CA1119725A - Magnetic ink character recognition waveform analyzer - Google Patents

Magnetic ink character recognition waveform analyzer

Info

Publication number
CA1119725A
CA1119725A CA000336613A CA336613A CA1119725A CA 1119725 A CA1119725 A CA 1119725A CA 000336613 A CA000336613 A CA 000336613A CA 336613 A CA336613 A CA 336613A CA 1119725 A CA1119725 A CA 1119725A
Authority
CA
Canada
Prior art keywords
character
waveform
video signal
signal
counter
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired
Application number
CA000336613A
Other languages
French (fr)
Inventor
Charles T. Kao
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Recognition Equipment Inc
Original Assignee
Recognition Equipment Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Recognition Equipment Inc filed Critical Recognition Equipment Inc
Application granted granted Critical
Publication of CA1119725A publication Critical patent/CA1119725A/en
Expired legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10Character recognition
    • G06V30/14Image acquisition
    • G06V30/144Image acquisition using a slot moved over the image; using discrete sensing elements at predetermined points; using automatic curve following means
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10Character recognition
    • G06V30/19Recognition using electronic means
    • G06V30/196Recognition using electronic means using sequential comparisons of the image signals with a plurality of references
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10Character recognition
    • G06V30/22Character recognition characterised by the type of writing
    • G06V30/224Character recognition characterised by the type of writing of printed characters having additional code marks or containing code marks
    • G06V30/2253Recognition of characters printed with magnetic ink

Abstract

MICR WAVEFORM ANALYZER
ABSTRACT OF THE INVENTION
A MICR waveform analyzer for reading E-13B magnetic ink characters including a phase lock loop to locate the character, a Normalizer to compress data and equalize ink strength dynamic range, and a waveform amplitude analyzer to identify the magnetically read character is described.

Description

~1~9~25 MICR WAVEFORM ANALYZER
FIELD OF INVENTION
This invention relates to automated Magnetic Ink Character Recognition (MICR) for E-13B font printed to the specification of the American Bankers Association, and more particularly to a low cost magnetic ink character recognition (MICR) system employing a phase lock loop to locate the character horizontal location and to analyze the waveform from a single channel, single gap read head.
DESCRIPTION OF THE PRIOR ART
High speed MICR readers are well known and have been manufactured by several companies to process checks. Once such system is manufactured by Recognition Equipment Incorporated and uses an AC modulated write signal and a multi-element read head. Such a system is too expensive for low speed, low volume applications.
SUMMARY OF THE INVENTION
In accordance with an aspect of the invention there is provided a waveform analyzer for readlng magnetic ink characters wherein an electrisal video signal is produced as the characters pass a pickup head, said video signal containing character data and possibly error data signals, comprising: means for integrating the video signal representative of an entire magnetic ink character to produce a character waveform; processing means for digitizing and then normalizing the digitized character `~
waveform to normalize character ink strength variations;
and sampling means for locating and selecting samples of the digitized character waveform as distinct from error data signals to present to a recognition system.

~ .
.~ ' :.
~ ~ .
.. .~ , ~ :

- 2a -The present invention was developed for a low speed compact document transport. A single write head magnetizes the ferrous material in the printed characters and a read head adjacent the write head senses the magnetic field of the magnetized characters. The system includes both a phase lock loop and an amplitude analyzer for producing signals from which a character decision is made.
Both horizontal and vertical data compression techniques are applied to reduce the time and hardware requirements for wavefor-m analyzing logie. A multi-weighted charaeter waveform ROM is then used for final analyzing the compressed waveform for eharaeter reeognition.

~3 ~ The foregoing and other features and technical advances of the invention will become apparent from the following and more detailed description of a preferred embodiment of the invention as illustrated in the accompanying dr awi ngs.
5 ¦ BRIEF DESCRIPTION OF THE DRAWINGS
Figure 1 is a functional representation of the transport in which the MICR
I reader is a part;
Figure 2 illustrates an E13-B '0" and the magnetic waveform representation thereof; and t0 Figure 3 is a block diagral-n of the MICR waveform analyzer.
Figur e 4 illustrates the block diagram of Ph~se Lock Loop.
Figure 5 is a block diagram of the Maximum Sample Detector and the Amplitude Norrnalizer.
F:i~ure 6 is a block diagram of Mismatch Accurnulator.
Figure 1 is a representation of the functional portion of a transport in which the magnetic sensing head is installed.
As a document is dropped into the ~eeder, it is sensed by an item presence detector (IPD-A) at the bottom of the feeder throat. The docurnent is deskewe and advanced forward to the transport belts.
Acting on a timed signal from the feeder IPD, the transport drive motor i initiated and the belts move in a clockwise or left to right mode. The document i pinched between the jam release belt and the pinch roller as the belts accelerat~
up to a constallt speed of 25 inches per second (IP.~).
Just prior to the read station, the jam release belt engages the drive bel 25 ~ a d the leading edge ot the documellt is sensed by a second IPD (IPD-B). This I

s~rrts a timing clock which tells the down range IPD (IPD-C) when to expect the document.
E-13B coded documents are read at the MICR station situated just below the R/V optical housing. The head assembly consists of two separate heads attached 5 to each other. The write head has a single .005" gap, û.6" in height. Its internal DC resistance may be for example, approximately 110 ohm and energized with a current of 45 ma when connected to the ~5v supply.
The read head has a single .003" gap, 0.6" in height. The head assembly is mounted in the transport so that the lower end o~ the gap aligns with the bottom10 of the paper path.

As the document enters the MICR station, the write head magnetizes the ferrous material in the printed characters. The adjacent read head then senses or reads this magnetic field and sends the characteristics signal to the MICR logic board. Since the strength of the magnetic field drops off significantly with 15 distance between the character and the read head, it is imperative to maintain proper document positioning. To maintain this positioning there are two spring loaded plastic platens built into the jam release directly opposite the two heads.
These platens work independently and provide the spring force required to keep a good document in position as well as flattening a previously creased document.
20 Set-up of the MICR piatens is a very critical balance of proper spring loading and minimum document drag which, when not properly adjusted, can ca~se skew an speed variations.
Characters of the E-13E~ font, printed with magnetic ink, are DC magnetize when they pass the write head gap. As a character passes the read head ~ap, a 25~ voltage i induced for each chanee in the arnount ot mdgnetic flux. Assumin¦

' ~

' ' : '' ' '~,. :~ .

unior~ll ink strength witll n a character, the flux changes will be due to character features. f~n increase in the arnount of ink, such as the leading edge of a vertical stroke, results in a signal of one polarity, while a decrease in ink results in a signal of the opposite polarity. Relative signal amplitude is a fucntion of the amount o~
5 flux density change.
Figure 2 sho~vs the E-13B character "0" and the corresponding waveform. It can be seen tilat the read head video signal is a differentiation of the character'~
rnagnetic intensity. By integrating this signal, a "character waveform" is developed which indicates the instantaneous arnount of ink passing the read head.
10 It is this waveform that is analyzed and recogni~ed by the decision logic o~ the MICR system.
All feature changes of the highly stylized E-13B characters occur at .013"
intervals, or rnultiples of this interval. The horizolltal location logic of the MICR

system generates a "character window" representing eight such intervals, and 15 adjusts systern timing such that character feature changes are aligned within these intervals.
The character w~vefor m is amplitude norrnalized to compensate for variations in ink streng-th from character to character. It is tl-en analyzed by comparinK its norlnalized arnplitude within cac1- of the eight hloclcs with expected
2~ values representing each of the fourteen characters stored in a r ead-only memory (R(:1M). f\t the end of the process, an output character code is generated corresponding to the best-rnatched ROM pattern. The degree of correlation must be wi thin a certain lil-nit and no other ROM pattern shall have been equally mdtched, or a reject code is output. The MICR Data Available signal is receivec 25 ¦ by he transport c~ntroller, which then tccept- the output che.racter code tn~

- ~1972~;

r~sets the Data Available Flag. If a measured amount of time passes without character activity, a space code is output to the transport controller.
Figure 3 is a block diagram of the MICR waveform analyzer. As previously discussed, the write head DC magnetizes the E-13B magnetic ink imprinted 5 characters and the adjacent read head senses the differentiation of the magnetic field. The video signal is connected to the recognition logic circuit which is divided into an analog section and a digital section. `
The analog section of the logic consists of the following functional blocks:

Preamplifier, 60 Hz Notch Filter, Noise Dead Band Circuit, Full wave Rectifier, 10 Peak Detector, Dynamic Threshold, Integrator and 8-Bit Analog/[)igital (A/D) converter.
The Preamplifier is a two stage circuit with a total gain of 4,800. The signal from the MICR Read head results from variations in the flux density of the magnetized ink, i.e., the amount of ink passing the MICR Read head gap.
The 60 Hz Notch ~ilter sharply attentuates the power~line component of the character signal. A long time constant integrator monitors the output of the filter and functions to provide a correction voltage to be returned to the Preamplifier, compensating for offsets and keeping the signal centered about zero.

The Noise Dead Band Circuit removes the first 40 millivolts of signal above 2a and beJow zero in order to eliminate any baseline noise.
The Full Wave Rect;fier inverts the negative portion of the signal, producing signal peaks of positive polarity.
The Peak Detector utilizes a differentiator and Q zero crossing detector to produce a digital signal, indicating the tirne when character feature changes 25 occur. This digital signal must meet a minimum width requirement to be passed ¦ dnd is ter ed 'Pl,AK'.

~-- The l~yndmic Thl-eshold circuit provides a Iurther qualifying term for the digitized edge signal. The threshold voltage is delived from a percentage of the rectiIied signal, and stored by a capacitor which has a controlled discharge rate.
A minimurn threshold voltage is provided by a diode 'OR' circult.
The amplified, filterecl, noise-reduced signal from the Dead Band Circuit is also applied to an Integrator. The output Irom the Integrator is a representation of the amount of magnetic ink passing the MICR Read head gap. The shape will be according to chilracter features, while the amplitude wilJ be accordin~ to the magnetic strength of the ink.
This logic changes the Int.grator's charac-ter wavcform to a digitized form for furtl)er processing and recogr1itior1 by the MICI~ logic. The conversion rate is at the MICR system sample rate of 32 rnicroseconds.
The Digital section of the logic functions to perform hori~ontal location of character data, based upon the time relationship of character peaks. The Digital 15 section consists of the following logic: Find Character Flip Flop, Phase Lock Loop (PLL), Delay Time Counter, False Start Coun-ter, Block Counter ancf Space I:)etector. The first peak of a character set, the Find Character Flip Flop, initializes the digital PLL, enables three timers and arrns the Space Detector. The three tirners are Peal< Interval Timer, Delay Time Counter (12~ Sample Delay), 20 and False ~tart Timer.

The leading and trailing edge of character strokes of the E-13B font are designed to occur at 13 mil intervals or multiples of 13 mils. The MICR logic creates a character windo~l~, equal in time to eight (8) such intervals. These are r denoted as blocks. /~t a Transport spced of 24 Inches Per Second (IPS), one blocl<
2~ is 520 microseconds. Eacl) block is subclivided into sixteen (16) parts or samples.

~97ZS

~ample Clock (SCLK) is the main System Clock and is 32.5 microseconds, derived from the Transport mioprocessor 1.97 MHz clock (PHE).
The Find Character Flip Flop is reset by the signal Character Start, (CHARSTR), allowing the logic to search for another character.
The function of the PLL logic is to synchronize a sixteen (16) step counter with character peaks, such that the peaks will coincide with the counter roll over frorn count IS to zero. The counter operates at the SCLK rate. The first peak of a character sets the counter to a count of one. Since character peaks occur at block intervals or multiples thereof, each subse~uent peak should coincide with the 10 counter rollover to zero. A ROM with correction factors stored for each of the counter states provides a load value for the counter at the time a character peak occurs and can advance or retard the count two steps. For example, assume that a character peak occurs and that the counter is at a count of 3. The next clock will load the ROM correction value 2, which is a correction of two Counts since the 15 next count would have been 4.

Another PLL correction method is the Peak Interval Timer, a sixteen (16) step counter that is loaded with the value 13 when a character peak occurs. If the counter is between a count of 12 and 15 when the next character peak occurs, indicating that the peaks were at block intervals, the signal PLLRST is generated 20 which forces the PLL counter to a count of one due to the validity of the peaks.

The Delay Time Counter is a 128 Sample Delay that is initialized and enabled by the Find Character Flip Flop. Character waveform data is delayed by 12~ sample clocks, one character period on the MICR logic. This one character delya allows the look-ahead capability for the horizontal location logic and also Z5 ~ for alnpl tude normalization by the MICR.

, ~:

: ,, " .: . ":::
- : : :......... , .: , ~ : -:: - , - ' . : ~ .' .' ', ,-:, .
: .. . . :. .. . ..

7;~i - -- The Delay l irne Counter rneasures this period and signifies the emer~ence of character data from its delay line. The output of the counter then vaits for the next PLCENl ER from the PLL ROM, which is a decode of the PLL Counter states 7, 8 or g, indicating the center of a block. lhis wilJ produce an important timing 5 signal, Character Start (CHARSTR). At this time the state of the PLL Counter is loaded into a slave counter to allow the PLL to operate on the next character. The slave counter is free running at the sample rate and therefore rolls over at block intervals. A cour,t of 8 indicates the center of a block, to optimum time to ar,alyze the character waveform. This tirning information is output to the MICR
10 logic by the signal IBLACK'.
The False Start Tirner functions to meas lre the interval between character peaks. If none occur within a specified period of tirne, the assurnption is made that the Find Character Flip F:lop was set by something other than a valid character and the flip flop will be reset.
The Block Counter is initialized by the signal 'CHARSTR' and allowed to increment when the slave PLL counter indicates the end of a block ~ime period.
The 3-bit block count is used by the MICR logic dur ing waveform analysis.
If no peaks occur within five (5) block periods, or 62.5 percent of a character time period, the Space Detector si,snifies a 'space'. An 8 block delay places the 20 space inclication in a time frame with the delayed character data.
The Space Dctector can only gellerate one consecutive space and then only if a character peal< has been previously found.
A character period is drfined as being 128 samples initiated by the first peak (first edge) of a chal-acter. This period is subsidividecl into 8 blocks of 16 samples, 25 each block representin~ one 13 mil interval of paper travel. Character features of the E-13B font occur at 13 rnil intervals or rnultiples thereot.

~~- The MIC~ logic performs hori~.on~al location of character data resulting in control siynals tn;lt are time related to the character. These control signals are used by ~he MICE~ logic for synchronizing the waveform analysis apparatus vith the character wa~eforrn.
The MICI~ logic has five functions. These are C`lock Divider 1~8 Scan Delay Line Character Wavcform Amplitude Norrnali;~.er Character Waveform Analyzer and Outpu~ Interface to the Transport Controller.
The Clock Divider divides the 1.~7 MHz clock ~PHE) of the Transport Controller down to the three Sys tern Clocks which are I microsecond clock 10 (CLKlM) 2 microsecond clock (ADCLK) used by the A/D Converter on MICR
logic and 32.5 rnicrosecond Sample Clock (SCLK).
The 8-bit word serially describing the character waYeshape is shifted through an 8-bit parallel 128-bit seria! delay line. The shift rate is at the sample rate ~SCLK). The one-character delay allows look-ahead capability for the 15 horizontal location logic and for amplitude nonnali~ation by the MICR logic.
Characters are recognized by their waveform which is a function of character ~eature but the waveform arnplitude is a function of the ir k signal strength.
The Amplitude Normalizer makes all chclracters the same size by a rationing 20 techIlique. The larges t sample of a character is found during the look-ahead tirne by the Maximum SarIlple Detector. At the end of the look-ahead period the maxiIrlurrl sample value is trar)sferred to the Maximum Data Buffer.
Combinational lo~ic operating on the maximum sample value steers Multiplexer to output the four Most Significant Bits (MS13) of the maximurn sample value anc 25 Multiplexers to output the c:orresponding four (4) bits of character data emergin~
frorn the delay line.

--- The four Maximum Sample bits and ~he correspondillg four data salnples bits are presented to the Norrnalizer ROM which perforrns the arithmetic operation:

S (Data Same~
MS (Max. Sample~ 7 S This results in a character waveforrn described by a succession of 3-bi~ words and where the tallest feature o~ the waveform will have the value 7.
Character recognition is accomplished by findin~ a correlation between the normalized waveshape and one of the defined character waveforrn patterns stored in ROM. During each of the eight (8) blocks that divide a character time, the 10 amplitude value of the incoming waveforrn addresses a misrnatch value, sequentially for all tourteerI charac-tcr patterns and adds it to the contents of the a~ccumulatol register corresporlding to the character pattern nurnber. The fourteen, 4-bit Inismatch accurnulators are contained by a 16 x 4 bit memory l/Cwhich is addressed by the Character Identity Counter.
As the misrnatches are accumulated during the last block, a register keeps track of the Char~cter IderItity Code which has the fe~est total misrnatches. Ifthe mismatch value was less than 12, and if no other character code had the samenumber of rnisrnatches, the decision character nurnber is converted by a ROM t ASCII and output to the Transport Controller. If a space has been detected by the 20 Horizontal Location Iogic, the ROM outputs the ASCII co<ie for space.
At the end of a character decision cycle, or if a space has been detected, a hancishake signal MICi~ Data Available (MICRDAV) is ~enerated. The Transpor Controller responds by reading the ASCII output and generating the Reset MICR
Data Available (RMlCRr)AV) signal.
~5 Fi~ure 4 illustrates a block dia~ram of di~ital phase lock loop. The purpos of the di ital phase lock loop, or simpIy called "PLL", is to phase lock on the peak~
3~

¦ of-a charactel- video signal so that the p~ Ise of e.lch sample relatir,g ~o the peaks can be identi~ied. As previously discussed, the A/l) convel-ter sarrlples -tne ¦ character wavefonn which is the integrated waveforrn of video si~,nal 16 tirnes per 13 mil of documeot moving tirne. In ttlis systern the sampling clock (SCLK) rateis 0.0131(16x25) = 32.5 microsecond. Since E-13B character fonts are designed for 13 mil spacing for each vertical stroke, all the peaks of an ideally printed character should be separated by a rnultiple of 16 s~mpling clocks array becauseeach peak irnplies the ed~e of a vertical stroke. But a large percenta~e of E-13a characters printed today cannot meet tha$ standard. There are 16 x 8 = 128 scanssampled for each integrated character Y~aveform as snown in Figure 2. Among the 128 scarls, only 8 scans are really representative to the ch~racter feature. These 8 sc.lrls must be sel~cted accurately between peaks. From experience, the stroke width o~ E-13B charac-ters can vary from 8 to l5 mil. In other words, the distance count between two adjacent peaks can vary frorn 10 to 19 counts instead of 16 co~nts due to printin~ quality control problerns. I
In some inst ~nces the first peal< of each chdl-acter is used as a reference for¦
sarrlpling. It ha~s been found that a lot of characters have a mislocated first peak and thus can mislead the sampling schcrme. One solution is to take all the peal<s of a character into cor,sideratiorl and let the rnajority of the peal<s dccide the phase.
This concept is acl~ie~cd by using the Pl L shown in ~igure 4. The center part of the PLL consists of a l'LL Master Counter and a PLL Phase Correction ROM. The other circuit is used to qualify the incomin~ peak and initially reset the PLL
(PLLRST). The PLI. Master Counter may be a 4-bit binary counter such as Texas Instruments SN 74163. The PLL Ph~se Correction ROM may be, for example, a 32 x 4 ROM. All timers and counters are driven by the sampling clock SCLK. The l~trizontal Location Logic detects the first peak of a character and sends it ~o PLI Logic. FIRST PEAK sets PLLE~ST high which causes the PLL Phase Cor rection ROM outputs (4 lines) to be 1. It also s~arts a Timer B to enable the PEAK to preload the PLL Master Counter w;th l's. The PLL Phase Correction 5 ROM is programmed to correct the PLL up to +2 counts by the following peaks.
For example, if the next peak occurs at PLL count 13, the PLL ROM output will be 0. If the next peak occurs at PLL Count 3, the ROM output will be 2. The PLL
Master Counter is then loaded with this new nurnber by each qualified peak and continues counting. Therefore, if the first peak is mislocated by a count of 6, the 10 PLL lo~ic will correct the phase after 3 consecutive upcoming correct peaks.
Timer B will time out an averap,e character time to block out any false peaks.
Timer A, which is also called Peak Interval Timer, is sta!ted for every peak and enables a window for resetting the phase Jock loop whenever it counts to a multiple of 16 ~ I SCLK pulses. In other words, ii the next incoming peak is 15 spaced at a multiple of 16 counts to the preceding one, the PLL will be immediately preset to I becallse a perfectly 13 mil spaced vertical stroke has been located. Thus aEter all peaks of a character pass through PLL, the P~L
Master Counter will lock on the "Average" peal<s (tE a character. A character start (CHARSTR) signal Irom hori~ontal loc:ation logic will load the PLL Master 2U Counter phase information into a PLL Slave Counter which is also a 4-bit binary coun ter. The PLL Master Counter then is ready to phase lock on the next character and let the Slave Counter continue counting to provide phase reference to block counter for character decision logic. The character decision logic will only analyze the 8 digitized integrated video signal with a phase count of 8 among 25 S ose 12~ scans. Th~ls a data base reduction factor of 16 is achieved becaus;

r~

. ~4 128-= 16. These 8 scarls represent the relative vertical stroke height. The rest of the samples are in a transition zone and thus useless for character recognition purposes. This PLL system results in significant savings on tirne and hardware requiremen-ts for the decision logic and also improves the read rate on poorly 5 printed characters.
Figure 5 shows the Maximum Sample Detector and Amplitude Normalizer Block Diagram.
One major problem in reading E-13E~ characters is that the magnetic ink . strength can vary from 30% of a norrnal ink strength to 300% of that strength.10 This, effectively, is a I to 10 dynamic range. Due to the fact that no signal strength reference can be pr ovided before the MICR Read head senses the individual character, it is impossible to take care of the I to 10 ink dynamic range problem by using the conventional AGC (Autornatic Gain Controi) methods. It is also econornically inefficient to use 8 bits to represent the character waveform 15 amplitude inforlnation for final waveform analysis. The Amplitude Normalizer will solve both problems by dividing the maximum sample value into the digitized data and then code the result in 3 bits.
The Maximum Sample Oetector is comprised of an 8-bit Amplitude Comparator, a Register and a e,uffer. At thc beginning of a c haracter, the 20 ¦ Horizontal Location LoE~ic resets the 8-bit Maximurn Value Register (MV~) to 0.
¦ TlIe output is connected to the 8-bit comparator whirh comparcs the 8-bi digitized character waveforrn input data with the register output. If the wavcform dat.l is gl ea~er than the output of the re~ister, a LOAD signal i gencrated to load the waveform data into the ~lVR. At the end of a characte 25 ¦ passing t~r~t gh thc Ma,limum Value Detector, MVR should cont.~in the matirnu~

value of that :haracter waveform. The Horiz.ollta! Loc:ation Logic will output aCHARSTR signal to loacl the maximum value into the maxirnum value buffer (MYB) and then the MVR is ready to find the maxirr um value of the next cnaracter. The MBY output is connected to a Decocler circuit to determine ~he rnost significant bit of that character waveforrrl. The maximll!7l value is alsoconnected to a 4 to I multiplexer (MUX 13) where only the 4 rnost significant bits of the maxirnum value is output to the Amplit-lde Normalization ROM. The i3ecoder output also is connected to the identical rnultiplexer (MUX A) where the 128 scan delayed input data is to be multiplexed so that the corresponding 4 most significant bits will be output to the t~orrnalizâtion ROM. The Normalization ROM
is a 256 x 4 ROM. It divides the waveforrn data by the 4 MSB of Maxirnum Value Samples and codcs it into 3 bits. For example if the 4 MSB s of Maximum Value isto be (A)16 â waveform ciata of (A)16 will be coded as (7)~ and a waveform data f (5)16 will be coded as (4)~. By using this technique a further data base compression factor of 8 = 2.6 is achieved and a dynamic range compensation of ¦ I to 16 is perforrnecl. When these 3 normalized waveIorm data are sampled withthe PLI output a total data base compression factor of 4 266% is obtained because (218 x S)/(8 x 3) = 1024/24 = 42.66 The effect of the data base cc)rnpression will be seen in the rnisrnatch acc urrlul;ltor illustrated in Figure 6.
Figure 6 is a block diagra n of a Mismatcil ~ ~cumulator.
A charactel^ waveform ROM with 1024 ?~ 4 bits stores the character waveform rnismatch counts. This ROM has 10 input address lines. The first thre address line inputs are the three normalized waveform bits as discussed in Figur~. The next three address line inputs are the thlce block count bits from th Blocl< Counter which iclentiEy the block number of the character. The last fou 2~

a~dress line inputs are from a Charactcr Identity Counter ~hich is a 4-bit binary counter counting from ()l6 to (D)~6 at I MHz frequency rate. This counter increments from ()16 to (D)16 each time the block counter increments by i. The
4 ROM output bits are programmed such that the mismatch counts are higher if
5 the corresponding normalized waveIorm ~mplitude value is farthcr away from the stored ideal waveform value. For example, for the scan of the f irst block for character 0, the block count will be (~8' the (~haracter Identity Count will be 0, and the stored ideal amplitude code should be (7)8 because the first block of the character 0 is the largest vertical s-troke. Therefore, the mismatch point for 10 address (7)1024 is ~ )n the contrary! if tht first scan amplitude is ~0)g, the mismatch count will be a ma~irnum (~ `- The rnismatch count for a waveforrn amplitude from ()S to (7)8 will decrease accordingly. In other words, the misrnatch count for each individual scan is weighted by 16 steps. These weighted mismatch counts are connected to a 4-bit adder where the mismatch points from 15 the previous blocks are added together and output to a 16 x ~ RAM for stora~e. At the beginning of a character mismatch accurnulation, the RAM output buffer i cleared by Start Decision Controiler (STDC) so that only the first block mismatc counts are stored in address locations ()16 to (13)16. RAM address locations I ¦

and 15 are not used because there are only 14 E-13~ character fonts. After th 20 first block, each tilne the character idel~tity counter increments, the R/W contro logic will force tl~e RAM to the read mode first for the systern to fetch th previously acculnul~tecl mismatch counts and storc them in the buffer. Secondly the R/W control changes to the Write mode for the system to write the newl accurrlul;3ted rni~rnc-tch count into the R~M. After the Block Counter counts t 25 (7)8 .Ind the char/~clcr idelltity counter counts to (D)~6, the 16 x 4 RAM wil L9';7J~

contain the mism;ltcil points of all the 14 pr~defille(i character waveforms for the input character waveform. 1hen the decision logic will cornpare the 14 accumulated misrnatch counts to determine which Qne is the best match. If there is only one least misrnatch count and that accumulated mismatch points is less 5 than 12, then a character decision is made by selecting the E~AM address that contains the least mismatch count as the character identity code.
Since all the waveform analyzing algorithms are based on 16 scans for each of eight 13 mil blocks of an E-13E~ character, this reader can be applied to the transports having speed other than 25 IPS by sirnply changing the Scan Clock Rate.
10 F;or example, if the transport speed is 20 IPS, the Scan Clock rate will be:

20 x IG = 40.625 rnicrose(:orld 11-~e rest of the logics remain the same.
It can be seen thal the original digitized data has 128 x 8 = 1024 bits for each character. Without any data compression techniqueS it would be an enormous task for the mismatch accurnulator to analyze those bits rather than only 24 (8 x 3) bits that resulted from the present invention.

Having describeci a preferred embodimeI-lt of the invention, further embodinnents and modificatior7s will be suggested to those skilled in the art, whicI
embodiments and modificdtior)s are deemed to be within the scope of the inventiorl as definecl in the followil7~ claims.

Claims (6)

The embodiments of the invention in which an exclusive property or privilege is claimed are defined as follows:
1. A waveform analyzer for reading magnetic ink characters wherein an electrical video signal is produced as the characters pass a pickup head, said video signal containing character data and possibly error data signals, comprising: means for integrating the video signal representative of an entire magnetic ink character to produce a character waveform; processing means for digitizing and then normalizing the digitized character waveform to normalize character ink strength variations;
and sampling means for locating and selecting samples of the digitized character waveform as distinct from error data signals to present to a recognition system.
2. The analyzer according to claim 1 wherein said sampling means includes a peak detector and phase lock loop, the phase lock loop locking onto valid peak pairs produced by said peak detector.
3. The sampling means according to claim 1 including horizontal location logic circuits for horizontally locating the character data waveform signals for each character.
4. The analyzer according to claim 1 wherein the means for integrating is reset at the beginning of each character read.
5. The analyzer according to claim 2 wherein the phase lock loop locks onto peaks which are 13 mils apart or multiples of 13 mils.
6. A waveform analyzer for reading magnetic ink characters wherein a video signal is produced as the characters pass a pickup head, said video signal containing character data and possible error data signals, comprising: an integrator circuit for integrating the video signal over a character period to produce a character waveform signal, an analog to digital converter for digitizing said character waveform signal; a time delay dircuit for delaying the waveform signal; a time delay circuit for delaying the integrated and digital character waveform signal for one character period; a normalizer for amplitude normalizing the character wave-form signal; a peak detector for detecting peaks in the video signal; a phase lock loop for locking onto valid pairs of peaks resulting from scanning a character; and a horizontal location circuit for identifying a character horizontally within the video signal, wherein the character waveform is gated by the horizontal location circuit for identifying the character producing the video signal.
CA000336613A 1978-11-13 1979-09-28 Magnetic ink character recognition waveform analyzer Expired CA1119725A (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US05/959,970 US4245211A (en) 1978-11-13 1978-11-13 MICR Waveform analyzer
US959,970 1978-11-13

Publications (1)

Publication Number Publication Date
CA1119725A true CA1119725A (en) 1982-03-09

Family

ID=25502637

Family Applications (1)

Application Number Title Priority Date Filing Date
CA000336613A Expired CA1119725A (en) 1978-11-13 1979-09-28 Magnetic ink character recognition waveform analyzer

Country Status (8)

Country Link
US (1) US4245211A (en)
JP (1) JPS5567871A (en)
CA (1) CA1119725A (en)
DE (1) DE2945519A1 (en)
FR (1) FR2441888A1 (en)
GB (1) GB2036401B (en)
IT (1) IT7950666A0 (en)
SE (1) SE7909325L (en)

Families Citing this family (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS5876978A (en) * 1981-10-30 1983-05-10 フンダサウン・ブラデスコ Manual linear unit for reading 7-code magnetic character printed on document
US4510615A (en) * 1981-12-29 1985-04-09 International Business Machines Corporation Magnetic character reader with double document detection
US4827531A (en) * 1983-04-11 1989-05-02 Magnetic Peripherals Inc. Method and device for reading a document character
US4797938A (en) * 1985-10-15 1989-01-10 International Business Machines Corporation Method of identifying magnetic ink (MICR) characters
US4764976A (en) * 1986-06-26 1988-08-16 Ncr Corporation Document reader module
US4922086A (en) * 1987-10-19 1990-05-01 Banctec, Inc. Write head apparatus for recognizing magnetic characters and method therefor
DE3854115T2 (en) * 1987-11-30 1996-02-29 Aloka Co Ltd Method and arrangement for ultrasonic measurement of the flow velocity.
US5014324A (en) * 1987-12-21 1991-05-07 Ncr Corporation MICR character reader using magnetic peaks to update timing clocks
US5121437A (en) * 1987-12-21 1992-06-09 Ncr Corporation Micr character reader independent of document transport speed
US6334108B1 (en) 1989-05-01 2001-12-25 Catalina Marketing International, Inc. Method and system for selective incentive point-of-sale marketing in response to customer shopping histories
US6684195B1 (en) 1989-05-01 2004-01-27 Catalina Marketing International, Inc. Method and system for selective incentive point-of-sale marketing in response to customer shopping histories
US5621812A (en) * 1989-05-01 1997-04-15 Credit Verification Corporation Method and system for building a database for use with selective incentive marketing in response to customer shopping histories
US5237620A (en) * 1989-05-01 1993-08-17 Credit Verification Corporation Check reader method and system for reading check MICR code
US8700458B2 (en) 1989-05-01 2014-04-15 Catalina Marketing Corporation System, method, and database for processing transactions
US5687322A (en) 1989-05-01 1997-11-11 Credit Verification Corporation Method and system for selective incentive point-of-sale marketing in response to customer shopping histories
US5201010A (en) * 1989-05-01 1993-04-06 Credit Verification Corporation Method and system for building a database and performing marketing based upon prior shopping history
US5305196A (en) * 1989-05-01 1994-04-19 Credit Verification Corporation Check transaction processing, database building and marketing method and system utilizing automatic check reading
US5644723A (en) 1989-05-01 1997-07-01 Credit Verification Corporation Method and system for selective incentive point-of-sale marketing in response to customer shopping histories
GB2237915B (en) * 1989-10-24 1993-09-29 Unisys Corp Character recognition apparatus
US5026974A (en) * 1989-12-27 1991-06-25 Ncr Corporation Method for recognizing the leading edge of a character in E13B font
DE4006426C2 (en) * 1990-03-01 1993-11-18 Siemens Nixdorf Inf Syst Method for evaluating binary information from a magnetic memory card
US6292786B1 (en) 1992-05-19 2001-09-18 Incentech, Inc. Method and system for generating incentives based on substantially real-time product purchase information
FR2715752B1 (en) * 1994-02-03 1996-04-26 Dassault Automatismes Telecomm Method and device for recognizing characters, in particular standardized characters of the E-13B type.
US6327378B1 (en) * 1995-02-15 2001-12-04 Banctec, Inc. Character recognition method
US5712564A (en) * 1995-12-29 1998-01-27 Unisys Corporation Magnetic ink recorder calibration apparatus and method
US6609104B1 (en) 1999-05-26 2003-08-19 Incentech, Inc. Method and system for accumulating marginal discounts and applying an associated incentive
US6993498B1 (en) 1999-07-15 2006-01-31 Midnight Blue Remote Access, Llc Point-of-sale server and method
US8023717B1 (en) * 2007-01-16 2011-09-20 Burroughs Payment Systems, Inc. Method and system for processing backwards documents in a document reader/imager
US8023718B1 (en) * 2007-01-16 2011-09-20 Burroughs Payment Systems, Inc. Method and system for linking front and rear images in a document reader/imager
WO2012142985A1 (en) * 2011-04-19 2012-10-26 Balluff Gmbh Method for operating a distance sensor and device for performing the method
US8636214B2 (en) * 2011-10-25 2014-01-28 International Business Machines Corporation Recognition of encoded information on documents

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
NL257640A (en) * 1959-11-12 1900-01-01
US3302174A (en) * 1963-07-30 1967-01-31 Gen Electric Signal position detection circuit
GB1023810A (en) * 1963-12-30 1966-03-23 Ibm Improvements relating to character recognition apparatus
US3535682A (en) * 1965-12-10 1970-10-20 Lundy Electronics & Syst Inc Waveform recognition system
US3528058A (en) * 1966-05-27 1970-09-08 Ibm Character recognition system
BE755936A (en) * 1969-09-09 1971-02-15 Ibm APPARATUS AND METHOD FOR IDENTIFYING CHARACTERS
US3851309A (en) * 1973-01-11 1974-11-26 Optical Recognition Syst Character recognition apparatus
US3930231A (en) * 1974-06-10 1975-12-30 Xicon Data Entry Corp Method and system for optical character recognition
US4080528A (en) * 1976-04-23 1978-03-21 Recognition Equipment Incorporated Data interlacing system
US4143355A (en) * 1977-08-29 1979-03-06 Signature Guardian Systems, Inc. Character recognition system
US4143356A (en) * 1977-10-31 1979-03-06 Ncr Canada Ltd - Ncr Canada Ltee Character recognition apparatus
US4148010A (en) * 1977-10-31 1979-04-03 Ncr Canada Ltd. - Ncr Canada Ltee Magnetic ink character reader system

Also Published As

Publication number Publication date
JPS5567871A (en) 1980-05-22
SE7909325L (en) 1980-05-14
DE2945519A1 (en) 1980-05-22
US4245211A (en) 1981-01-13
IT7950666A0 (en) 1979-10-24
FR2441888A1 (en) 1980-06-13
GB2036401A (en) 1980-06-25
GB2036401B (en) 1983-07-20

Similar Documents

Publication Publication Date Title
CA1119725A (en) Magnetic ink character recognition waveform analyzer
US4399553A (en) Character reader
US4740675A (en) Digital bar code slot reader with threshold comparison of the differentiated bar code signal
US4797938A (en) Method of identifying magnetic ink (MICR) characters
US4143355A (en) Character recognition system
CA1306542C (en) Speed compensation scheme for reading micr data
US3535682A (en) Waveform recognition system
GB2098768A (en) Card readers
EP0096066B1 (en) Method and system for processing data signals representing an unknown character
GB1118790A (en) Character recognition apparatus
EP0380601A1 (en) Magnetic character reading apparatus.
US4080528A (en) Data interlacing system
JPS5951024B2 (en) magnetic signal detector
US5729621A (en) Method and apparatus for magnetic ink character recognition using a magneto-resistive read head
US4243983A (en) Method and apparatus for detecting the multiple feeding of documents
US3354432A (en) Document reading system
US4058706A (en) MICR data lift system
US3879707A (en) Character recognition system for bar coded characters
CA1059238A (en) Code converter
US3303469A (en) Method and apparatus for reading characters
CA1151301A (en) Magnetic ink character recognition apparatus
GB1014511A (en) Character recognition apparatus
US3212058A (en) Null dependent symbol recognition
EP0451252B1 (en) Ratio detection character recognition apparatus
EP0774728A2 (en) Magnetic ink character detection apparatus and method

Legal Events

Date Code Title Description
MKEX Expiry