Note: Descriptions are shown in the official language in which they were submitted.
2194710
Method and Apparatus for Improved Coin, Bill and Other
Currency Acceptance and Slug or Counterfeit Rejection
This is a division of copending Canadian Patent
Application Serial No. 2,069,875 filed on May 28, 1992,
based on PCT/US91/07548 filed October 9, 1991.
Technical Field
The present invention relates to the
examination of coins, bills or other currency for purposes
such as determining their authenticity and denomination,
and more particularly to methods and apparatus for
achieving a high level of acceptance of valid coins or
currency while simultaneously maintaining a high level of
rejection of nonvalid coins or currency, such as slugs or
counterfeits. While the present invention is applicable
to testing of coins, bills and other currency, for the
sake of simplicity, the exemplary discussion which follows
is primarily in terms of coins. The application of the
present invention to the testing of paper money, banknotes
and other currency will be immediately apparent to one of
ordinary skill in the art.
Backqround Art
It has long been recognized in the field of
coin and currency testing that a balance must be struck
between the conflicting goals of "acceptance" and
"rejection"--perfect acceptance being the ability to
correctly identify and accept all genuine items no matter
their condition, and perfect rejection being the ability
to correctly discriminate and reject all non-genuine
items. When testing under ideal conditions, no difficulty
arises when trying to separate ideal or perfect coins from
slugs or counterfeit coins that have different character-
istics even if those differences are relatively slight.
Data identifying the characteristics of the ideal coins
. 2194710
-
--2--
-- can be stored and compared with data measured
from a coin or slug to be tested. By narrowly
defining coin acceptance criteria, valid coins
that produce data falling within these criteria
can be accepted and slugs that produce data
falling outside these criteria can be rejected.
A well-known method for coin acceptance and
slug rejection is the use of coin acceptance
windows to define criteria for the coin
acceptance. One example of the use of such
windows is described in U.S. Patent Nos.
3,918,564 and 3,918,565, both assigned to the
assignee of the present invention.
Of course, in reality, neither the
test conditions nor the coins to be tested are
ideal. Windows or other tests must be set up
to accept a range of characteristic coin data
for worn or damaged genuine coins, and also to
compensate for environmental conditions such as
extreme heat, extreme cold, humidity and the
like. As the acceptance windows or other coin
testing criteria are widened or loosened, it
becomes more and more likely that a slug or
counterfeit coin will be mistakenly accepted as
genuine. As test criteria are narrowed or
tightened, it becomes more likely that a
genuine coin will be rejected.
U.K. Application Serial No. 89/23456.1
filed Oct. 18, 1989, and assigned to the
assignee of the present invention, is one
response to the real world compromise between
achieving adequately high levels of acceptance
and rejection at the same time. This U.K.
application describes techniques for
establishing non-uniform windows that maintain
a high level of acceptance while achieving a
high level of rejection.
Another prior art approach is found in
the Mars Electronics IntelliTrac~ Series
products. The IntelliTrac Series products
2194710
-
-- 3
operate substantially as described in European Patent
Application EP 0 155 126, which is assigned to the
assignee of the present invention.
SUMMARY OF THE INVENTION
In accordance with one aspect of the invention
there is provided a method of operating a money validation
apparatus for discriminating genuine items of different
types from counterfeit items, comprising: sensing data
characteristic of at least two characteristics of each of
a plurality of genuine items representative of the
universe of items to be validated; converting the sensed
data into a plurality of vectors for each item type;
storing the vectors in a look-up table in memory;
calculating a mean vector for each item type; testing an
item and generating a vector corresponding to said at
least two characteristics for the item; calculating the
difference between the item vector and the mean vector for
an item type; comparing the difference to a first mean
vector tolerance; incrementing an item denomination index,
recalculating the difference and comparing the difference
to a mean vector tolerance for another item type if the
comparison did not fall within the first mean vector
tolerance; searching an item type look-up table if the
difference falls within the corresponding mean vector
tolerance; and accepting the item if its vector is found
in a look-up table, or rejecting the item if its vector is
not found.
In accordance with another aspect of the
invention there is provided a method in an item validation
apparatus having a sensor circuit and a processing and
control circuit, for discriminating genuine items from
counterfeit items, comprising the steps of: sensing data
characteristic of at least two characteristics from a
plurality of genuine items of different item types;
converting the sensed data into a plurality of data points
for each item type; selecting data points to form clusters
of data representing an acceptance criteria for each
genuine item type; storing the clusters; defining a center
data point for each cluster; setting a deviation limit
which is small in comparison to the distance from the
219~710
-
- 3a -
center data point to a cluster boundary data point;
testing an item and generating a data point for the item;
accepting the item as being a particular type if the data
point is within a cluster corresponding to that type; and
modifying the acceptance criteria by incrementing or
decrementing the center data point of a cluster if enough
accepted items of that type had data points within the
deviation limit.
In accordance with yet another aspect of the
invention there is provided an item validation apparatus
for discriminating genuine items from counterfeit items,
comprising: means for sensing data characteristic of at
least two characteristics from a plurality of genuine
items of different item types; means for converting the
sensed data into a plurality of data points for each item
type; means for selecting data points to form clusters of
data representing an acceptance criteria for each genuine
item type; means for storing the clusters; means for
defining a center data point for each cluster; means for
setting a deviation limit which is small in comparison to
the distance from the center data point to a cluster
boundary data point; means for testing an item and
generating a data point for the item; and means for
accepting the item if the data point is within a cluster
and for modifying the acceptance criteria if enough
accepted items of that type had data points within the
deviation limit.
In accordance with yet another aspect of the
invention there is provided a method of operating a money
validation apparatus having at least one sensor circuit
and a processing and control circuit, which utilizes
acceptance criteria corresponding to genuine items of
different types, wherein the acceptance criteria is
comprised of characteristic data having a center point,
comprising: setting a deviation limit which is small in
comparison to the distance from the center point to a
boundary point of the acceptance criteria; measuring a
rest value for each sensor; testing an item by measuring
shift values of the sensors; calculating exponentially
weighted moving averages based on the rest values;
calculating relative values for the item based on the
219~710
,
- 3b -
shift values, the rest values and the exponentially
weighted moving average; generating characteristic data
for the item based on the relative values; accepting the
item as being of a particular type if its characteristic
data is within the acceptance criteria corresponding to
that type; calculating the absolute difference between the
characteristic data of an accepted item and the center
point of the acceptance criteria; adding the difference of
the center point and the characteristic data of the
accepted item to a cumulative sum if the absolute
difference is less than or equal to the deviation limit;
incrementing the center point by a preset amount if the
cumulative sum exceeds a predetermined limit, or
decrementing the center point by a preset amount when the
cumulative sum is less than a predetermined negative
limit; and resetting the cumulative sum.
In accordance with yet another aspect of the
inveniton there is provided a method of operating a money
validation apparatus having a sensor circuit and a
processing and control circuit, comprising: defining a
coordinate system having its origin at an idle operating
point of the money validation apparatus; sensing data
representative of at least two characteristics of each of
a plurality of genuine money items; combining the sensed
characteristic data for each genuine item into vectors
wherein the idle operating point is used as the origin of
each vector; mapping the vectors onto the coordinate
system to form an acceptance cluster; storing the
acceptance cluster; sensing an item inserted into the
validation apparatus and generating data representative of
said at least two characteristics; converting the
generated data for the inserted item into a test vector;
comparing the test vector to the stored acceptance
cluster; and accepting the item as a genuine item if the
test vector matches a vector within the acceptance
cluster.
In accordance with yet another aspect of the
invention there is provided a method for increasing the
level of counterfeit rejection in a money validation
system, wherein the money validation system generates at
least one value corresponding to at least one character-
2194710
-- 4
istic of an inserted item and compares the generated valueto predetermined acceptance criteria values, comprising:
inserting a plurality of known counterfeit items into the
validation apparatus; generating counterfeit values for
each counterfeit item;
subtracting the counterfeit values from the acceptance
criteria values to form an improved acceptance criteria;
and utilizing the improved acceptance criteria to validate
subsequently inserted items.
In accordance with yet another aspect of the
invention there is provided a method of operating a money
validation system having at least two sensors and a
processing and control circuit, comprising: defining an
activity vector originating at the zero electrical
activity point and terminating at an idle operating point
for the system; sensing data with each sensor
representative of at least two characteristics of each of
a plurality of genuine money items; combining the sensed
data into a plurality of acceptance vectors; mapping the
acceptance vectors onto a coordinate system; defining a
mean vector having its origin at the idle operating point
and terminating at the center of the mapped region;
defining a tolerance value for the mean vector; generating
an acceptance region based on the tolerance value and the
mean vector; sensing an inserted item and generating data
representative of said at least two characteristics;
converting the generated data of the inserted item into a
test vector; comparing the test vector to the acceptance
region; and accepting the item as a genuine item if the
test vector falls within the acceptance region.
In accordance with yet another aspect of the
invention there is provided a money validation apparatus,
comprising: at least one sensor circuit which senses at
least two characteristics of money items; means for
defining an idle operating point of the system; means for
converting sensed characteristic data for genuine items
into acceptance vectors having an origin at the idle
operating point; means for mapping the acceptance vectors
onto a coordinate system; means for defining a mean vector
having an origin at the idle operating point and
terminating at the center of the mapped region; means for
' 219~710
- 4a -
defining a tolerance value; means for generating an
acceptance region based on the tolerance value and the
mean vector; means for sensing an inserted item and
generating data representative of said at least twc
characteristics; means for converting the generated data
of the inserted item into a test vector; and means for
comparing the test vector to the acceptance region and for
accepting the item as a genuine item if the test vector
falls within the acceptance region.
The present invention can be applied to a wide
range of electronic tests for measuring one or more
parameters indicative of the acceptability of a coin,
currency or the like. The various aspects of the
invention may be employed separately or in conjunction
depending upon the desired application.
Brief Description of the Drawinqs
The present invention taken in conjunction with
the invention disclosed in copending Canadian Patent
Application Serial No. 2,069,875 filed on May 28, 1992,
based on PCT/US91/07548 filed October 9, 1991, will be
described hereinbelow with the aid of the accompanying
drawings in which:
Fig. 1 is a schematic block diagram of an
embodiment of electronic coin testing apparatus, including
sensors, suitable for use with the invention;
Fig. 2 is a schematic diagram indicating
suitable positions for the sensors of the embodiment of
Fig. 1;
Fig. 3 is a graphical representation of a prior
art coin acceptance window for testing three coin
acceptance criteria;
Fig. 4 is a graphical representation of one
aspect of the present invention, namely improved coin
acceptance criteria using coin acceptance clusters;
Fig. 5 is a flow chart of the operation of the
coin acceptance clusters for the improved definition of
coin aceeptance eriteria of the present invention;
Fig. 6 is a graphical representation of a
typical line distribution curve of certain measured
criteria for a genuine coin;
' 2194710
-
- 4b -
Fig. 7A is a graphical representation of the
line distribution for the genuine coin criteria of Fig. 6
drawn to include a line distribution for the same criteria
of an invalid coin, to illustrate the anti-fraud or anti-
cheat aspect of the present invention;
Fig. 7B is an additional graphical
representation showing substantial overlap for certain
measured criteria of a genuine coin line distribution and
an invalid coin line distribution;
Figs. 7C and 7D are additional graphical
representations showing minimal overlap for certain
measured criteria for certain genuine coin
2194710
line distributions and invalid coin line
distributions;
Fig. 8 is a flow chart of the operation
of the anti-fraud or anti-cheat aspect of the
present invention;
Fig. 9 is a flow chart of the operation
of the aspect of the present invention relating to
minimizing the effects of counterfeit coins and
slugs on the self-adjustment process for the
center of the coin acceptance window;
Fig. 10 is a flow chart of a portion of
the operation of the present invention relating to
relative value computation and conservation of
memory space and minimization of microprocessor
computation time in a microprocessor based coin
validation system; and
Fig. 11 is a graphical representation
concerning that aspect of the present invention
describing the modification of the measured
response in the validation apparatus due to the
presence of large changes to the reference
parameter; Fig. 11 is located on the same sheet
of drawings containing Figs. 6 and 7~.
Detailed DescriPtion
The coin examining apparatus and methods
of this invention may be applied to a wide range
of electronic coin tests for measuring a parameter
indicative of a coin's acceptability and to the
identification and acceptance of any number of
3~ coins from the coin ~ets of many countries. In
particular, the following description concentrates
on the details for setting the acceptance limits
for particular tests for particular coins, but the
application of the invention to other coin tests
and other coins will be clear to those skilled in
the art.
~19471~
--6--
The figures are intended to be
representational and are not drawn to scale.
Throughout this specification, the term ~coin~ is
intended to include genuine coins, tokens,
counterfeit coins, slugs, washers, and any other
item which may be used by persons in an attempt to
use coin-operated devices. Also, th,e disclosed
invention may suitably be applied to validation of
bills and other currency, as well as coins. It
will be appreciated that the present invention is
widely applicable to coin, bill and other currency
testing apparatus generally.
The presently preferred embodiment of the
method and apparatus of this invention is
implemented as a modification of an existing
family of coin validators, the Mars Electronics
IntelliTrac Series. The present invention employs
a revised control program and revised control
data. The IntelliTrac Series operates
substantially as described in European Application
EP 0 155 126.
Fig. 1 shows a block schematic diagram of
a prior art electronic coin testing apparatus 10
suitable for implementing the method and npparatus
of the present invention by making the
modifications described below. The mechanical
portion of the electronic coin testing apparatus
10 is shown in Fig. 2. The electronic coin
testing apparatus 10 includes two principal
~ections: a coin examining and ~ensing circuit 20
including individual ~ensor circuits 21, 22 and
23, and a processing and control circuit 30. The
processing and control circuit 30 includes a
programmed microprocessor 35, an analog to digital
(A/D) converter circuit 40, a signal shaping
- 2194710
circuit 45, a comparator circuit 50, a counter 55,
and NOR-gates 61, 62, 63, 64 and 65.
Each of the sensor circuits 21, 22
includes a two-sided inductive 6ensor 24, 25
having its series-connected coil6 located adjacent
opposing 6idewalls of a coin passageway. As shown
in Fig. 2, sensor 24 is preferably of a large
diameter for testinq coins of wideranging
diameters. Sensor circuit 23 includes an
inductive 6ensor 26 which ic preferably arranged
as shown in Fig. 2.
Sensor circuit 21 is a high-frequency,
low-power oscillator used to test coin parameters,
such as diameter and material. As a coin passes
the sensor 24, the frequency and amplitude of the
output of sensor circuit-21 change as a result of
coin interaction with the sensor 24. This output
is shaped by the shaping circuit 45 and fed to the
comparator circuit 50. When the change in the
amplitude of the signal from shaping circuit 45
exceeds a predetermined amount, the comparator
circuit 50 produces an output on line 36 which is
connected to the interrupt pin of microprocessor
35.
The output from shaping circuit 45 is
also fed to an input of the A/D converter circuit
40 which converts the analog signal at its input
to a digital output. This digital output is
serially fed on line 42 to the microprocessor 35.
The digital output is monitored by microprocessor
35 to detect the effect of a passing coin on the
amplitude of the output of sensor circuit 21. In
conjunction with frequency shift information, the
amplitude information provides the microprocessor
35 with adequate data for particularly reliable
testing of coins of wideranging diameters and
materials using a single sensor 21.
219~710
The output of sensor circuit 21 is also
connected to one input of NOR gate 61 the output
of which is in turn connected to an input of NOR
gate 62. NOR gate 62 is connected as one input of
NOR gate 65 which has its output connected to the
counter 55. Frequency related information for the
sensor circuit 21 is generated by 6electively
connecting the output of sensor circuit 21 through
the NOR gates 61, 62 and 65 to the counter 55.
Frequency information for sensor circuits 22 and
23 is similarly generated by selectively
connecting the output of either sensor circuit 22
or 23 through its respective NOR gate 63 or 64 and
the NOR gate 65 to the counter 55. Sensor circuit
22 is also a high-frequency, low-power oscillator
and it is used to test cDin thickness. Sensor
circuit 23 is a strobe sensor commonly found in
~ending machines. As shown in Fig. 2, the sensor
26 is located after an accept gate 71. The output
of sensor circuit 23 is used to control such
functions as the granting of credit, to detect
coin jams and to prevent customer fraud by methods
such as lowering an acceptable coin into the
machine with a string.
The microprocessor 35 controls the
selective connection of the outputs from the
sensor circuits 21, 22 and 23 to counter 55 as
described below. The freguency of the oscillation
at the output of the sensor circuits 21, 22 and 23
is sampled by counting the threshold level
crossings of the output signal occurring in a
predetermined sample time. The counting is done
by the counter circuit 55 and the length of the
predetermined sample time is controlled by the
microprocessor 35. One input of each of the NOR
gates 62, 63 and 64 i5 connected to the output of
its associated sensor circuit 21, 22 and 23. The
2194710
g
output of sensor 21 is connected through the NOR
gate 61 which is connected as an inverter
amplifier. The other input of each of the NOR
gates 62, 63 and 64 is connected to its respective
control line 37, 38 and 39 from the microprocessor
35. The signals on the control lines 37, 38 and
39 control when each of the sensor circuits 21, 22
and 23 is interrogated or ~ampled, or in other
words, when the outputs of the sensor circuits 21,
22 and 23 will be fed to the counter 55. For
example, if microprocessor 35 produces a high
(logic ~1~) signal on lines 38 and 39 and a low
signal (logic ~0~) on line 37, sensor circuit 21
is interrogated, and each time the output of the
NOR gate 61 goes low, the NOR gate 62 produces a
high output which is fed through NOR gate 65 to
the counting input of counter 55. Counter 55
produces an output count signal and this output of
counter 55 is connected by line 57 to the
microprocessor 35. Microprocessor 35 determines
whether the output count signal from the counter
55 and the digital amplitude information from A/D
converter circuit 40 are indicative of a coin of
acceptable diameter and material by determining
whether the outputs of counter 55 and A/D
converter circuit 40 or a value or values computed
therefrom are within ~tored acceptance limits.
When sensor circuit 22 is interrogated,
microprocessor 35 determines whether the counter
output is indicative of a coin of acceptable
thickness. Finally, when sensor circuit 23 i6
interrogated, microprocessor 35 determines whether
the counter output is indicative of coin presence
or absence. When both the diameter and thickness
tests are satisfied, a high degree of accuracy in
discrimination between genuine and false coins is
achieved.
-' 219-4710
--10--
A person skilled in the art would readily
be able to implement in any number of ways the
specific logic circuits for the block diagram set
forth in Fig. 1 and described above. Preferably,
the circuitry ~uitable for the embodiment of Fig.
1 is incorporated in an application ~pecific
integrated circuit (ASIC) of the type presently
part of the TA100 stand alone Acceptor ~old by
Mars Electronics, a subsidiary of the assignee of
the present invention. Another ~pecific way to
implement the circuitry of Fig. 1 is shown and
described in European Patent Application EP 0 155
126, referenced above, which is assigned to the
assignee of the present invention.
The methods of tne present invention will
now be described in the context of setting coin
acceptance limits based upon the frequency
information from sensor circuit 21. As a coin
approaches and passes inductive sensor 24, the
frequency of its associated oscillator varies from
the no coin idling frequency, fO and the output of
sensor circuit 21 varies accordingly. Also, the
amplitude of the envelope of this output signal
varies. Microprocessor 35 then computes a maximum
change in frequency f, where f equals the
maximum absolute difference between the freguency
measured during coin passage and the idling
frequency. The f value is also sometimes
referred to as the shift value. f=max(f~.~ured ~
fO). A dimensionless quantity F= f/fO is then
computed and compared with ~tored Acceptance
limits to see if this value of F for the coin
being tested lies within the acceptability range
for a valid coin. The F value is also sometimes
referred to as the relative value.
219471Q
As background to such measurements and
computations, see U.S. Patent No. 3,918,564
assigned to the assignee of the present
application. As discussed in that patent, this
type of measurement technique also applies to
parameters of a sensor output signal other than
frequency, for example, amplitude. Similarly,
while the present invention is specifically
applied to the setting of coin acceptance limits
for particular sensors providing amplitude and
frequency outputs, it applies in general to the
setting of coin acceptance limits derived from a
statistical function for a number of previously
accepted coins of the parameter or parameters
measured by any sensor.
In the prior art, if the coin was
determined to be acceptable, the ~ value was
stored and added to the store of information used
by microprocessor 35 for computing new acceptance
limits. For example, a running average of stored
F values was computed for a predetermined number
of previously accepted coins and the acceptance
limits were established as the running average
plus or minus a stored constant or a stored
percentage of the running average. Preferably,
both wide and narrow acceptance limits were stored
in the microprocessor 35. Alternatively these
limits could be stored in RAM or ROM. In the
embodiment shown, whether the new acceptance
limits were set to wide or narrow values was
controlled by external information supplied to the
microprocessor through its data communication bus.
Alternatively, a ~election ~witch connected to one
input of the microprocessor 35 could be used. In
the latter arrangement, microprocessor 35 tested
for the state of the ~witch, that is, whether it
was open or closed and adjusted the limits
- 2194710
depending on the state of the switch. The narrow
range achieved very good protection against the
acceptance of slugs; however, the tradeoff was
that acceptable coins which were worn or damaged
were likely to be rejected. The ability to select
between wide and narrow acceptance limits allowed
the owner of the apparatus to adjust the
acceptance limits in accordance with his
operational experience. As described further
below in conjunction with a discussion of Figs. 4
and 5, the present invention has an improved and
more sophisticated approach to the
acceptance/rejection tradeoff.
Other ports of the microprocessor 35 are
connected to a relay control circuit 70 for
controlling the gate 71 shown in Fig. 2, a clock
75, a power supply circuit 80, interface lines 81,
82, 83 and 84, and debug line 85. The
microprocessor 35 can be readily programmed to
control relay circuit 70 which operates a gate to
separate acceptable from unacceptable coins or
perform other coin routing tasks. The particular
details of controlling such a gate do not form a
part of the present invention.
The clock 75 and power supply 80 supply
clock and power inputs required by the
microprocessor 35. The interface lines 81, 82, 83
and 84 provide a means for connecting the
electronic coin testing apparatus 10 to other
apparatus or circuitry which may be included in a
coin operated vending mechanicm which includes the
electronic coin testing apparatus 10. The details
of such further apparatus and the connection
thereto do not form part of the present invention.
3~ Debug line 85 provides a test connection for
monitoring operation and debugging purposes.
2194710
Fig. 2 illustrates the mechanical portion of
the coin testing apparatus 10 and one way in which
sensors 24, 25 and 26 may be suitably positioned
adjacent a coin passageway defined by two spaced
side walls 32, 38 and a coin track 33, 33a. The
coin handling apparatus includes a conventional
coin receiving cup 3~, two spaced sidewalls 32 and
38, connected by a conventional hinge and spring
assembly 34, and coin track 33, 33a. The coin
track 33, 33a and sidewalls 32, 38 form a coin
passageway from the coin entry cup 31 past the
coin sensors 24, 25. Fig. 2 also shows the sensor
26 located after the gate 71, which in Fig. 2 is
shown for separating acceptable from unacceptable
1~ coins.
It should be understood that other
positioning of sensors may be advantageous, that
other coin passageway arrangements are
contemplated and that additional sensors for other
coin tests may be used.
The various aspects of the present
invention will now be described.
COIN CLUSTERS - IMPRO~ED DEFINITION OF COIN
ACCEPTANCE CRITERIA
When validating coins, two or more
independent tests on a coin are typically
performed, and the coin is deemed authentic or of
a specific denomination or type only if all the
test results equal or come close to the results
expected for a coin of that denomination. For
example, the influence of a coin on the fields
generated by two or more sensors can be compared
to measurements known for authentic coins
corresponding to thickness, diameter and material
content. This is represented graphically in Fig.
3, in which each of the three orthogonal axes P1,
2I94710
P2 and P3 represent three independent coin
characteristics to be measured. For a coin of
type A, the measurement of characteristic P, is
expected to fall within a range (or window) W~,
which lies within the upper and lower limits U~
znd L~. Similarly, the characteristics or
properties P2 and P, of the coin are expected to
lie within the ranges W~ and W~, respectively. If
all three measurements lie within these ranges or
windows, the coin is deemed to be an acceptable
coin of type A. Under these circumstances, the
measurements for acceptable coins will lie within
the three-dimensional acceptance region designated
as RA in Fig. 3. A coin validator arranged to
validate more than one type of coin would have
different acceptance regions R~, ~, etc., for
different coin types B, C, etc.
As discussed further in connection with
Figs. 7B, 7C and 7D below, counterfeit coins or
slugs may have sensor measurement distributions
which fall within or overlap those for a genuine
coin. For example, a slug may have
characteristics which fall within region RA Of Fig.
3 because the slug exhibits properties which
overlap those of a valid coin of that
denomination. Although tighter limits on the
acceptance region RA may screen out such slugs,
such a restriction will also increase the
rejection of genuine coins.
The present invention, in order to
provide improved coin acceptance criteria which
are better defined, takes into account two
observations concerning the vast majority of
counterfeit coins. First, counterfeit coins do
not produce the same distribution of sensor
responses as do valid coins. Second, most
counterfeit coins falling within an acceptance
219471~
region, such as region RA ~hown in Fig. 3, were on
the periphery of the acceptance region and
exhibited very little overlap with the values
found for genuine coins. See, e.g., the
histograms designated as Figs. 7B, 7C and 7D,
which show the overlap for three ceparate coin
tests, between a large cet of empirically tested
United States twenty-five cents coins and a large
set of empirically tested foreign coins. The coin
measurement criteria are represented on the
abscissa of each histogram; the percentage of
tested coins having specified measurement criteria
may be determined from the ordinate of each
histogram. It is noted that there is very little
overlap on Figs. 7C and 7D.
Looking at Fig. 7B, it is seen that the
data for the twenty-five cents coins significantly
overlaps the data for the foreign coin for the
material test illustrated in this figure. No
adjustment of this test criteria can practically
reduce the acceptance of the foreign coin without
also rejecting the vast majority of genuine
twenty-five cents coins. On the other hand, for
the thickness and diameter tests of Figs. 7C and
7D, the areas of overlap are much smaller and
individual adjustments of the acceptance criteria
could be made that would ~ignificantly increase
the rejection of the foreign coin while ctill
accepting a large number of genuine twenty-five
cents coins. In its presently preferred
embodiment, the present invention takes a more
subtle approach than just described in that it
recognizes that coin acceptance criteria such as
material, thickness, diameter and the like are
generally not independent of one another. For
example, a slug which has coin thickness which
overlaps that typical of a genuine coin may be
219~710
-16-
much more statistically likely to have a coin
diameter that also overlaps tbat typical of a
genuine coin. The present invention takes into
account such interrelationships as further
described below.
~or a particular denomination coin,
sensor response data from several different sets
of 6ensor6 and for a large population of genuine
coins was collected. one ~uch distribution is
illustrated in Figs. 7B, 7C and 7D, which show the
peak change in sensor response for a large number
of representative twenty-five cents coins
submitted through a coin mechanism in a normal
manner. All this data was then mapped into a
three dimensional coordinate system to form a
~cluster~ of acceptance ~alues. Likewise, data
was collected and mapped for known counterfeit
coins or slugs. The data for one such foreign
coin often used as a slug is also illustrated in
Figs. 7B, 7C and 7D. This data was similarly
mapped into a three dimensional coordinate system,
and certain points were ruled out as acceptance
points.
Fig. 4 represents a mapping of coin
sensor values in a three dimensional coordinate
system. The point f10, f20, AOat the intersection
of the Xl ~ X2 ~ X3 coordinate axes (~x coordinate
system~) represents the point of zero electrical
activity for the sensing circuits, while the point
f10, f20, Ao represents an idle operating point for
the system. The point f10, f20, Ao is an arbitrary
starting point shown for exemplary purposes only
and can be changed in response to environmental
factors or the like. A vector C0 terminates at
this steady Gtate idle operating point, and is
utilized to perform a mapping from the x
coordinate system, or the zero electrical activity
2194710
system, to an x' coordinate system, the idle
sensor response coordinate system.
The regions RA~ RB~ and F~ represent
linear acceptance regions such as shown in Fig. 3
S for use in detecting genuine coins of three
differing denominations, while the regions C~, C~
and Cc represent cluster regions for these same
three genuine coins. Regions S~ and S8 are
examples of counterfeit coin cluster regions.
Vectors Vl, V2 and V~, which originate from the
origin of the x' coordinate system, terminate at
the genuine coin cluster centers for the sensor
response distributions for each of the coin
denominations, in effect mapping from the x'
system to x'' systems for each of the coin
clusters. This additional mapping to the x''
coordinate system saves on memory requirements and
computation time for the microprocessor.
Additional beneficial effects of this mapping
approach are discussed below.
Coin clusters are formed and optimized
for two sets of criteria. First, a mean vector
for each coin type, represented by vectors V1, V2
and V3 in Fig. 4, is created. These vectors are
determined based on empirical statistical data for
each coin. Once these vectors are determined,
increased flexibility in acceptance criteria can
be accomplished by allowing and increasing
~tolerance~ for the location of each vector.
Typically, a tolerance of plus and minus one count
for each vector is needed to maintain acceptance
rates greater than 90%. The cluster center can
also be offset by a tolerance of plus or minus two
count permutations from its true position, and
augmented again to achieve a higher acceptance
rate of genuine coins.
- 219~71~
-18-
The second criteria is to minimize slug
acceptance. The goal of attaining the required
slug rejection rate is addressed by removing the
portion of the augmented coin cluster that
overlAps the cluster region of a slug or slugs.
An example of a portion that would be removed is
shaded portion 0~ in Fig. 4. This portion 0~ has a
very low frequency of occurrence for valid coins,
and thus its removal minimally affects the coin
acceptance rate. In the presently preferred
embodiment, the resulting coin acceptance cluster
is represented by points in a three dimensional
space stored in a look-up table in memory.
Fig. 5 is a flow chart showing the
operation of this aspect of the invention. For an
initial coin denomination identification i=l
(block 503), the differences ( ~ ) between the
measured characteristics of the coins (X,,...~)
(block 502) and the respective center point for
each vector (Cntri,.. Cntr~) (block 504) are
compared against upper and lower limits (block
506). In terms of the variables used on Fig. 5, i
is the coin denomination index, m is the number of
measured coin parameters, (~ ) are the
lower limits and (U~ U~L) are the upper limits.
If the values do not fall within the
appropriate limits, then the coin denomination
index i is incremented (block 508) ~nd the
values are compared against the limits for another
coin denomination. When the ~alues are within
the limits, the system checks to see if the vector
formed by the values is in the look up table
(block 510); if the vector is in the table, then
the coin is accepted (block 512). The coin
denomination variable will be incremented until
valid data is determined or until all valid
denomination values have been ~earched (blocks
- 2194710
--19--
514, 516). Each time the, coin denomination index
~in is incremented, the system looks to that
portion of the look-up table relating to that coin
denomination.
In this manner a specific level of coin
acceptance is achieved while maintaining a high
level of slug rejection. ~urther, the method and
apparatus of the present invention attains the
rejection of slugs that produce sensor responses
that are not distinguishable from those of genuine
coins following an approach as illustrated in Fig.
3.
A further advantaqe stems from the fact
that the points defining the clusters may be
represented as vectors whose components are all
integer numbers and the cluster volume is a finite
set of integer values. Sensor response
measurements are taken relative to the x'
coordinate system allowing the use of a smaller
set of numbers than if the measurements were taken
relative to the x coordinate system. In addition,
the V vectors map the x' coordinate system to the
x'' coordinate system. If the mean is again
removed from each measurement, then an even
smaller set of integer numbers is needed to
represent the cluster volume. Conseguently, a
canonical code may represent the cluster volumes.
Representation of the coin clusters by canonical
codes makes practical the use of low cost
microprocessors having limited memory ~pace, in
that the ~pecific function for each cluster can be
easily stored in memory in a look-up table.
Fur,ther, a large degree of common~lity
was found to exist between clusters of different
coin types relative to the x~ coordinate system,
This commonality permits the large common portion
of cluster information for all coins to be stored
2194710
-20-
only once, and the remaining coin specific values
to be stored separately in microprocessor memory.
Consequently, a savings in memory requirements is
realized.
In the preferred embodiment, the look-up
table is stored in memory in a sorted fashion in
order to permit a fast ~search through the table.
The search starts in the middle of the table, and
uses a search technique for fast identification of
the portions of the table which contain the data
of interest.
It should be noted that in order to
stabilize the measurements and maintain a high
degree of genuine coin acceptance with varying
environmental changes, historical information for
each of the C0 and V vectors must be maintained,
and these vectors must also be varied when system
parameters change due to temperature, humidity,
component wear and the like. These vectors point
to the idle operating state of the system and are
functions of parameters which may experience step
changes as well as slow variations, all of which
require compensation and adaptive tracking to
provide a stable operating platform. Also, while
the V vectors for all coin types are compensated
in exactly the same manner, they can also be
compensated as a function of coin denomination.
It should also be noted that the coin
acceptance cluster may be created in two
dimensions rather than three, based on measurement
of two coin characteristics rather than three.
ANTI-FRAUD AND ANTI-CHEAT
Another aspect of the present invention
involves an improved method and apparatus for
avoiding a fraud practice where ~lugs have been
used in a prior art coin validator in an attempt
219471~1
to move the acceptance window toward the slug
distribution. The prior art method may be
understood by taking all f variables as
representing any function which might be tested,
such as frequency, amplitude and the like, for any
coin test. The specific discussion of the prior
art which follows will be in terms of frequency
testing for United States 5-cent coins using
circuitry as shown in Fig. 1 programmed to operate
as described below.
For initial calibration and tuning, a
number of acceptable coins, such as eight
acceptable 5-cent coins, are inserted to tune the
apparatus for 5 cent-coins. The frequency of the
output of sensor circuit 21 is repetitively
sampled and the frequency values f~ur~ are
obtained. A maximum difference value, f~ is
computed from the maximum difference between
f~ured and f~ during passage of the first 5-cent
coin. f=max(f~,~u~,~ - fo)~
Next, a dimensionless quantity, F, is
calculated by dividing the maximum difference
value f by fD where F=( f/f~). The computed F for
the first 5-cent coin is compared with the stored
acceptance limits to see if it lies within those
limits. Since the first 5-cent coin is an
acceptable 5-cent coin, its F value iE within the
limits. The first 5-cent coin is accepted and
microprocessor 35 obtains a coin count C for that
coin.
The coin count C is incremented by one
every time an acceptable coin is encountered until
it reaches a predetermined threshold number.
Until that threshold number is reached, new F
values are stored based on the last coin accepted.
When that threshold number i5 reached, a flag is
set in the software program to use the latest F
219~71~
-22-
value as the center point to determine the
acceptance limits of the acceptance ~window~ for
subsequently inserted coins. The originally
stored limits are no longer used, and the new
limits may be based on the latest F value plus or
minus a constant, or computed from the latest F
value in Any logical manner. Once t~e apparatus
is tuned as discussed above, it is capable of
performing in an actual operating environment.
The coin mechanism was designed to
continually recompute new F values and acceptance
limits as additional coins were inserted. If a
counterfeit coin was inserted, its F value
theoretically would not be within the acceptance
limits so the coin would be rejected. After
rejection of a counterfeit coin a new idling
frequency, f0, was measured and then the
microprocessor 35 awaited the next coin arrival.
Recomputation of the F values and
acceptance limits in this manner allowed the
system to self-tune and recalibrate itself and
thus to compensate for component drift,
temperature changes, other environmental shifts
and the like. In order for beneficial
compensation to be achieved, the computation of
new F values was done so that these values were
not overly weighted by previously accepted coins.
While achieving many benefits, the prior
art system has suffered because in practice a slug
exists whose measured characteristics overlap
those for a known acceptable coin as illustrated
in Fig. 7A. In Fig. 7A, the item designated 710
is a line distribution for certain measurement
criteria of a genuine coin. Curve 720 i6 a line
distribution for the same measurement criteria of
a slug. The overlap is shown as the shaded area
730 in Fig. 7A. As a result, the repeated
2194710
-23-
insertion of these slugs will move the window
center point toward the slug by tracking as those
slugs are accepted. Eventually, acceptance will
be 100% for the slug and poor for the valid coin.
The present invention addresses this
problem as discussed below.
Acceptance criteria for any given
denomination coin may be illustrated by the
measured distribution of coin test data from the
center point of a coin acceptance window. In the
preferred embodi~ent of the present invention, as
discussed earlier in this application, the
dimensionless quantity F is computed and then
compared with stored acceptance limits to see if
the computed value of F for the coin being tested
lies within a certain di-stribution in the coin
acceptance window. Fig. 6 is a representation of
such a distribution having a center point at zero
and acceptance limits at J+3~ and J-3J. Item 610
in Fig. 6 represents a measured criteria line
distribution for a genuine coin.
In practice, invalid coins have
distributions that slightly overlap those of
genuine coins. Item 710 in Fig. 7A depicts the
genuine coin line distribution of Fig. 6 having a
center point at JO~ ~ and the overlapping line
distribution of an invalid coin or slug having a
center point at J5~. The invalid coin line
distribution is designated as 720. Of course,
there are distributions for invalid coins other
than that chown in Fig. 7A, including
distributions to the left of the genuine coin
distribution 710. The genuine coin distribution
and the invalid coin distribution shown in Figs. 6
and 7A are exemplary only.
It is readily seen that the line
distribution of characteristic data for the
-
21~4710
genuine coin overlaps with the line distribution
for the invalid coin in the shaded area 730 shown
in Fig. 7A. For a coin mechanism employing window
self-adjustment, such a5 that described zbove with
respect to the prior art, repeated insertion of
invalid coins, some of which have characteristics
just within the outer edges of the genuine coin
acceptance window, will cause the ~ystem to move
the center point of the coin acceptance window
toward the distribution pattern of the invalid
coin. This ~tracking~ eventually results in
acceptance of invalid coins and rejection of
genuine coins. A person wishing to cheat or
defraud the coin mechanism need only repeatedly
insert a certain invalid coin into the coin
mechanism, thereby in effect programming the
system to accept non-genuine coins, resulting in a
significant loss of revenue.
To combat such behavior, the present
invention provides for improved invalid coin
rejection by preventing this ~tracking~ of the
center point of the acceptance window toward the
invalid coin distribution. This is accomplished
by sensing any invalid coin that has parameters
which fall close to the outer limits of the coin
acceptance window, ~uch ~s within ~ ~near miss~
area ~z~ in the invalid coin distribution between
points ~3~ and ~4~ on the graph in Fig. 7A.
The seguence of steps followed for thi~
method are ~et forth in the flow chart of Fig. 8.
First, a determination is ~ade whether a submitted
coin is valid (bl~ck 812, Fig. 8). Coins having
specified parameters within the genuine coin
acceptance window, for example as defined by
3s cymmetrical limits ~+3~ and ~-3~ around the center
point ~0~ of the genuine coin distribution of
Fiqs. 6 and 7A, are considered valid; those coins
2194710
-25-
outside of that coin acceptance window are
considered not valid.
If the coin is not valid, the system
determines whether the cheat mode flag is set
(block 802). If that flag is not set, a
determination is made whether the invalid coin
fits within the ~near miss~ area, 'z,~ between ~3~
and ~4~ on Fig. 7A (block 804). If the answer to
that inquiry is yes, the system moves the center
of the coin acceptance window a preset amount away
from the invalid coin distribution curve (block
806). For example, with reference to Fig. 7A, the
center of the coin acceptance window is moved from
~on to ~-ln. Alternatively, the right acceptance
boundary may be moved from ~3~ to ~2n. In either
case, very few genuine coins will not be accepted,
but essentially all invalid coins will now be
rejected, thereby preventing any attempted fraud.
A cheat counter is then cleared (block
808), and the cheat mode flag is set (block 810).
If another invalid coin is then inserted into the
mechanism, the system recognizes that the cheat
mode flag is set (block 802), and no changes are
made to the center position of the coin acceptance
window.
With regard to the Fig. 7A example, the
center of the coin acceptance window is maintained
at its ~ position until a preset, threshold
number of valid coins of the ~ame denomination are
counted in the cheat counter. The cheat counter
can be reset to zero if another invalid coin is
submitted to the mechanism which has a
characteristic which fits within the ~near miss~
area ~z~ on Fig. 7A.
Once the cheat counter reaches the
desired threshold number, the cheat mode flag is
cleared and the center of the coin acceptance
2194710
-26-
window is moved back to its original position.
These steps are 6hown on the Fig. 8 flowchart, in
the left-hand column, blocks 812 to 824.
Specifically, after block 812 determines
that the coin is valid, block 814 recognizes that
the cheat mode flag is set. If the valid coin is
the ~ame denomination as what triggered the cheat
mode flag (block 816), then the cheat counter is
incremented (block 818). When the cheat counter
reaches its preset threshold limit (block 820),
the cheat mode flag is cleared (block 822), and
the acceptance window is returned to its original
position (block 824).
In the Fig. 7A example, the center of the
coin acceptance window is moved from n_l~ back to
non once the threshold number of valid coins is
counted in the cheat counter.
By this method, attempts to train the
coin mechanism to accept counterfeit coins, slugs
and the like are thwarted, in that the center of
the coin acceptance window will not move toward
the invalid coin distribution if the user
repeatedly insert6 a number of the invalid coins
into the coin mechanism, even though some of these
coins would normally be acceptable and some would
only miss being ~cceptable by a 6mall amount such
that a slight movement of the acceptance criteria
would result in their acceptance. In fact,
according to thi6 aspect of the present invention,
the coin acceptance window moves away from the
invalid coin di~tribution for certain non-valid
coins or ~lugs, until such time A S a threshold
number of valid coins are counted.
The above described method can be used
for any denomination coins. Further, the value of
various parameters is adjustable, including but
not limited to the threshold value of genuine
21947~ ~
-27-
coins required to clear the cheat mode flag, the
width of that portion of the invalid coin
distribution which triggers the cheat mode (area
~z~ in ~ig. 7A), and the distance that the center
of the coin acceptance window i6 moved away from
the invalid coin distribution. These and other
parameters may be customized for each denomination
coin and any other special conditions relating to
the coin mechanism or the coins. For example, if
it is known that a counterfeit coin having a
certain distri~ution is often mistaken for a
genuine U.S. twenty-five cents coin, then the
acceptance window for this coin can be programmed
to move a distance out of the range of that
counterfeit coin and to stay there for a minimum
of lo or more genuine U.S. quarter coin
validations.
This anti-fraud and anti-cheat method and
apparatus may be used independently of the other
aspec~s of this invention in any coin testing
apparatus in which the coin criteria can be
adjusted by the control logic which controls the
coin, bill or other currency test apparatus.
However, the presently preferred embodiment is to
2~ incorporate this anti-fraud, anti-cheat aspect in
conjunction with the other aspects of the present
invention in one system.
~MPROVED COIN ACCEPTANCE WINDOW CENTER SE~F-
~DJ~STMENT
A method for self-adjustment of the
center of the
coin acceptance window involves accumulating a sum
of the deviations from the center of the coin
acceptance window for each coin. When the sum of
3~ deviations equals or exceeds a pre-set value, the
2194710
-28-
center position of the coin acceptance window is
adjusted.
By one aspect of the present invention,
only small or gradual deviations from the center
point of the coin acceptance window are added to
the running sum of deviations. Abrupt or large
deviations in the coin variables outside of this
small deviation band are ignored in terms of
center adjustment, as it is recognized that
adiustment based on such large deviations tends to
unduly shift the coin acceptance windows toward
the acceptance of counterfeit coins, slugs and the
like~ and away from acceptance of genuine coins.
Fig. 9 is a flow chart showing the steps
involved in this aspect of the present invention.
First, the coin mechanism is ~taught~ in the usual
manner, e.g., utilizing 8 valid coins to establish
the necessary information concerning the coin
acceptance window. Outside limits are then set
for the window in any one of a number of
conventional manners or using the cluster
technique described above. These steps are
combined in block 902, which states that the
window is established. If the coin is not
accepted as valid (block 904), no adjustment to
the center of the coin adjustment window
(designated in Fig. 9 as CNTR) is made and the
system waits for the next coin (block 903).
If the coin is determined to be valid
(block 904), then the absolute value difference
between M, the measured criteria for that
particular coin, and C~TR is compared to the
center adjustment deviation limit DEV (block 906).
If this absolute value difference is less than the
limit DEV, then the cumulative sum value CS is
modified by adding to it the value ~CNTR - M~
(block 908).
- -
2194710
-29-
If the absolute value difference between
M and CNTR exceeds the limit DEV (block 906), then
no adjustment is made to the cumulative sum CS,
and the system awaits arrival of the next coin.
When the cumulative sum CS equals or
exceeds a certain positive cumulative sum limit,
or is equal to or less than a negative cumulative
sum limit (block 910), the value of 'CNTR is
incremented by a preset amount or is decremented
by a preset amount, as appropriate (block 912).
The cumulative sum CS is then adjusted
accordingly, and the system awaits the arrival of
the next coin.
Thus, it is seen that only valid coins
having small deviations from the center value CNTR
of the coin adjustment window affect the self-
adjustment of that center value. Coins which
deviate outside this limited deviation range do
not effect the center self-adjustment. Since
counterfeit coins and slugs will almost in all
cases deviate from the center point CNTR more than
the limit DEV amount, this method virtually
insures that counterfeit coins, slugs and the like
will not affect the center self-adjust mechanism.
The method for protecting the center
self-adjustment mechanism described above allows a
wider coin acceptance window to be utilized,
thereby increasing the frequency that genuine
coins will be accepted by the system.
In the preferred embodiment, this
improved coin acceptance window center ~elf-
adjustment is utilized in combination with all
other aspects of the present invention. However,
it is to be understood that this center-adjust
method may be used independently of, or in various
combinations with, the aspects of the present
invention.
'' 219~710
-30-
RELATIVE VALUE COMPUTATION
It is beneficial to employ a low-cost
microprocessor to calculate the dimensionless F
value discussed above, which may also be referred
to as the relative value. To this end, in order
to perform calculations based upon the F value, a
scaling factor of 256 was utilized to ease
processing, and the resulting number was truncated
to the nearest integer.
This method of calculation resulted in
some loss of resolution. For example, when the
ratio of the scaling factor of 256 and the rest
value fO was greater than one, not all integer
values existed within the range covered by the
relative values F for a certain rest value f0. For
example, if the rest value f~ was 128 KHz, then the
relative value F would be even numbers. (F= f/128
*256 = f* 2). Similarly, only odd values of F
existed if f0 was an odd number. Further, when the
rest value fo changed, the list of non-existing
values changed also. Consequently, an expanded
look-up table was required in order to accomodate
all possible relative values F. This consumed
expensive memory space, and increased the
computation time ~pent for coin validation.
Also, use of such a high scaling factor
as 256 meant that oftentimes the integer value of
F was much greater than unity, and therefore extra
memory space was required to store the necessary
data for the F value, the center of the coin
acceptance window and the limits of that window.
Further, for sensors operating at high
frequencies, validation resolution was lost, as
one integer relative value F represented several
possible actual shift values f, due to
truncation. For example, if a sensor operated at
2194710
-31-
fo=1024 KHz, then 256 divided by 1024 equals 1/4,
which became the multiplier for the shift value
f. In this example, for f values of 4, 5, 6 and
7KHz, at fo=1024 KHz, F=1 for all four f values.
This resulted in a loss in resolution which
reduced the ability of the coin mechanism to
separate counterfeit from genuine coins.
~astly, in the prior art ~ystems,
truncation of the calculation of the F relative
value resulted in a 0.5 bias of the center of the
coin adjustment window. This is because all
values between integers were truncated downward.
Since window centers could only be adjusted in
increments of plus or minus one, the center was
always biased by plus or minus 0.5 in steady
state. This further reduced the coin acceptance
rate. If a plus or minus one expansion of the
window width was used to compensate for the
reduced coin acceptance rate, the result was
increased acceptance of counterfeit coins.
Another aspect of the present invention,
described below, provides additional resolution
over the usage in the prior art systems of the 256
scaling factor. The relative value F is now
preferably calculated according to the following
equation:
F= f * E(fo)/fo, where E(fo) is the exponentially
weighted moving average (also referred herein to
as the EWMA) of the rest value (f0) calculated for
each variable and coin denomination separately.
The theoretical eguation for the exponentially
weighted moving average ~t coin increment is:
EQUATION A: E(fo)1 ~ E(fo)l~ + W* (fOl - E(fo)1l) +
0.5 where W ~ weighing factor, and has a value
between 0 and l. The result is rounded as opposed
to truncated to eliminate the 0.5 bias error. For
the first validation measurement, E(fo) is set to
219471~
--32--
equal fO where fO is the rest value during the
~teaching~ of the unit, as that teaching is
described earlier in this application. Through
computer simulation, it has been determined that a
value for W of 1/40 results in the best
performance of the coin mechanism. Over time, the
ratio of E(fo)Jfol approaches unity in the steady
state of f0.
The ratio of the exponentially weighted
moving average (E ( f0) I) and the instantaneous rest
value (f0l) will have moderate deviations from
unity, with larger deviations being rare. On
those occasions when an abrupt change of the rest
value fO occurs, the ratio of E ( f0) JfO may
significantly deviate from unity, partially
compensating for the shift value f change. This
makes it possible for window center self-
adjustment without a significant expansion of the
window. Further, while the window is being self-
ad justed the ratio of the E (f,,) ,/f0l gradually comes
back to unity if no new perturbations occur for a
large enough amount of submitted coins.
Fig. ll shows a step change of the rest
value fO to fO' and the curve of the exponentially
weighted moving average E(fo)l shown as a dotted
line. Any step changes in rest values, g, that
would easily throw the shift values f outside the
acceptance window must be compensated for by E ( fO)
to provide a smooth transi.tion from one operating
point to another. Referring to Fig. 11, this
~;mooth transition ~hould be at a rate that is
slower than the tracking rate of the system.
E(fo)/fo allows the window center to track the
shift value with some delay as shown in Fig. ll.
3 5 As long as the relative deviation of the
rest value f0 from its exponentially weighted
moving average, multiplied by the shift value f,
219471~
-33-
is within the range plus or minus 0.5, this aspect
of the present invention does not create gaps
between relative values F. This method provides
for a sufficient coin acceptance rate allowing for
fast ~elf-adjustment of centers of coin acceptance
windows following abrupt and large changes in rest
values f0 in most cases. Further, the new method
produces relative values F having no loss of
resolution and al60 eliminates the 0.5 bias by
rounding, allowing for improved counterfeit coin
rejection. Another advantage is ease of
microprocessor implementation since the
exponentially weighted moving average can be
easily calculated. Current values of the
exponentially weighted moving average need to be
calculated separately for each rest value and
stored, and only one constant value of W need be
stored.
It should be noted that EQUATION A for
the exponentially weighted moving average given
above is just one example of an equation having
the required characteristics. The required
characteristics include that the ratio (E(fo)Jfol)
must go to unity in steady state, and that during
a transition in rest the ratio (E(fo)/fo) must be
such that when multiplied by the shift value f,
the relative value F must fall within the
acceptance window, so that an adjustment of the
center of the coin acceptance window can be made.
The exponentially weighted moving average
(EWMA) can be calculated to compensate for various
changes such as unit aging, wear, contamination
and cleaning, ambient temperature, etc. This can
be accomplished in the following manner, as shown
in the flow chart of Fig. 10.
The initial EWMA (E(fo)) equals the rest
value f0 at the time the mechanism is ~taught~.
' - 2194710
-34-
Deviations between the subsequently computed EW~
and the relevant rest value fOL are then summed
(block 102, Fig. 10). When the absolute value of
the sum of deviations (Sl) exceeds a threshold
value 1/~ (block 104), then the EWMA is
incremented or decremented by a preset ~mount
(depending on the sign of the deviation sum), and
the deviation sum is adjusted accordingly (block
106). In the preferred embodiment, the EWMA is
moved ~+1~ or ~ when the sum of deviations
exceeds the threshold value of l/W. If the sum of
deviations does not exceed the threshold, the
system awaits arrival of the next coin (block
112).
In place of freguency, any parameter
having a rest value (such as amplitude) may be
used.
A further aspect of the present invention
involves co~bining all of the above disclosed
methods in one coin, bill or other currency
validation apparatus. Of course, other
combinations and permutations of the above aspects
are also contemplated and may be found beneficial
by those skilled in the art.
In the preferred embodiment, with regard
to certain aspects of the present invention, the
microprocessor 35 is programmed according to the
attached printout appended hereto as an Appendix;
however, the operation of the electronic coin
testing apparatus 10 and the methods described
herein, will be clear to one skilled in the art
from the above discussion.
