the first recognition phase, various types of processing are executed using any one of these two types of reference waveform data.
For example, in the first recognition phase, various processes are executed using the reference waveform data for offset printing, and thereafter, various processes similar to those in the first recognition phase are performed using the reference waveform data for laser printing. It is good also as a structure which performs magnetic ink character recognition using both of two types of data by performing separately.

FIG. 14 is a diagram for explaining the difference amount detected in step SC2. In FIG. 14, the waveform indicated by the character waveform data is indicated by a thin line, and the waveform indicated by the reference waveform data is indicated by a thick line.
The difference amount is the size of the area shown by hatching in FIG. 14, and more specifically, in the Y-axis direction of the waveform indicated by the character waveform data in each sampling unit defined on the X-axis. This is the sum of absolute values of the difference between the value and the value in the Y-axis direction of the waveform indicated by the reference waveform data. The smaller the difference between the waveform indicated by one character waveform data and the waveform indicated by one reference waveform data, the smaller the waveform indicated by the one character waveform data and the waveform indicated by the one reference waveform data. And the character type of the magnetic ink character corresponding to the one character waveform data is highly likely to be the character type corresponding to the one reference waveform data.

After calculating the amount of difference between the waveform indicated by the character waveform data of the character to be processed and the waveform of the reference waveform data corresponding to the character type to be compared, the character recognition unit 80 performs the following operation for all 14 types of characters. It is determined whether or not the calculation of the difference amount has been completed (step SC3).
If the calculation of the difference amount has not been completed for all 14 character types (step SC3: NO), the character recognition unit 80 returns the processing procedure to step SC1, and the calculation of the difference amount for all character types is performed. If completed (step SC3: YES), the process proceeds to step SC4.

In step SC4, the character recognizing unit 80 sets the character type corresponding to the reference waveform data having the smallest difference amount as the first candidate C11 for pass 1, and then selects the character type corresponding to the reference waveform data having the smallest difference amount. The second candidate for path 1 is C12.
As described above, the character type set as the first candidate C11 for pass 1 has the highest probability of being the character type of the processing target character, and the character type set as the second candidate C12 for pass 1 is the character type of the processing target character. The probability of being a character type is the second highest.

Next, the character recognition unit 80 compares the difference amount between the reference waveform data corresponding to the character type determined as the first candidate C11 for pass 1 and the character waveform data of the character to be processed, and the threshold value T1 for pass 1. Then, it is determined whether or not the difference amount exceeds the pass 1 threshold value T1 (step SC5).
The pass 1 threshold value T1 is a value serving as a reference when determining whether or not to determine the character type set as the first candidate C11 for pass 1 as the character type of the character to be processed. In the case of the threshold value T1 or less, if other conditions described later are satisfied, the character recognition unit 80 determines the character type set as the first candidate for pass 1 C11 as the character type of the processing target character.

The threshold value T1 for pass 1 is such that only the amount of difference between the character waveform data of the character to be processed and the reference waveform data corresponding to the correct character type of the character to be processed is equal to or less than the threshold value for pass 1 T1, and The amount of difference between the character waveform data and the reference waveform data corresponding to another character type is set appropriately from the viewpoint of exceeding the pass 1 threshold value T1. Specifically, in the present embodiment, the pass 1 threshold T1 is set for each character type as follows.
That is, in a preliminary test, for each of the 14 character types of magnetic ink characters, it corresponds to the character waveform data when one magnetic ink character is actually read and the correct character type of the one magnetic ink character. The amount of difference from the reference waveform data is detected a plurality of times. Then, the pass 1 threshold T1 for one magnetic ink character is expressed as “pass 1 threshold T1 = (average value of a plurality of detected difference amounts) + (standard deviation of a plurality of detected difference amounts) × (predetermined coefficient) ) ”. The (predetermined coefficient) is a value set for each character type based on the results of previous experiments and simulations for the purpose of setting the pass 1 threshold value T1 according to the above-described viewpoint. In this way, by setting the pass 1 threshold value T1, it is possible to set the pass 1 threshold value T1 to an appropriate value in consideration of variations in the actually calculated difference amount.

If the difference amount exceeds the pass 1 threshold T1 (step SC5: YES), the character recognition unit 80 ends the first recognition pass without determining the character type of the processing target character.
When the difference amount is equal to or smaller than the pass 1 threshold value T1 (step SC5: NO), the character recognition unit 80 further includes the reference waveform data corresponding to the character type determined as the pass 1 second candidate C12 and the character to be processed. The difference between the character waveform data and the pass 1 threshold T1 is compared, and it is determined whether or not the difference exceeds the pass 1 threshold T1 (step SC6).

When the difference amount exceeds the pass 1 threshold value T1 (step SC6: YES), the character recognition unit 80 determines the character type set as the first pass 1 candidate C11 as the character type of the processing target character (step step). SC7), the first recognition pass is terminated.
On the other hand, when the difference amount is equal to or less than the pass 1 threshold T1 (step SC6: NO), the character recognition unit 80 ends the first recognition pass without determining the character type of the processing target character.

Thus, in the present embodiment, when the difference amount between the reference waveform data corresponding to the character type set as the first candidate C11 for pass 1 and the character waveform data of the character to be processed is equal to or less than the threshold value T1 for pass 1. Even if the difference amount between the reference waveform data corresponding to the character type determined as the second candidate C1 for pass 1 and the character waveform data of the character to be processed is equal to or smaller than the threshold value T1 for pass 1, the processing is performed. Does not determine the character type of the target character. This is due to the following reason.
That is, the threshold value T1 for pass 1 is such that only the amount of difference between the character waveform data of the character to be processed and the reference waveform data corresponding to the correct character type of the character to be processed is equal to or less than the threshold value T1 for pass 1. The difference amount between the character waveform data and the reference waveform data corresponding to another character type is a value set so as to exceed the pass 1 threshold value T1. Therefore, when the check 4 is normally conveyed and the magnetic ink character string 4A is normally read by the magnetic head 54, the character type related to the difference amount below the pass 1 threshold T1 is limited to one. For this reason, when there are a plurality of character types related to the difference amount below the pass 1 threshold value T1, the situation may be caused by a reading error by the magnetic head 54, a conveyance error of the check 4, or any other cause. This is because there is a possibility of appearing, and if the character type of the character to be processed is determined in such a situation, there is a risk of erroneous recognition.

FIG. 7 is a flowchart showing the operation of the character recognition unit 80 in the second recognition pass of step SB3 of FIG.
In the second recognition pass, the character recognizing unit 80 specifies the character type to be compared in the process of step SD2 and step SD3 from the 14 character types (step SD1). Hereinafter, the character type specified in step SD1 is referred to as “comparison character type”. The processing in step SD2 and step SD3 is performed once in order, with each of the 14 character types being compared.

  Next, the character recognizing unit 80 between the waveform of the character waveform data of the character to be processed and each of the waveforms of the reference waveform data corresponding to the character type to be compared slid by a predetermined distance within a predetermined range. A difference amount is calculated (step SD2).

  The processing of step SD2 will be described with an example. For example, in each character waveform data, the X-axis range occupied by the waveform corresponding to one magnetic ink character is defined as 70 sampling units. In this case, in step SD2, for example, the character recognizing unit 80 is configured to slide the entire waveform of the reference waveform data by 7 sampling units leftward in the X-axis direction, and the waveform of the character waveform data of the character to be processed. The difference amount between is calculated. Similarly, the character recognizing unit 80 slides the entire waveform of the reference waveform data to the left in the X-axis direction by 6, 5, 4, 3, 2, 1, 0 sampling units, A difference amount between the waveform of the character waveform data of the character to be processed is calculated. At the same time, the entire waveform of the reference waveform data is slid to the right in the X-axis direction by 7, 6, 5, 4, 3, 2, 1 sampling units, and the character waveform data of the character to be processed. The amount of difference between the waveforms is calculated. In this case, the difference amount is calculated 15 times, and the difference amount for 15 times is calculated.

  Here, depending on the state of conveyance of the check 4, the state of reading of the magnetic ink character string 4 </ b> A by the magnetic head 54, and the state of the character cutting process in step SA <b> 1 in FIG. The position in the X-axis direction may deviate by several sampling units from the ideal position of the waveform. Based on this, by sliding the waveform of the reference waveform data within a predetermined range as described above, and detecting the difference amount from the waveform of the character waveform data of the character to be processed, the above-described “deviation” Thus, the difference amount can be appropriately calculated.

  Next, the character recognition unit 80 sets the difference amount having the smallest value among the difference amounts calculated in step SD2 as the difference amount of the comparison target character type (step SD3). In the above example, the character recognizing unit 80 specifies the smallest difference value of the calculated 15 differences as the difference amount of the comparison target character type.

Next, the character recognition unit 80 determines whether or not the calculation of the difference amount (step SD2) and the specification (step SD3) have been completed for all 14 character types (step SD4).
When the calculation and identification of the difference amount are not completed for all character types (step SD4: NO), the character recognition unit 80 returns the processing procedure to step SD1.
On the other hand, when the calculation and identification of the difference amount have been completed for all character types (step SD4: YES), the character recognition unit 80 determines that the difference amount identified in step SD3 is the 14 types of character types. The smallest character type is set as the first candidate C2 for pass 2, and the character type with the second smallest specified difference amount is set as the second candidate C22 for pass 2 (step SD5).

Next, the character recognizing unit 80 compares the difference amount specified in step SD3 with the pass 2 threshold value T2 for the character type determined as the pass 2 first candidate C21, and the difference amount is the pass 2 threshold value. It is determined whether or not T2 is exceeded (step SD6).
The threshold value T2 for pass 2 is a value serving as a reference when determining whether or not the character type set as the first candidate C21 for pass 2 is determined as the character type of the character to be processed. In the case of the threshold value T2 or less, if other conditions described later are satisfied, the character recognition unit 80 determines the character type set as the first candidate C21 for pass 2 as the character type of the processing target character.

The pass 2 threshold value T2 is calculated and set in the same manner as the pass 1 threshold value T1 described above. That is, in the preliminary test, for each of the 14 character types of magnetic ink characters, the processing of step SD2 and step SD3 is executed, and the character waveform data when one magnetic ink character is actually read; A difference amount between the reference waveform data corresponding to the correct character type of the one magnetic ink character is specified a plurality of times.
Then, the pass 2 threshold T2 for one magnetic ink character is expressed as “pass 2 threshold T2 = (average value of a plurality of specified difference amounts) + (standard deviation of a plurality of specified difference amounts) × (predetermined coefficient) ) ”. The (predetermined coefficient) is a value set for each character type based on the results of previous experiments and simulations for the purpose of setting the threshold value T2 for pass 2 in accordance with the above-described viewpoint. In this way, by setting the pass 2 threshold value T2, it is possible to set the pass 2 threshold value T2 to an appropriate value in consideration of variation in the actually calculated difference amount.

When the difference amount exceeds the threshold value T2 for pass 2 (step SD6: YES), the character recognition unit 80 ends the second recognition pass without determining the character type of the processing target character.
When the difference amount is equal to or less than the threshold value T2 for pass 2 (step SD6: NO), the character recognition unit 80 further determines the difference amount specified in step SD3 for the character type determined as the second candidate for pass 2 C22. The pass 2 threshold value T2 is compared, and it is determined whether or not the difference amount exceeds the pass 2 threshold value T2 (step SD7).

When the difference amount exceeds the pass 2 threshold value T2 (step SD7: YES), the character recognition unit 80 determines the character type set as the pass 2 first candidate C21 as the character type of the processing target character (step step). SD8), the second recognition pass is terminated.
On the other hand, when the difference amount is equal to or smaller than the pass 2 threshold T2 (step SD7: NO), the character recognition unit 80 ends the second recognition pass without determining the character type of the processing target character.

As described above, in this embodiment, even if the difference amount specified in step SD3 is equal to or smaller than the pass 2 threshold T2 for the character type that is the first pass C2 candidate C21, For the character type determined as the second candidate C22, when the difference amount specified in step SD3 is equal to or less than the threshold value T2 for pass 2, the character type of the character to be processed is not determined. This is due to the same reason as described for the above-described pass 1 threshold T1.
Note that the difference amount specified for each of the character types that are the first candidate C21 for pass 2 and the second candidate C22 for pass 2 in step SD3 is stored until the subsequent steps.

  Now, referring to FIG. 4 described above, after the execution of the first recognition phase in step SA3, the character recognition unit 80 determines whether or not the character type of the character to be processed has been confirmed in the first recognition phase (step SA4). ).

If the character type of the character to be processed can be confirmed (step SA4: YES), the character recognition unit 80 proceeds to step SA5 without performing the second recognition phase. In step SA5, it is determined whether or not all magnetic ink characters included in the magnetic ink character string 4A are characters to be processed. In other words, for all the magnetic ink characters included in the magnetic ink character string 4A, the characters for step SA1. Is cut out, and it is determined whether or not processing relating to character recognition has been executed.
If not all magnetic ink characters included in the magnetic ink character string 4A are characters to be processed (step SA5: NO), the character recognition unit 80 returns the processing procedure to step SA1, and the next magnetic ink On the other hand, when the character is cut out and all the magnetic ink characters included in the magnetic ink character string 4A have been processed (step SA5: YES), the processing procedure proceeds to step SA6.

  On the other hand, if it is determined in step SA4 that the character type of the processing target character cannot be determined in the first recognition phase (step SA4: NO), the character recognition unit 80 executes the second recognition phase (step SA7).

FIG. 8 is a flowchart showing the operation of the character recognition unit 80 during the execution of the second recognition phase in step SA7.
In the second recognition phase, first, the character recognition unit 80 calculates the character spacing of the processing target character based on the detection signal waveform data (step SE1).

The character spacing of the character to be processed is the character start position of the character to be processed and the magnetic ink character next to the character to be processed (magnetic ink arranged next to the front end 4g side of the character to be processed in the magnetic ink character string 4A). Is the distance from the character start position. For example, the character recognition unit 80 detects each peak position (X-axis value corresponding to the maximum value and the minimum value) based on the waveform indicated by the detection signal waveform data, and based on the detected peak position. The character start position of the character to be processed and the character start position of the next magnetic ink character after the character to be processed are detected, and the separation amount of these character start positions is detected based on the detected character start position of each character. The detected separation amount is defined as a character interval.
As described above, in this embodiment, the conveyance control of the check 4 is executed so that the character spacing of each magnetic ink character included in the magnetic ink character string 4A is constant (for example, 67 sampling units), and sampling is performed. The unit length is specified. Hereinafter, the character spacing defined to be constant is referred to as “reference character spacing”.

Next, the character recognition unit 80 compares the character spacing of the processing target character detected at step SE1 with the reference character spacing (step SE2).
In step SE2, when the character interval of the character to be processed is equal to the reference character interval (step SE2: “character interval = reference character interval”), the character recognition unit 80 executes the third recognition pass (step SE3). When the character type of the processing target character can be determined in the third recognition pass, the second recognition phase is terminated, and when the character type cannot be determined, it is determined that the character type of the processing target character cannot be determined. Then, the second recognition phase ends.

  In step SE2, when the character interval of the character to be processed is equal to the value obtained by subtracting “one sampling unit” from the reference character interval (step SE2: “character interval = reference character interval−1”), the character recognition unit 80 The recognition pass is executed step by step in the order of the fourth recognition pass (step SE4), the fifth recognition pass (step SE5), and the third recognition pass (step SE6). At this time, if the character type of the character to be processed can be determined in any recognition pass, the second recognition phase is terminated without executing the next recognition pass. If the character type of the processing target character cannot be determined in any recognition pass, it is determined that the character type of the processing target character cannot be determined, and the second recognition phase ends.

  In step SE2, when the character interval of the character to be processed is less than the value obtained by subtracting “one sampling unit” from the reference character interval (step SE2: “character interval <reference character interval−1”), the character recognition unit 80 The recognition pass is executed step by step in the order of 5 recognition pass (step SE7) and third recognition pass (step SE8). At this time, if the character type of the character to be processed can be determined in any recognition pass, the second recognition phase is terminated without executing the next recognition pass. If the character type of the processing target character cannot be determined in any recognition pass, it is determined that the character type of the processing target character cannot be determined, and the second recognition phase ends.

  In step SE2, when the character interval of the character to be processed is equal to the value obtained by adding “one sampling unit” to the reference character interval (step SE2: “character interval = reference character interval + 1”), the character recognition unit 80 The recognition pass is executed step by step in the order of the fourth recognition pass (step SE9), the fifth recognition pass (step SE10), and the third recognition pass (step SE11). At this time, if the character type of the character to be processed can be determined in any recognition pass, the second recognition phase is terminated without executing the next recognition pass. If the character type of the processing target character cannot be determined in any recognition pass, it is determined that the character type of the processing target character cannot be determined, and the second recognition phase ends.

  In step SE2, when the character interval of the character to be processed exceeds the value obtained by adding “one sampling unit” to the reference character interval (step SE2: “character interval> reference character interval + 1”), the character recognition unit 80 The recognition pass is executed step by step in the order of the recognition pass (step SE12) and the third recognition pass (step SE13). At this time, if the character type of the character to be processed can be determined in any recognition pass, the second recognition phase is terminated without executing the next recognition pass. If the character type of the processing target character cannot be determined in any recognition pass, it is determined that the character type of the processing target character cannot be determined, and the second recognition phase ends.

FIG. 9 is a flowchart showing the operation of the character recognition unit 80 during execution of the third recognition pass.
In the third recognition pass, both-side expansion / contraction comparison, which will be described later, is performed, and based on the result of this both-side expansion / contraction comparison, the character type of the processing target character is determined or the character type of the processing target character cannot be determined. Is called.

Referring to FIG. 9, the character recognition unit 80 specifies a character type to be compared in the process after step SF2 from the 14 character types (step SF1). Hereinafter, the character type specified in step SF1 is referred to as “comparison character type”. The processing after step SF2 is performed 14 times in order, with each of the 14 character types being compared.
Next, the character recognition unit 80 refers to the both-side expansion / contraction table 90 (step SF2).

FIG. 15 is a diagram schematically showing the data structure of the double-side expansion / contraction table 90.
As shown in FIG. 15, in the double-side expansion / contraction table 90, a record is generated for each of 14 character types, and each record has an expansion / contraction point field 91 in which information indicating the expansion / contraction point of the corresponding character type is stored. There are three fields: a maximum contraction amount field 92 in which data indicating the maximum contraction amount is stored, and a maximum expansion amount field 93 in which data indicating the maximum expansion amount is stored. The expansion / contraction point, the maximum contraction amount, and the maximum extension amount will be described later.

  Next, the character recognition unit 80 acquires a record corresponding to the character type to be compared in the referred double-side expansion / contraction table 90, and stores the expansion / contraction point field 91, maximum contraction amount field 92, and maximum expansion amount field 93 of the record. With reference to each field, the respective values of the expansion / contraction point, the maximum contraction amount, and the maximum expansion amount related to the character type to be compared are acquired (step SF3).

Next, the character recognition unit 80 corrects the waveform of the reference waveform data corresponding to the character type to be compared based on the expansion / contraction point, the maximum contraction amount, and the maximum expansion amount acquired in step SF3, and each of the corrected waveforms is corrected. And the amount of difference between the waveform of the character waveform data corresponding to the processing target character (step SF4).
Hereinafter, the process of step SF4 will be described in detail.

FIG. 16 is a diagram used for explaining the processing of step SF4, and more specifically, is a diagram schematically showing the content of the reference waveform data corresponding to the character type “0” in the E-13B font. In FIG. 16, the waveform of the reference waveform data corresponding to the character type “0” is indicated by a solid line, and the width of the waveform in the X-axis direction is 70 sampling units.
In the following description, it is assumed that the character type to be compared is the character type “0”. In step SF4, the corrected waveform of the reference waveform data corresponding to the character type “0” and the character waveform corresponding to the character to be processed are displayed. The process of step SF4 will be described by taking as an example the case of calculating the amount of difference from the data waveform.

As shown in FIG. 15, in this embodiment, the expansion / contraction point of the character type “0” is “the 42nd sampling unit”, the maximum contraction amount is “3 sampling units”, and the maximum expansion amount is “6 sampling units”. It is.
In step SF4, first, the character recognizing unit 80 uses the maximum contraction amount for the waveform on the right side in the X-axis direction from the expansion / contraction point (42th sampling unit) among the waveforms of the reference waveform data corresponding to the character type “0”. Slide to the left in the X-axis direction by a certain “3 sampling units”. Thereby, referring to FIG. 16, the waveform of the reference waveform data (the waveform indicated by the solid line) is corrected to become the waveform indicated by the broken line.

Next, the character recognition unit 80 calculates a difference amount between the corrected waveform of the reference waveform data and the waveform of the character waveform data corresponding to the character to be processed. When the difference amount is calculated in this way, the waveform to the right of the 42th sampling unit among the waveforms of the reference waveform data is slid to the left in the X-axis direction by 3 sampling units. Therefore, for example, the difference between the value of the 42nd sampling of the waveform of the character waveform data corresponding to the character to be processed and the value of the 45th sampling of the waveform of the reference waveform data for correction “before” is taken. The 43rd sampling value of the waveform of the character waveform data corresponding to the character is taken as a difference from the 46th sampling value of the waveform of the reference waveform data before correction.
That is, when calculating the difference amount, regarding the waveform on the left in the X-axis direction from the n (where n ≧ 42) sampling unit, the value of the n sampling unit of the reference waveform data of the character to be processed and the reference before correction By calculating the difference from the value of the “n + 3” sampling unit of the waveform of the waveform data, the difference amount is calculated. When n exceeds 70, the value of the 70th sampling unit is used.

  After calculating the difference amount related to the maximum contraction amount in this way, the character recognition unit 80 in the X-axis direction from the expansion / contraction point (the 42nd sampling unit) among the waveforms of the reference waveform data corresponding to the character type “0”. The right waveform is slid to the left in the X-axis direction by “1 sampling unit” subtracted from “3 sampling units” which is the maximum contraction amount, that is, “2 sampling units”. By this slide, the waveform of the reference waveform data is corrected, and a difference amount between the corrected waveform of the reference waveform data and the waveform of the character waveform data of the character to be processed is detected.

  In this manner, the character recognition unit 80 corrects the waveform of the reference waveform data based on the maximum contraction amount after subtraction while subtracting the maximum contraction amount by “one sampling unit”, and the corrected reference waveform data A difference amount between the waveform and the waveform of the character waveform data of the character to be processed is calculated. This difference amount is detected until the value of the maximum contraction amount after subtraction becomes “one sampling unit”. In other words, the character recognizing unit 80 divides between the waveform of the reference waveform data corrected by sliding to the left of the X axis by 3, 2, 1 sampling unit and the waveform of the character waveform data corresponding to the character to be processed. Each difference amount is calculated.

  Further, the character recognizing unit 80 sets “6 sampling” which is the maximum expansion amount for the waveform on the right side in the X-axis direction from the expansion / contraction point (42th sampling unit) among the waveforms of the reference waveform data corresponding to the character type “0”. Slide to the right in the X-axis direction by “unit”. Thereby, referring to FIG. 16, the waveform of the reference waveform data (the waveform indicated by the solid line) is corrected to become the waveform indicated by the two-dot chain line.

Next, the character recognition unit 80 calculates a difference amount between the corrected waveform of the reference waveform data and the waveform of the character waveform data corresponding to the character to be processed. When the difference amount is calculated in this way, the waveform to the right of the 42th sampling unit among the waveforms of the reference waveform data is slid to the right in the X-axis direction by 6 sampling units. The difference between the value at the 60th sampling of the waveform of the character waveform data corresponding to the character and the value at the 54th sampling of the waveform of the reference waveform data before correction is taken. Also, the 61st sampling value of the waveform of the character waveform data corresponding to the character to be processed is the difference from the 55th sampling value of the waveform of the reference waveform data before correction.
That is, when calculating the difference amount, regarding the waveform on the left in the X-axis direction from the n (where n ≧ 42) sampling unit, the value of the n sampling unit of the reference waveform data of the character to be processed and the reference before correction The difference is calculated by taking the difference from the value of the “n-6” sampling unit of the waveform of the waveform data. In the case of n = 43 to 48, the value of the 42th sampling unit is used.

  After calculating the difference amount related to the maximum elongation in this way, the character recognition unit 80 in the X-axis direction from the expansion / contraction point (42th sampling unit) in the waveform of the reference waveform data corresponding to the character type “0”. The right waveform is slid rightward in the X-axis direction by “1 sampling unit” subtracted from “6 sampling units” which is the maximum extension amount, that is, “5 sampling units”. As a result, the waveform of the reference waveform data is corrected, and a difference amount between the corrected waveform of the reference waveform data and the waveform of the character waveform data of the character to be processed is detected.

  In this manner, the character recognition unit 80 corrects the waveform of the reference waveform data based on the maximum extension amount after subtraction while subtracting the maximum extension amount by “one sampling unit”, and the corrected reference waveform data A difference amount between the waveform and the waveform of the character waveform data of the character to be processed is calculated. This difference amount is detected until the maximum elongation value after subtraction reaches “one sampling unit”. That is, the character recognizing unit 80 converts each of the waveforms of the reference waveform data corrected by sliding to the right of the X axis by 6, 5, 4, 3, 2, 1 sampling units and the character waveform corresponding to the processing target character. Each difference between the waveform of the data is calculated.

  Further, the character recognition unit 80 does not correct the waveform of the reference waveform data, and the amount of difference between the waveform of the character waveform data corresponding to the character to be processed and the waveform of the reference waveform data corresponding to the character type “0”. Is calculated.

  As in the above example, in step SF4, the waveform of the reference waveform data corresponding to the comparison target character type is corrected, and the waveform between each of the corrected waveforms and the waveform of the character waveform data corresponding to the processing target character is corrected. A difference amount is calculated. In the above example, in step SF4, the waveform to the right of the expansion / contraction point is slid to the left of the X axis by 3, 2, 1 sampling unit, 6, 5, 4, 3, 2, 1 sampling. A total of ten different amounts of difference are calculated for each of the unit slid to the right in the X-axis direction and the non-slidable unit.

  Here, when the magnetic ink character string 4A is recorded on the check 4, the magnetic ink character is printed in the horizontal direction (direction corresponding to the X axis) due to the magnetic ink character being printed out of the center value of the printing standard. May be printed in a contracted state or in an extended state. In this case, the waveform of the character waveform data generated based on the detection signal waveform data may extend or contract in the X-axis direction. In addition, depending on the state of conveyance of the check 4, the waveform of the character waveform data may extend or contract in the X-axis direction. Based on this, as described above, the waveform of the reference waveform data is slid within a predetermined range from a predetermined position, and then the difference amount is detected from the waveform of the character waveform data of the character to be processed. The amount of difference can be calculated appropriately reflecting the stretch and shrinkage.

The stretch points, maximum shrinkage, and maximum stretch for each character type are based on the results of prior tests and simulations for the purpose of appropriately calculating the difference, reflecting the stretch and shrinkage of the magnetic ink characters. Based on the following points.
That is, the expansion / contraction point is set to a point corresponding to the division position when the waveform corresponding to the reference waveform data of the magnetic ink character is divided according to the density of the peak position, for example. Further, as the expansion / contraction point, a point with a small change amount per unit is set for the waveform corresponding to the reference waveform data of the magnetic ink character.
In addition, the maximum shrinkage amount and the maximum stretch amount are the same even when the waveform of the reference waveform data corresponding to one character type is corrected based on the maximum shrinkage amount or the maximum stretch amount. These values are set large if they are not approximated to the waveform of the reference waveform data of a character type other than the character type, whereas these values are set small if they are approximated.

  In step SF4, after correcting the waveform of the reference waveform data corresponding to the type of character to be compared and calculating the difference between each of the corrected waveforms and the waveform of the character waveform data corresponding to the character to be processed, the character The recognizing unit 80 identifies the calculated difference amount having the smallest value, and identifies the identified difference amount as the difference amount of the comparison target character type (step SF5).

Next, the character recognition unit 80 determines whether or not the calculation of the difference amount (step SF4) and the specification (step SF5) have been completed for all 14 character types (step SF6).
When the calculation and identification of the difference amount are not completed for all character types (step SF6: NO), the character recognition unit 80 returns the processing procedure to step SF1.
On the other hand, when the calculation and identification of the difference amount have been completed for all character types (step SF6: YES), the character recognition unit 80 has the smallest identified difference amount among the 14 character types. The character type is set as the first candidate C3 for pass 3, and the character type having the second smallest difference is specified as the second candidate C32 for pass 3 (step SF7).

Next, the character recognizing unit 80 compares the difference amount specified in step SF5 with the pass 3 threshold value T3 for the character type determined as the first pass 3 candidate C31, and the difference amount is determined as the pass 3 threshold value. It is determined whether or not T3 is exceeded (step SF8).
The threshold value T3 for pass 3 is a value serving as a reference when determining whether or not the character type set as the first candidate C31 for pass 3 is determined as the character type of the character to be processed. In the case of the threshold value T3 or less, if other conditions to be described later are satisfied, the character recognition unit 80 determines the character type set as the first candidate C31 for pass 3 as the character type of the processing target character.

  The pass 3 threshold value T3 is calculated and set by the same method as the pass 1 threshold value T1 described above. That is, in the preliminary test, the processing of step SF2 to step SF5 is executed for each of the 14 types of magnetic ink characters, and the character waveform data when one magnetic ink character is actually read, and the 1 A difference amount between the reference waveform data corresponding to the correct character type of one magnetic ink character is specified a plurality of times. Then, the pass 3 threshold value T3 relating to one magnetic ink character is expressed as “pass 3 threshold value T3 = (average value of a plurality of specified difference amounts) + (standard deviation of a plurality of specified difference amounts) × (predetermined coefficient) ) ”. The (predetermined coefficient) is a value set for each character type based on the results of previous experiments and simulations for the purpose of setting the pass 3 threshold T3 in accordance with the above-described viewpoint. By setting the pass 3 threshold value T3 in this way, it is possible to set the pass 3 threshold value T3 to an appropriate value in consideration of the variation in the actually calculated difference amount.

If the difference amount exceeds the threshold 3 for pass 3 (step SF8: YES), the character recognition unit 80 ends the third recognition pass without determining the character type of the processing target character.
When the difference amount is equal to or less than the pass 3 threshold value T3 (step SF8: NO), the character recognition unit 80 further determines the difference amount specified in step SF5 for the character type that is the second candidate for pass 3 C32. The pass 3 threshold value T3 is compared to determine whether or not the difference amount exceeds the pass 3 threshold value T3 (step SF9).

When the difference amount exceeds the pass 3 threshold value T3 (step SF9: YES), the character recognition unit 80 determines the character type set as the pass 3 first candidate C31 as the character type of the processing target character (step SF10), the third recognition pass is terminated.
On the other hand, when the difference amount is equal to or less than the pass 3 threshold T3 (step SF9: NO), the character recognition unit 80 determines that the difference amount related to the character type determined as the pass 3 second candidate C32 is the pass 3 pass number. It is determined whether or not there is a size of 5/4 or more of the difference amount related to the character type set as one candidate C31 (step SF11).

If the size is 5/4 or more (step SF11: YES), the character recognition unit 80 determines the character type set as the first candidate C31 for pass 3 as the character type of the processing target character (step SF10). ), The third recognition pass is terminated.
On the other hand, when the size is not 5/4 or more (step SF11: NO), the character recognition unit 80 ends the third recognition pass without determining the character type of the processing target character.

As described above, in this embodiment, when both the difference amount related to the first candidate C31 for path 3 and the difference amount related to the second candidate C32 for path 3 are less than the threshold value T3 for path 3, It is determined whether or not the difference amount related to the second candidate C32 has a size that is 5/4 or more of the difference amount related to the first candidate C31 for pass 3, and if there is a size that is 5/4 or more times The character type determined as the first candidate C31 for pass 3 is determined as the character type of the processing target character. This is due to the following reason.
That is, when the difference amount related to the second candidate C32 for path 3 is greater than or equal to 5/4 times the difference amount related to the first candidate C31 for path 3, the difference amount related to the first candidate C31 for path 3 Can be said to be a sufficiently small value compared to the difference amount relating to the second candidate C32 for pass 3, and therefore, the character type set as the first candidate C31 for pass 3 is another character type. This is because it can be said that the probability of the character type of the processed character is sufficiently high.

Next, the fourth recognition path will be described with reference to FIG.
The operation of the fourth recognition pass is the same as the processing of step SF11 in FIG. This is different from the operation of the third recognition pass in that there is no process for determining whether or not there is a problem. Thereby, when both the difference amount related to the first candidate C31 for pass 3 and the difference amount related to the second candidate C32 for pass 3 are less than the threshold value T3 for pass 3, the character type of the character to be processed If it is determined that there is a risk of erroneous recognition, the character type of the character to be processed is not determined.
In this way, the condition for determining the character type of the character to be processed is stricter in the fourth recognition pass than in the third recognition pass, and as shown in FIG. Executed.

FIG. 10 is a flowchart showing the operation of the character recognition unit 80 during execution of the fifth recognition pass.
Referring to FIG. 10, in the fifth recognition pass, first, the character recognition unit 80 determines whether “shrinkage” may occur in the waveform of the character waveform data of the character to be processed, or “elongation” occurs. It is determined whether or not there is a possibility (step SG1).

The character recognition unit 80 determines whether or not “shrinkage” may occur in the waveform of the character waveform data of the processing target character, and whether or not “elongation” may occur. The determination is made as follows.
That is, when the character interval of the processing target character is smaller than the reference character interval, the width of the waveform of the processing target character on the X axis is reduced compared to a normal state. In this case, The character recognition unit 80 determines that “shrinkage” may occur in the waveform of the character waveform data of the character to be processed. On the other hand, when the character interval of the processing target character is larger than the reference character interval, the width of the waveform of the processing target character on the X axis is expanded as compared with the normal state. In this case, The character recognition unit 80 determines that “elongation” may occur in the waveform of the character waveform data of the character to be processed.

If it is determined in step SG1 that there is a possibility that “shrinkage” has occurred (step SG1: “shrinkage”), the character recognition unit 80 performs processing in step SG3 and subsequent steps from among 14 character types. A character type to be compared is specified (step SG2). Hereinafter, the character type specified in step SG2 is referred to as “comparison character type”. The processes after step SG3 are performed once in turn for each of the 14 character types.
Next, the character recognition unit 80 refers to the shrinkage table 94 (step SG3).

FIG. 17A is a diagram schematically showing the data structure of the shrinkage table 94.
As shown in FIG. 17A, in the contraction table 94, information indicating the expansion / contraction point and information indicating the maximum contraction amount are stored in association with each other for every 14 character types.
Note that, for the same character type in the both-side expansion / contraction table 90 and the contraction table 94, the maximum contraction amount value in the contraction table 94 is equal to or greater than the maximum contraction amount value in the double-side expansion / contraction table 90.

  Next, the character recognizing unit 80 refers to the record corresponding to the comparison target character type in the referred contraction table 94, and acquires the respective values of the expansion / contraction point and the maximum contraction amount related to the comparison target character type (step S40). SG4).

  Next, the character recognition unit 80 corrects the waveform of the reference waveform data corresponding to the character type to be compared based on the expansion / contraction point and the maximum contraction amount acquired in step SG4, The difference between the waveform of the character waveform data corresponding to the character is calculated (step SG5).

More specifically, the character recognizing unit 80 sets the maximum amount of shrinkage with the maximum shrinkage amount as an upper limit for the waveform to the right of the X axis from the stretch point among the waveforms of the reference waveform data corresponding to the character type to be compared. While subtracting “one sampling unit” at a time, the waveform of the reference waveform data is corrected based on the maximum contraction amount after the subtraction, and between the waveform of the corrected reference waveform data and the waveform of the character waveform data of the character to be processed The amount of difference is calculated. The calculation of the difference amount is performed until the value of the maximum contraction amount after subtraction becomes “0 sampling unit”.
For example, when the comparison target character type is the character type “0”, the maximum contraction amount is “3 sampling units” with reference to the contraction table 94 in FIG. In this case, the character recognizing unit 80 processes each of the waveforms of the reference waveform data corrected by sliding the waveform to the right of the expansion / contraction point by sliding it to the left of the X axis by 3, 2, 1, 0 sampling units, and the processing target. Each difference between the waveform of the character waveform data corresponding to the character is calculated.

  Next, the character recognizing unit 80 specifies the smallest difference value calculated in step SG5, and identifies the identified difference amount as the difference amount of the comparison target character type (step SG6).

Next, the character recognition unit 80 determines whether or not the calculation of the difference amount (step SG5) and the specification (step SG6) have been completed for all 14 character types (step SG7).
When the calculation and identification of the difference amount are not completed for all character types (step SG7: NO), the character recognition unit 80 returns the processing procedure to step SG2.
On the other hand, when the calculation and identification of the difference amount have been completed for all the character types (step SG7: YES), the character recognition unit 80 determines that the value of the identified difference amount is the highest among the 14 character types. The smaller character type is set as the first candidate C51 for pass 5, and the character type whose value of the specified difference amount is the second smallest is set as the second candidate C52 for pass 5 (step SG8).

Next, the character recognition unit 80 compares the difference amount specified in step SG6 with the pass 5 threshold value T5 for the character type determined as the pass 5 first candidate C51, and the difference amount is determined as the pass 5 threshold value. It is determined whether or not T5 is exceeded (step SG9).
The pass 5 threshold value T5 is a value serving as a reference when determining whether or not the character type set as the first candidate C51 for pass 5 is to be determined as the character type of the character to be processed. In the case of the threshold value T5 or less, if other conditions described later are satisfied, the character recognition unit 80 determines the character type set as the first candidate C51 for pass 5 as the character type of the processing target character.

When the difference amount exceeds the threshold value 5 for pass 5 (step SG9: YES), the character recognition unit 80 ends the fifth recognition pass without determining the character type of the processing target character.
When the difference amount is equal to or less than the pass 5 threshold value T5 (step SG9: NO), the character recognition unit 80 further determines the difference amount specified in step SG6 for the character type determined as the pass 5 second candidate C52. The pass 5 threshold value T5 is compared to determine whether or not the difference amount exceeds the pass 5 threshold value T5 (step SG10).

When the difference amount exceeds the pass 5 threshold value T5 (step SG10: YES), the character recognition unit 80 determines the character type set as the first pass 5 candidate C51 as the character type of the processing target character (step SG11), the fifth recognition pass is terminated.
On the other hand, when the difference amount is equal to or less than the pass 5 threshold value T5 (step SG10: NO), the character recognition unit 80 ends the fifth recognition pass without determining the character type of the processing target character.

  As described above, in the present embodiment, even if the difference amount specified in step SG6 is less than or equal to the pass 5 threshold T5 for the character type that is the first candidate C51 for pass 5, For the character type determined as the second candidate C52, when the difference amount specified in step SG6 is equal to or less than the pass 5 threshold T5, the character type of the character to be processed is not fixed. This is due to the same reason as described for the above-described pass 1 threshold T1.

  In this embodiment, if there is a possibility that “shrinkage” has occurred in the waveform of the character waveform data corresponding to the character to be processed, the character recognition unit 80 uses the reference waveform corresponding to the character type to be compared. Among the data waveforms, the waveform on the right side of the X axis from the expansion / contraction point is slid to the X axis “left” by a predetermined amount with the maximum shrinkage amount as the upper limit. The waveform of the reference waveform data is corrected by this slide, and the difference amount between the corrected waveform of the reference waveform data and the waveform of the character waveform data of the character to be processed is calculated.

Here, among the waveforms of the reference waveform data, when the waveform to the right of the X-axis from the expansion / contraction point is slid to the left in the X-axis direction, the entire waveform of the reference waveform data is left in the X-axis within a predetermined range. It will be in the state shrunk toward (for example, refer to Drawing 16).
In other words, in this embodiment, if there is a possibility that “shrinkage” has occurred in the waveform of the character waveform data, the waveform of the reference waveform data is shrunk in correspondence with the “shrinkage” and then the difference is made. The amount is calculated. As a result, the difference amount can be calculated appropriately in response to the “shrinkage” of the waveform of the character waveform data, and the reference waveform data corrected for “elongation” and the character waveform of the character to be processed Since a difference amount is not calculated from data, processing efficiency is improved.

On the other hand, if it is determined in step SG1 that there is a possibility that “elongation” has occurred (step SG1: “elongation”), the character recognizing unit 80 performs step SG13 and subsequent steps out of 14 character types. A character type to be compared in the process is specified (step SG12). Hereinafter, the character type specified in step SG12 is referred to as “comparison character type”. The process after step SG13 is performed 14 times in order for each of the 14 character types.
Next, the character recognition unit 80 refers to the stretch table 95 (step SG13).

FIG. 17B is a diagram schematically showing the data structure of the extension table 95.
As shown in FIG. 17B, in the stretch table 95, information indicating the expansion / contraction point and information indicating the maximum expansion amount are stored in association with each other for each of the 14 character types.
For the same character type in the both-side expansion / contraction table 90 and the extension table 95, the value of the maximum extension amount in the extension table 95 is equal to or greater than the value of the maximum extension amount in the both-side extension table 90.

  Next, the character recognizing unit 80 refers to the record corresponding to the character type to be compared in the referenced expansion table 95, and acquires the respective expansion / contraction points and the maximum expansion amount related to the character type to be compared (step). SG14).

  Next, the character recognition unit 80 corrects the waveform of the reference waveform data corresponding to the character type to be compared based on the expansion / contraction point acquired in step SG14 and the maximum extension amount, A difference amount between the waveform of the character waveform data corresponding to the character is calculated (step SG15).

More specifically, the character recognizing unit 80 sets the maximum elongation amount with respect to the waveform of the reference waveform data corresponding to the character type to be compared with the maximum elongation amount as the upper limit for the waveform to the right of the X axis from the expansion / contraction point. While subtracting by “one sampling unit”, the waveform of the reference waveform data is corrected by sliding to the right of the X axis by the maximum extension amount after subtraction, and the waveform of the corrected reference waveform data and the character of the processing target character are corrected. The amount of difference between the waveform of the character waveform data is calculated. This difference amount is detected until the maximum elongation value after subtraction becomes “0 sampling unit”.
For example, when the character type to be compared is the character type “0”, the maximum extension amount is “6 sampling units” with reference to the extension table 95 in FIG. In this case, the character recognizing unit 80 corrects the waveform of the reference waveform data corrected by sliding the waveform on the right side of the expansion / contraction point to the right of the X axis by 6, 5, 4, 3, 2, 1, 0 sampling units. And the amount of difference between the waveform and the waveform of the character waveform data corresponding to the character to be processed is calculated.

  Next, the character recognizing unit 80 identifies the difference amount calculated in step SG15 having the smallest value, and identifies the identified difference amount as the difference amount of the comparison target character type (step SG16).

  Next, the character recognizing unit 80 determines whether the calculation of the difference amount (step SG15) and the specification (step SG16) have been completed for all 14 character types (step SG17).

When the calculation and identification of the difference amount are not completed for all character types (step SG17: NO), the character recognition unit 80 returns the processing procedure to step SG12.
On the other hand, when the calculation and identification of the difference amount have been completed for all character types (step SG17: YES), the character recognition unit 80 executes the processes of step SG18 to step SG21. Since the process of step SG18-step SG21 is the same as the process of step SG8-step SG11 mentioned above, the description is abbreviate | omitted.

  As described above, in this embodiment, when there is a possibility that “elongation” has occurred in the waveform of the character waveform data corresponding to the character to be processed, the character recognition unit 80 corresponds to the character type to be compared. Of the waveform of the reference waveform data, the waveform on the right side of the X-axis from the expansion / contraction point is slid to the right of the X-axis by a predetermined amount with the maximum elongation value as the upper limit. The waveform of the reference waveform data is corrected by this slide, and the difference amount between the corrected waveform of the reference waveform data and the waveform of the character waveform data of the character to be processed is calculated.

Here, among the waveforms of the reference waveform data, when the waveform on the X axis right side from the expansion / contraction point is slid to the X axis right side, the entire waveform of the reference waveform data is X axis right side within a predetermined range. It will be in the state extended toward (for example, refer to Drawing 16).
In other words, in this embodiment, if there is a possibility that “elongation” has occurred in the waveform of the character waveform data, the waveform of the reference waveform data is expanded corresponding to the “elongation”, and then the difference amount Is calculated. As a result, it is possible to appropriately calculate the difference amount corresponding to the “elongation” of the waveform of the character waveform data, and to perform the correction corresponding to “shrinkage” and the character waveform of the character to be processed. Since a difference amount is not calculated from data, processing efficiency is improved.

Now, referring to FIG. 8, in the present embodiment, when the character spacing of the character to be processed is equal to the reference character spacing (step SE2: “character spacing = reference character spacing”), the third recognition pass causes the processing target to be processed. Recognize characters. This is due to the following reason.
In other words, when the character spacing of the character to be processed is equal to the reference character spacing, the waveform of the character waveform data of the character to be processed is often not stretched or shrunk, and the stretch or shrinkage has occurred. Even if it is, it cannot be said that there is a high possibility that one of the expansion and the contraction has occurred. Based on this, in the above case, it is configured to execute only the third recognition pass that performs recognition of the character to be processed after calculating the difference amount reflecting both expansion and contraction, thereby improving processing efficiency. I am trying. In the present embodiment, if the character spacing of the character to be processed is equal to the reference character spacing, it is highly likely that the reading by the magnetic head 54 or the conveyance of the check 4 has been executed normally. Although the configuration is such that the fourth recognition pass is not performed with strict type determination conditions, the configuration may be such that the fourth recognition pass is performed before the execution of the third recognition pass.

Referring to FIG. 8, in the present embodiment, when the character interval of the character to be processed is equal to the value obtained by subtracting “one sampling unit” from the reference character interval (step SE2: “character interval = reference character interval— 1 "), the character recognition unit 80 executes the recognition pass step by step in the order of the fourth recognition pass (step SE4), the fifth recognition pass (step SE5), and the third recognition pass (step SE6). This is due to the following reason.
That is, when the character interval of the processing target character is equal to the value obtained by subtracting “one sampling unit” from the reference character interval, the character interval is smaller by “one sampling unit” than the reference character interval, and the waveform of the processing target character That is, the width on the X-axis is reduced by one sampling unit as compared with the normal state. In this case, it can be said that the possibility that the waveform of the character to be processed is contracted is higher than the possibility that the waveform of the character to be processed is extended. However, compared with the case where the difference between the character spacing and the reference character spacing is “1 sampling unit” or more, the possibility that “shrinkage” occurs in the waveform of the character to be processed is “stretching”. Compared to the possibility of occurrence of the occurrence of a "shrinkage", it is not extremely high, and even if a "shrinkage" occurs, the possibility of a very small shrinkage corresponding to "one sampling unit" Is expensive.
In consideration of this, in the present embodiment, in the above case, the process related to the recognition is executed step by step in the order of the fourth recognition pass, the fifth recognition pass, and the third recognition pass.

  In detail, considering that the possibility that “shrinkage” has occurred in the waveform of the character to be processed is not significantly higher than the possibility that “elongation” has occurred, The fourth recognition pass for executing the recognition of the processing target character after calculating the difference amount by reflecting both the contraction and the contraction is performed to efficiently recognize the processing target character. If the character type of the processing target character cannot be determined in the fourth recognition pass, the possibility that the waveform of the processing target character is contracted is higher than the possibility that the waveform of the processing target character is expanded Then, the fifth recognition pass is executed to recognize the processing target character specialized when the waveform of the processing target character is contracted.

If there is a possibility that “shrinkage” has occurred in the waveform of the character to be processed (step SG1: “shrinkage” in FIG. 10), the shrinkage table 94 will be referred to in the fifth recognition pass. As described above, for the same character type in the both-side expansion / contraction table 90 and the contraction table 94, the value of the maximum contraction amount in the contraction table 94 is greater than or equal to the value of the maximum contraction amount in the both-side expansion / contraction table 90. ing. Therefore, the fifth recognition pass can realize the recognition of the processing target character corresponding to the “shrinkage” of the waveform of the processing target character than the fourth recognition pass.
When the character type of the processing target character cannot be determined in the fifth recognition pass, the third recognition pass is executed, which has a looser condition for determining the character type than the fourth recognition pass. This improves the success rate of character type determination while ensuring the reliability that the determined character type is the correct character type.

Referring to FIG. 8, in the present embodiment, when the character interval of the character to be processed is less than a value obtained by subtracting “one sampling unit” from the reference character interval (step SE2: “character interval <reference character interval−1”). ), The character recognition unit 80 executes the recognition pass step by step in the order of the fifth recognition pass (step SE7) and the third recognition pass (step SE8). This is due to the following reason.
That is, when the character spacing of the processing target character is less than the value obtained by subtracting “one sampling unit” from the reference character spacing, it can be said that the possibility that the waveform of the processing target character is contracted is relatively high.
Based on this, first, the character recognition unit 80 first executes a fifth recognition pass for recognizing the processing target character specialized when the waveform of the processing target character is contracted, and efficiently recognizes the processing target character. I do. If the character type of the processing target character cannot be determined in the fifth recognition pass, the processing unit character recognition is performed after calculating the difference amount reflecting both expansion and contraction. A recognition pass is in progress.

Referring to FIG. 8, in the present embodiment, when the character interval of the character to be processed is equal to the value obtained by adding “one sampling unit” to the reference character interval (step SE2: “character interval = reference character interval + 1”). ”), The character recognition unit 80 executes the recognition pass step by step in the order of the fourth recognition pass (step SE9), the fifth recognition pass (step SE10), and the third recognition pass (step SE11). This is due to the following reason.
That is, when the character interval of the processing target character is equal to the value obtained by adding “one sampling unit” to the reference character interval, the character interval is “one sampling unit” larger than the reference character interval, and the waveform of the processing target character is This means that the width on the X-axis is extended by one sampling unit as compared with the normal state. In this case, it can be said that the possibility that the waveform of the processing target character is expanded is higher than the possibility that the waveform of the processing target character is contracted. However, compared with the case where the difference between the character spacing and the reference character spacing is “1 sampling unit” or more, the possibility that “elongation” has occurred in the waveform of the character to be processed is “shrinkage”. Compared to the possibility of occurrence of, it is not extremely high, and even if "elongation" occurs, there is a possibility of very small elongation corresponding to "one sampling unit" Is expensive.
In consideration of this, in the present embodiment, in the above case, the process related to the recognition is executed step by step in the order of the fourth recognition pass, the fifth recognition pass, and the third recognition pass.

  In detail, considering that the possibility of “stretching” in the waveform of the character to be processed is not significantly higher than the possibility of “shrinking”, The fourth recognition pass for executing the recognition of the processing target character after calculating the difference amount by reflecting both the contraction and the contraction is performed to efficiently recognize the processing target character. If the character type of the character to be processed cannot be determined in the fourth recognition pass, the possibility that the waveform of the character to be processed is stretched is higher than the possibility that the waveform of the character to be processed is contracted Then, the fifth recognition pass is executed to recognize the processing target character specialized when the waveform of the processing target character is extended.

When there is a possibility that “elongation” has occurred in the waveform of the character to be processed (step SG1: “elongation” in FIG. 10), the extension table 95 is referred to in the fifth recognition pass. As described above, in the both-side expansion / contraction table 90 and the extension table 95, for the same character type, the value of the maximum extension amount in the extension table 95 is equal to or greater than the value of the maximum extension amount in the both-side extension table 90. ing. Therefore, the fifth recognition pass can realize the recognition of the processing target character corresponding to the “elongation” of the waveform of the processing target character than the fourth recognition pass.
When the character type of the processing target character cannot be determined in the fifth recognition pass, the third recognition pass is executed, which has a looser condition for determining the character type than the fourth recognition pass. This improves the success rate of character type determination while ensuring the reliability that the determined character type is the correct character type.

Referring to FIG. 8, in the present embodiment, when the character spacing of the character to be processed exceeds a value obtained by adding “one sampling unit” to the reference character spacing (step SE2: “character spacing> reference character spacing + 1”). The character recognition unit 80 executes the recognition pass step by step in the order of the fifth recognition pass (step SE12) and the third recognition pass (step SE13). This is due to the following reason.
That is, if the character interval of the character to be processed exceeds the value obtained by adding “one sampling unit” to the reference character interval, it can be said that there is a relatively high possibility that the waveform of the character to be processed is extended.
Based on this, first, the character recognition unit 80 first executes the fifth recognition pass for recognizing the processing target character specialized when the waveform of the processing target character is extended, and efficiently recognizes the processing target character. I do. If the character type of the processing target character cannot be determined in the fifth recognition pass, the processing unit character recognition is performed after calculating the difference amount reflecting both expansion and contraction. A recognition pass is in progress.

  The processing in the second recognition phase has been described above. However, if the character type of the processing target character cannot be determined in the second recognition phase, the third recognition pass is always executed, and the execution of the third recognition pass is performed. Thus, the first candidate for path 3 C31 and the second candidate for path 3 C32 are identified. Further, the difference amounts specified for the respective character types that are the first candidate for pass 3 C31 and the second candidate for pass 3 C32 are stored until the subsequent steps.

Now, referring to FIG. 4, the character recognition unit 80 determines whether or not the character type of the character to be processed has been determined in the second recognition phase after the second recognition phase (step SA7) is executed (step SA8). ) If it can be confirmed (step SA8: YES), the process proceeds to step SA5.
When it cannot be determined (step SA8: NO), the difference specified for each of the first candidate C21 for pass 2 and the second candidate C22 for pass 2 specified in the second recognition pass of the first recognition phase The amount is compared with the difference amount specified for each of the path 3 first candidate C31 and the path 3 second candidate C32 specified in the third recognition pass of the second recognition phase (step SA9).

As a result of the comparison of the difference amount in step SA9, the character recognition unit 80 determines the character type of the candidate having the smallest difference amount among the four candidates as the first candidate C41 in magnetic ink character recognition, and the difference amount is The second smallest character type is determined as the second candidate C42 in magnetic ink character recognition (step SA10), and the processing procedure proceeds to step SA5. However, if the character type of the candidate having the second smallest amount of difference is the same as the character type of the candidate having the smallest amount of difference, the character type having the next smallest amount of difference is designated as the second candidate in magnetic ink character recognition. C42.
In an optical recognition process (step SA18), which will be described later, it is determined whether or not these first candidate C41 and second candidate C42 match the characters temporarily determined in the optical character recognition.

Here, in the second recognition pass of the first recognition phase, assuming that there is a deviation in the waveform of the character waveform data of the character to be processed, the difference amount is calculated after reflecting the deviation. The expansion and contraction of the waveform is not reflected. On the other hand, in the third recognition pass of the second recognition phase, the difference amount is calculated after reflecting the expansion / contraction assuming that the waveform of the character waveform data of the character to be processed has expansion / contraction. The waveform shift is not reflected.
Therefore, for example, when there is a shift in the character cutout position in step SA1, the difference amount calculated in the second recognition pass of the first recognition phase is the difference amount calculated in the third recognition pass of the second recognition phase. Is likely to be smaller. Further, when there is expansion or contraction of magnetic ink characters due to printing or the like, the difference amount calculated in the third recognition pass of the second recognition phase is larger than the difference amount calculated in the second recognition pass of the first recognition phase. Is likely to be smaller.
Therefore, the difference amount calculated in the second recognition pass of the first recognition phase is compared with the difference amount calculated in the third recognition pass of the second recognition phase, and the character type with the smaller difference amount is selected. By using candidates, it is possible to obtain character type candidates with higher certainty.

Now, referring to FIG. 4 described above, when it is determined in step SA5 that all magnetic ink characters included in the magnetic ink character string 4A have been processed (step SA5: YES), the character recognition unit 80 The number of magnetic ink characters included in the magnetic ink character string 4A is detected (step SA6).
Next, the character recognition unit 80 determines whether or not the character type has been determined for all the magnetic ink characters included in the magnetic ink character string 4A (step SA16).

When the character type is determined for all the magnetic ink characters (step SA16: YES), the host-side control unit 73 assumes that the magnetic ink character string 4A has been successfully read, and that the magnetic ink character string 4A has been successfully read. The process to be performed is performed (step SA17). The processing to be performed when the reading of the magnetic ink character string 4A is successful includes, for example, storing the information indicated by the magnetic ink character string 4A in the storage unit 78, and causing the control unit 71 to control the recording device 56, A predetermined image related to the endorsement is recorded on the back surface of the check 4, and the first check discharge unit 7 discharges the check 4.
On the other hand, when at least one of the magnetic ink characters included in the magnetic ink character string 4A has not been determined (step SA16: NO), the character recognition unit 80 has not determined the character type. An optical recognition process (step SA18) for executing optical character recognition on the magnetic ink character is executed.

  Next, the character recognition unit 80 determines whether or not the character type has been determined for all the magnetic ink characters whose character type has not been determined by the optical recognition process (step SA19), and for all the magnetic ink characters. If the character type is determined (step SA19: YES), processing to be performed when the magnetic ink character string 4A is successfully read is performed (step SA17). On the other hand, if there is at least one magnetic ink character whose character type has not been determined (step SA19: NO), the host-side control unit 73 performs processing to be performed when reading of the magnetic ink character string 4A fails. (Step SA20).

  As a process to be performed when reading of the magnetic ink character string 4A fails, the host-side control unit 73 controls the control unit 71 to perform the second check discharge unit without printing an image related to the endorsement. The process of conveying the check 4 to 8 is performed. The check 4 discharged to the second check discharge unit 8 is subjected to processing such as investigation of the cause of reading failure and re-reading.

FIG. 11 is a flowchart showing the operation of the character recognition unit 80 when the optical recognition process in step SA18 is executed.
In the optical recognition process, first, the character recognition unit 80 is included in the magnetic ink character string 4A based on data indicating the image of the surface of the check 4 generated based on the reading result of the front-side contact image sensor 52. Optical character recognition of magnetic ink characters is executed (step SK1).

In step SK1, for example, the character recognition unit 80 specifies the range of data corresponding to the image of the magnetic ink character string 4A in the image data indicating the surface of the check 4, and uses the magnetic ink character for the specified range of data. The image data is cut out every time.
Next, the character recognition unit 80 tentatively determines each character type of the magnetic ink character by comparing the bitmap pattern corresponding to each of the 14 character types with the image data of the cut out magnetic ink character. . Note that the character type determined in step SK1 is not final character type determination, and is therefore expressed as “provisional determination”.

  Next, the character recognition unit 80 detects the number of magnetic ink characters included in the magnetic ink character string 4A based on the result of optical character recognition in step SK1 (step SK2).

  Next, the character recognition unit 80 matches the number of characters detected in step SA6 in FIG. 4, that is, the number of characters detected based on the detection signal waveform data read by the magnetic head 54, and the number of characters detected in step SK2. It is determined whether or not to perform (step SK3).

If the number of characters does not match (step SK3: NO), the character recognition unit 80 determines that reading of the magnetic ink character string 4A has failed (step SK4), and ends the optical recognition process. This is due to the following reason.
That is, the processing after step SK5 is processing for attempting to determine the character type for the magnetic ink character whose character type has not been determined using the result of the optical character recognition processing at step SK1. Therefore, when the number of characters is different, certainty can be guaranteed for the correspondence between each magnetic ink character cut out based on the detection signal waveform data and each magnetic ink character cut out by optical character recognition. This is because the character type cannot be determined using the result of the optical character recognition process in step SK1.

  When the number of characters matches (step SK3: YES), the character recognition unit 80 processes one magnetic ink character from the magnetic ink characters determined to be unable to determine the character type in step SA8 in FIG. (Step SK5). The processes after step SK5 are sequentially performed for each magnetic ink character whose character type cannot be determined by magnetic ink character recognition. Hereinafter, the magnetic ink character specified in step SK5 is referred to as a “processing target character”.

Next, the character recognition unit 80 acquires the character type temporarily determined in step SK1 for the processing target character (step SK6).
Next, the character recognition unit 80 determines the character type temporarily determined in step SK1, the character type determined as the first candidate C41 in step SA10, and the character type determined as the second candidate C42 for the processing target character. It is determined whether or not any of the above matches (step SK7).

If they match (step SK7: YES), the character recognizing unit 80 determines the temporarily determined character as the character type of the processing target character (step SK8), and shifts the processing procedure to step SK9.
On the other hand, if they do not match (step SK7: NO), the character recognition unit 80 determines that the character type of the processing target character cannot be determined even by character recognition using optical character recognition (step SK10), and the processing procedure is stepped. Move to SK9.

In step SK9, the character recognizing unit 80 determines whether or not all the magnetic ink characters determined to be unable to determine the character type in step SA8 in FIG. 4 have been processed. If not all magnetic ink characters are characters to be processed (step SK9: NO), the character recognition unit 80 returns the processing procedure to step SK5.
On the other hand, when all the magnetic ink characters have been processed (step SK9: YES), the character recognition unit 80 determines whether there is even one magnetic ink character whose character type cannot be determined, that is, in step SK10. Then, it is determined whether or not there is even one magnetic ink character determined that the character type cannot be determined (step SK11).

If there is no character type that cannot be determined (step SK11: NO), the character recognition unit 80 determines that the magnetic ink character string 4A has been successfully read (step SK12), and ends the optical recognition process. To do.
On the other hand, if there is even one character type that cannot be determined (step SK11: YES), the character recognition unit 80 determines that reading of the magnetic ink character string 4A has failed (step SK4), and optical recognition is performed. The process ends.

As described above, in the present embodiment, the character type of the character to be processed is provisionally determined in the optical character recognition process of step SK1, and the character type that has been provisionally determined is the character type that is determined as the first candidate C41 in step SA10, and Only when it matches one of the character types determined as the second candidate C42, the temporarily determined character type is determined as the character type of the processing target character. This is due to the following reason.
That is, in the optical character recognition process in step SK1, the character type of the magnetic ink character is provisionally determined by optical character recognition based on the image data indicating the surface of the check 4. Here, in many cases, a signature is written on the surface of the check 4 with a pen or the like, and it is known that the signature may cause erroneous recognition of the magnetic ink character string 4A.
Based on this, in the present embodiment, the character type is not determined simply by optical character recognition for the magnetic ink character for which the character type cannot be determined by magnetic processing. The character type temporarily determined by the optical character recognition is the character type related to the first candidate C41 that is most likely to be the correct character type from the recognition results in the first recognition phase and the second recognition phase, and is correct. The character type temporarily determined is determined as the character type of the processing target character only when the character type matches with any one of the character types related to the second candidate C42 that is considered to be the second highest in probability. As a result, erroneous recognition is effectively suppressed.

  In particular, the first candidate C41, which has the highest probability of being the correct character type from the recognition results of both the first recognition phase and the second recognition phase, and the second highest probability of being the correct character type The possible second candidates C42 are selected, and it can be said that the reliability of the character type as a candidate is high. Therefore, by using these first candidate C41 and second candidate C42 to determine the character type, it is possible to obtain a highly reliable recognition result in which erroneous recognition is suppressed.

  For each of the magnetic ink characters included in the magnetic ink character string 4A, the character type determined by magnetic ink character recognition and the character type determined by optical character recognition are determined, and the character type determined by magnetic processing is determined. If the character types determined by optical character recognition match, the final determination of the character type of the magnetic ink character may be performed to reduce the erroneous recognition of the magnetic ink character. However, in this case, it is conceivable that the success rate (recognition rate) of character type determination is reduced, although the erroneous recognition is reduced. In this embodiment, since the operation | movement mentioned above is performed, the fall of a recognition rate can be prevented effectively, suppressing misrecognition.

As described above, this embodiment includes the first recognition phase and the second recognition phase for recognizing magnetic ink characters by different methods, and the results of magnetic ink character recognition in these processing phases. Based on each, the character type which a magnetic ink character shows is decided.
Thereby, based on the results of magnetic ink character recognition in the two processing phases for recognizing magnetic ink characters by different methods, the character type indicated by the magnetic ink characters can be recognized in many ways, and erroneous recognition of magnetic ink characters can be performed. Can be suppressed.

In the present embodiment, in the first recognition phase, according to the difference between the detected waveform obtained by reading the magnetic ink character and the waveform obtained by sliding each of the reference waveforms of a plurality of character types defined in the standard by a predetermined distance. The first difference C21 for pass 2 and the second candidate C22 for pass 2 are selected as candidates for the character type of the magnetic ink character based on the difference amount calculated by reflecting the difference.
Thereby, when the position of the detected waveform is deviated from the ideal waveform, the magnetic ink character can be recognized while suppressing the influence of the deviation.

In the present embodiment, in the second recognition phase, a difference between a detection waveform obtained by reading a magnetic ink character and a waveform obtained by expanding and contracting each of a plurality of character type reference waveforms defined by the standard by a predetermined distance. Based on the difference amount calculated by reflecting the expansion and contraction, the first candidate C31 for pass 3 and the second candidate C32 for pass 3 are selected as the character type candidates for the magnetic ink characters. Sort out.
Thereby, when expansion / contraction has occurred in the detected waveform, the influence of the expansion / contraction can be suppressed and magnetic ink characters can be recognized.

Furthermore, in this embodiment, when the character type of the character to be processed cannot be determined until the second recognition phase, in step SA9, the first candidate C21 for pass 2 selected in the first recognition phase and the pass 2 Each difference amount of the second candidate C22 for use is compared with each difference amount of the first candidate C31 for pass 3 and the second candidate C32 for pass 3 selected in the second recognition phase. In step SA10, the first candidate C41 and the second candidate C42 of the character type of the magnetic ink character are determined in ascending order of the difference amount from the comparison result in step SA9.
Therefore, even when the character type of the character to be processed cannot be determined by magnetic ink character recognition, it is considered that the probability of being the correct character type is the highest from the recognition results in both the first recognition phase and the second recognition phase. A candidate C41 and a second candidate C42 that is considered to have the second highest probability of being the correct character type are selected. Thereby, since a candidate with higher certainty can be obtained as compared with the case of selecting character type candidates from the result of only one processing phase, erroneous recognition of magnetic ink characters can be effectively suppressed.

(Second Embodiment)
Next, a recording medium processing apparatus, a control method for the recording medium processing apparatus, and a program according to the second embodiment will be described.

  The recording medium processing apparatus, the recording medium processing apparatus control method, and the program according to the second embodiment are the same as the recording medium processing apparatus, the recording medium processing apparatus control method, and the program according to the first embodiment. Thus, the difference is that the processes of steps SA11, SA12, SA13, and SA14 are further included, but the other configurations are substantially the same. FIG. 18 is a flowchart showing the operation of the character recognition unit 80 according to the second embodiment. In addition, about the structure which is common in 1st Embodiment, the same code | symbol is attached | subjected and the description is abbreviate | omitted.

  In the second embodiment, the difference between the first candidate C21 for pass 2 and the second candidate C22 for pass 2 selected in the first recognition phase in step SA9 shown in FIG. The difference amounts of the first candidate C31 for pass 3 selected in the phase and the second candidate C32 for pass 3 are compared.

In the comparison of step SA9, when the difference amount of the first candidate C21 for pass 2 selected in the first recognition phase is smaller than the difference amount of the first candidate C31 for pass 3 selected in the second recognition phase (step (SA11: YES), the character recognition unit 80 slides the waveform of the reference waveform data by, for example, the amount (slide amount) slid with respect to the first candidate C21 for pass 2 to thereby slide the first recognition phase. Is reflected (step SA12).
It is assumed that the slide amount of the waveform of the reference waveform data in step SD2 when the difference amount is calculated in the first recognition phase is stored.

  Next, the second recognition phase is executed again in a state in which the slide amount in the first recognition phase is reflected in the waveform of the reference waveform data (step SA13). The process in the second recognition phase (step SA13) for the second time is the same as the process in the second recognition phase (step SA7) for the first time except that the process is performed by reflecting the slide amount in the first recognition phase. is there.

  After execution of the second recognition phase (step SA13) for the second time, the character recognition unit 80 determines whether or not the character type of the character to be processed has been confirmed (step SA14), and if it can be confirmed (step SA14: YES), the process proceeds to step SA5.

If it cannot be determined (step SA14: NO), the difference amounts of the first candidate C31 for pass 3 and the second candidate C32 for pass 3 newly selected in the second second recognition phase, The difference amounts of the first candidate C21 for pass 2 and the second candidate C22 for pass 2 selected in the first recognition phase are compared.
As a result of the comparison of the difference amounts, the character type of the candidate having the smallest difference amount among the four candidates is determined as the first candidate C41 in the magnetic ink character recognition, and the character type having the second smallest difference amount is determined. It determines as the 2nd candidate C42 in magnetic ink character recognition (step SA10), and transfers a process sequence to step SA5.
However, if the character type of the candidate having the second smallest amount of difference is the same as the character type of the candidate having the smallest amount of difference, the character type having the next smallest amount of difference is designated as the second candidate in magnetic ink character recognition. C42.

Thus, in the comparison of step SA9, when the difference amount of the first candidate C21 for pass 2 in the first recognition phase is smaller than the difference amount of the first candidate C31 for pass 3 in the second recognition phase, the first recognition The reason why the second recognition phase is executed again by reflecting the slide amount in the phase is as follows.
That is, the difference amount of the character type selected as the first candidate C21 for pass 2 in the first recognition phase is smaller than the difference amount of the character type selected as the first candidate C31 for pass 3 in the second recognition phase. It is considered that there is a high possibility that the position of the detected waveform is shifted due to the shift of the character cutout position in step SA1. In such a case, even if the difference amount is calculated by reflecting the expansion / contraction on the reference waveform in a state where the position of the waveform remains shifted (a state where the slide amount is not reflected), the difference amount may not be reduced. high.
Therefore, in such a case, if the second recognition phase is executed for the second time after the waveform of the reference waveform data is slid by the slide amount in the first recognition phase, the displacement of the position of the detected waveform is reflected, Furthermore, since the expansion / contraction of the waveform is reflected, it is possible to recognize the magnetic ink character while suppressing the influence of the position shift of the detection waveform and the expansion / contraction. Thereby, a recognition rate can be improved.

The difference amount of the character type selected as the first candidate C21 for pass 2 in the first recognition phase is larger than the difference amount of the character type selected as the first candidate C31 for pass 3 in the second recognition phase. However, it cannot be said that there is a high possibility that the position of the detected waveform is shifted. In such a case, if the slide amount in the first recognition phase is reflected and the second recognition phase is executed again, the difference amount between the detected waveform and the reference waveform may be increased.
Therefore, in the present embodiment, the second recognition is performed by reflecting the slide amount in the first recognition phase only when the difference amount of the first candidate C21 for pass 2 is smaller than the difference amount of the first candidate C31 for pass 3. The phase is to be executed again.

(Third embodiment)
Next, a recording medium processing device, a control method for the recording medium processing device, and a program according to a third embodiment will be described.

  The recording medium processing apparatus, the control method for the recording medium processing apparatus, and the program according to the third embodiment are the first candidate C21 for pass 2 selected in the first recognition phase with respect to the second embodiment, And when the second candidate C22 for path 2 is “2” and “5”, it is different in that the processes of steps SA11, SA12, SA13, and SA14 are not executed, but the other configurations are almost the same. is there. FIG. 19 is a flowchart showing the operation of the character recognition unit 80 according to the third embodiment. In addition, about the structure which is common in 2nd Embodiment, the same code | symbol is attached | subjected and the description is abbreviate | omitted.

  In the third embodiment, the character types of the first candidate C21 for pass 2 and the second candidate C22 for pass 2 selected in the first recognition phase in step SA9 shown in FIG. 19 are “2” and “ 5 ”(step SA15).

  One of the second candidate C21 for path 2 and the second candidate C22 for path 2 is either one of “2” and “5”, and the first candidate C21 for path 2 and path 2 When the other second candidate C22 is the other of “2” and “5” (step SA15: YES), the processes of steps SA11, SA12, SA13, and SA14 are not executed. In this case, the candidate character type having the smallest difference amount among the four candidates is determined as the first candidate C41 in magnetic ink character recognition, and the character type having the second smallest difference amount is determined as the magnetic ink character recognition. Is determined as the second candidate C42 (step SA10), and the processing procedure proceeds to step SA5.

  On the other hand, when the first candidate for path 2 C21 and the second candidate for path 2 C22 are other than “2” and “5” (step SA15: NO), as in the second embodiment, steps SA11, The processing of SA12, SA13, and SA14 is executed.

  The reason why the processes of steps SA11, SA12, SA13, and SA14 are not executed when the first candidate C21 for path 2 and the second candidate C22 for path 2 are “2” and “5” will be described below. FIG. 20A is a diagram schematically showing the content of the reference waveform data corresponding to the character type “2” in the E-13B font, and FIG. 20B shows the character type “5” in the E-13B font. It is a figure which shows typically the content of corresponding reference waveform data.

As shown in FIGS. 20A and 20B, the character waveform data of the character type “2” and the character waveform data of the character type “5” in the E-13B font are similar to each other. More specifically, the portion W2a from the first sampling unit to the 28th sampling unit on the one end side in the character waveform data “2” shown in FIG. 20A and the character “5” shown in FIG. In the waveform data, the waveform and the peak position are substantially the same in the portion W5a of the first sampling unit to the twenty-eighth sampling unit on one end side.
Further, in the character waveform data “2” shown in FIG. 20A, the portion W2b from the 29th sampling unit to the 57th sampling unit on the other end side and the character waveform data “5” shown in FIG. The waveform is substantially the same in the portion W5b from the 36th sampling unit to the 64th sampling unit on the other end side. The peak position of the portion W2b and the peak position of the portion W5b are shifted by about 7 sampling units. When the portion W5b is slid to the left in the X-axis direction in FIG. 20B by 7 sampling units, the portion W2b The waveform and the peak position are substantially the same in the portion W5b.

As described above, in the second recognition pass (step SD2) of the first recognition phase, the entire waveform of the reference waveform data is slid to the left by 7 sampling units in the X axis direction, and the difference amount is calculated. Slide to the right for 7 sampling units to calculate the difference. Then, when the character type of the comparison target is “2” or “5”, when the waveform of the reference waveform data is slid by 7 sampling units, the character type of the reference waveform data is “2” or “5”. However, since either the portion W2a and the portion W5a on the one end side or the portion W2b and the portion W5b on the other end side overlap with the waveform of the character waveform data of the character to be processed, the difference amount becomes small. .
Therefore, when discriminating between the character types “2” and “5”, it is difficult to say that sorting with high certainty can be performed even if the amount of difference is compared in the first recognition phase, reflecting the waveform shift. If the second recognition phase is executed again by reflecting the slide amount in the first recognition phase, there is a risk of erroneous recognition.

  In the present embodiment, when the first candidate for path 2 C21 and the second candidate for path 2 C22 selected in the first recognition phase are “2” and “5”, steps SA11, SA12, SA13, Since the first candidate C41 and the second candidate C42 are determined without executing the process of SA14, erroneous recognition or inability to recognize between the character types “2” and “5” is prevented. it can.

  The above-described embodiment is merely an aspect of the present invention, and can be arbitrarily modified and applied within the scope of the present invention.

  For example, in the above-described embodiment, the host computer 70 includes the character recognition unit 80 and executes the recognition of the magnetic ink characters. However, the check reader 1 has the function of the character recognition unit 80. In addition, the check reading device 1 may be configured to execute various processes related to the recognition of the magnetic ink characters alone. In this case, the check reading device 1 functions as a recording medium processing device.

  DESCRIPTION OF SYMBOLS 1 ... Check reading apparatus (recording medium processing apparatus), 4 ... Check (recording medium), 4A ... Magnetic ink character string, 54 ... Magnetic head (magnetic reading part), 70 ... Host computer (recording medium processing apparatus), 71 ... Control unit 73... Host side control unit.

Claims (6)

A reading unit that magnetically reads magnetic ink characters recorded on a recording medium;
A storage unit for storing a reference waveform of a plurality of character types defined in the standard;
A control unit for controlling the reading unit and the storage unit,
The controller is
A difference amount corresponding to a difference between a detection waveform obtained by reading the magnetic ink character by the reading unit and a waveform obtained by sliding a plurality of character type reference waveforms stored in the storage unit by a predetermined distance And a first recognition phase for selecting character type candidates for the magnetic ink characters based on the calculated difference amount;
A difference amount according to a difference between the detected waveform and a waveform obtained by expanding and contracting each of the plurality of character type reference waveforms by a predetermined distance is calculated, and the character of the magnetic ink character is calculated based on the calculated difference amount. Performing a second recognition phase for selecting candidate types,
The difference amount of the character type selected as the candidate in the first recognition phase and the difference amount of the character type selected as the candidate in the second recognition phase are compared, and the magnetic ink character A recording medium processing apparatus for determining a character type candidate. The controller is
In each of the first recognition phase and the second recognition phase, at least a first candidate and a second candidate of the character type of the magnetic ink character are selected in ascending order of the calculated difference amount,
The difference amount of each character type selected as the first candidate and the second candidate in the first recognition phase, and the first candidate and the second candidate in the second recognition phase The first and second candidates for the character type of the magnetic ink character are determined in order of increasing difference. Item 8. A recording medium processing apparatus according to Item 1. The controller is
When the difference amount of the character type selected as the first candidate in the first recognition phase is smaller than the difference amount of the character type selected as the first candidate in the second recognition phase, the plurality of differences Each of the reference waveforms of the character types is slid by the distance slid in the first recognition phase, and the second recognition phase is executed for the second time.
The difference amount of each character type selected as the first candidate and the second candidate in the second recognition phase of the second time, and the first candidate and the second candidate in the first recognition phase The difference amount of each of the selected character types is compared, and the first candidate and the second candidate of the character type of the magnetic ink character are determined in ascending order of the difference amount. The recording medium processing apparatus according to claim 2.   In the second recognition phase for the second time, the font of the magnetic ink character is an E-13B font, the character type selected as the first candidate in the first recognition phase, and the selection as the second candidate The recording medium processing apparatus according to claim 3, wherein the recording medium processing apparatus is executed except when the character types are “2” and “5”. A control method of a recording medium processing apparatus including a reading unit that magnetically reads magnetic ink characters recorded on a recording medium,
Calculates the difference amount according to the difference between the detection waveform obtained by the reading unit reading the magnetic ink characters and the waveform obtained by sliding each of the reference waveforms of a plurality of character types defined by the standard by a predetermined distance And a first recognition phase for selecting character type candidates of the magnetic ink characters based on the calculated difference amount;
A difference amount according to a difference between the detected waveform and a waveform obtained by expanding and contracting each of the plurality of character type reference waveforms by a predetermined distance is calculated, and the character of the magnetic ink character is calculated based on the calculated difference amount. Performing a second recognition phase for selecting candidate types,
The difference amount of the character type selected as the candidate in the first recognition phase and the difference amount of the character type selected as the candidate in the second recognition phase are compared, and the magnetic ink character A control method for a recording medium processing apparatus, wherein a character type candidate is determined. A program executed by a control unit that controls each unit of a recording medium processing apparatus including a reading unit that magnetically reads magnetic ink characters recorded on a recording medium,
The control unit
A difference amount corresponding to a difference between a detection waveform obtained by reading the magnetic ink character and a waveform obtained by sliding a reference waveform of a plurality of character types determined by the standard by a predetermined distance is calculated and calculated. A first recognition phase for selecting character type candidates for the magnetic ink character based on the difference amount;
A difference amount according to a difference between the detected waveform and a waveform obtained by expanding and contracting each of the plurality of character type reference waveforms by a predetermined distance is calculated, and the character of the magnetic ink character is calculated based on the calculated difference amount. Performing a second recognition phase for selecting candidate types,
The difference amount of the character type selected as the candidate in the first recognition phase and the difference amount of the character type selected as the candidate in the second recognition phase are compared, and the magnetic ink character A program which functions as means for determining a character type candidate.

Images