Note: Descriptions are shown in the official language in which they were submitted.
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
METHODS AND APPARATUS FOR FACIAL RECOGNITION
CROSS-REFERENCE TO RELATED APPLICATIONS
[1] This application claims the benefit of U.S. Patent Application
16/297,366, filed on
March 8, 2019, which is a Continuation-in-Part of U.S. Patent Application No.
16/104,826,
filed on August 17, 2018, which is a Continuation-in-Part of U.S. Patent
Application No.
15/649,144, filed on July 13, 2017, which is a Continuation of U.S. Patent
Application No.
14/022,080, filed on September 9, 2013, now U.S. Patent No. 9,740,917, issued
August 22,
2017, which claims the benefit of U.S. Provisional Application No. 61/792,922,
filed on
March 15, 2013, and U.S. Provisional Application No. 61/698,347, filed on
September 7,
2012, the contents of which are expressly incorporated by reference herein in
their entireties.
BACKGROUND
[2] There has been a growing need for stronger identity verification to
protect personal
property, both physical and electronic. For example, it is important to
control access to
premises, vehicles, and personal property so that only authorized requesters
are allowed
access. A requester may be a user/person that requests access to an access
controlled assets
and/or infrastructure. In a traditional example, a requester may carry and use
a key, which is
designed to fit a lock to allow the requester of the key to open the lock and
gain entry. A loss
or damage to the key, however, can render access impossible. In another
example, a
requester may use a key fob to remotely lock or unlock the doors of a vehicle
by, e.g.,
pressing a button on the fob to generate an infrared ("IR") or radio frequency
("RF") signal,
which is detected by a sensor in the vehicle, which controls the doors. Such
vehicle keyless
access systems may require the requester to operate the ignition system. Other
similar
keyless access implementations may involve inserting and presenting a magnetic
card or the
like in a slot or a card reader/detector, or enabling an authorized requester
to key in a numeric
or alphanumeric code on a provided keypad. In each of these conventional
techniques,
however, it is very difficult to determine if the person holding the key/card
is the actual
authorized requester. An imposter may steal or duplicate a valid key and gain
unauthorized
accesses to the premise, vehicle, and/or personal property.
131 While traditional biometrics access control systems may mitigate some
shortcomings
of keys/cards-based access control systems, there may be limitations as well.
Traditional
biometric sensors, such as iris detection sensors, may be limited to specific
light conditions
significantly reducing both the effectiveness of the biometric sensor as well
as the possible
1
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
environments to apply same. The performance of biometric sensors may be
compromised in
direct sunlight due to glares, shadows, and other artifacts. Even with the
emergence of mega-
pixel camera technology, the features of each face may be obscured by ambient
lighting, the
position of the face, changes to the face, the background behind the face and
the quality of
the camera.
Motion blur, insufficient resolution, environmental impacts, lighting,
background, and camera angles collude to obscure subject details, making
heterogeneous
facial recognition (the matching of video and other probe images to large
databases of frontal
photographs) difficult.
[4]
Other factors may also increase the false acceptance and/or false recognition
rates of
traditional biometric sensors. For example, biometric sensors also have
difficulties obtaining
the necessary data in the absence of light. Light source shadowing and other
changes in
intensity may create contrasts on the face that may be misinterpreted as
facial features, and/or
slightly distort the measurement of the real facial features. Another major
source of
inaccuracy is the increased probability of similar measured features between
faces in a
growing population. Further, the problem of capturing the features of each
face may be
compounded by the desire for low maintenance and/or low complexity facial
recognition
systems. Therefore, improvement in access control may be desired.
SUMMARY
151 The
following presents a simplified summary of one or more aspects in order to
provide a basic understanding of such aspects. This summary is not an
extensive overview of
all contemplated aspects, and is intended to neither identify key or critical
elements of all
aspects nor delineate the scope of any or all aspects. Its sole purpose is to
present some
concepts of one or more aspects in a simplified form as a prelude to the more
detailed
description that is presented later.
[6]
Some aspects of the present disclosure include methods for generating a
sampled
profile including a plurality of sampling points having a plurality of
characteristic values
associated with the detected non-visible light, identifying one or more
macroblocks each
includes a subset of the plurality of sampling points, selecting a local
pattern value,
calculating a number of occurrences of the local pattern value within each
subset of the
plurality of the sampling points for each of the one or more macroblocks,
generating a first
array including a plurality of weighted values by calculating the plurality of
weighted values
based on the numbers of occurrences of the local pattern value and
corresponding sizes of the
one or more macroblocks, assigning a unique index to each of the plurality of
weighted
2
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
values, generating a second array of the unique index by ranking the plurality
of weighted
values, and generating a third array including a plurality of ranking
distances.
171 Certain aspects of the present disclosure include an edge capture
device (ECD)
ECDhaving an illumination source configured to emit an incident non-visible
light, an optical
sensor configured to detect a detected non-visible light, wherein the detected
non-visible light
includes a reflected non-visible light and a radiated non-visible light, one
or more processors
operatively coupled to the illumination source and the optical sensor, the one
or more
processors are configured to construct a biometric template of a requester
requesting access
to an entry point by generating a sampled profile including a plurality of
sampling points
having a plurality of characteristic values associated with the detected non-
visible light,
identifying one or more macroblocks each includes a subset of the plurality of
sampling
points, selecting a local pattern value, calculating a number of occurrences
of the local pattern
value within each subset of the plurality of the sampling points for each of
the one or more
macroblocks, generating a first array including a plurality of weighted values
by calculating
the plurality of weighted values based on the numbers of occurrences of the
local pattern
value and corresponding sizes of the one or more macroblocks, assigning a
unique index to
each of the plurality of weighted values, generating a second array of the
unique index by
ranking the plurality of weighted values, and generating a third array
including a plurality of
ranking distances.
[8] Aspects of the present disclosure include a computer readable medium
having code
stored therein that, when executed by one or more processors, cause the one or
more
processors to execute code for generating a sampled profile including a
plurality of sampling
points having a plurality of characteristic values associated with the
detected non-visible
light, code for identifying one or more macroblocks each includes a subset of
the plurality of
sampling points, code for selecting a local pattern value, code for
calculating a number of
occurrences of the local pattern value within each subset of the plurality of
the sampling
points for each of the one or more macroblocks, code for generating a first
array including a
plurality of weighted values by calculating the plurality of weighted values
based on the
numbers of occurrences of the local pattern value and corresponding sizes of
the one or more
macroblocks, code for assigning a unique index to each of the plurality of
weighted values,
generating a second array of the unique index by ranking the plurality of
weighted values,
and code for generating a third array including a plurality of ranking
distances.
191 An aspect of the present disclosure includes a system having means for
generating a
sampled profile including a plurality of sampling points having a plurality of
characteristic
3
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
values associated with the detected non-visible light, means for identifying
one or more
macroblocks each includes a subset of the plurality of sampling points, means
for selecting a
local pattern value, means for calculating a number of occurrences of the
local pattern value
within each subset of the plurality of the sampling points for each of the one
or more
macroblocks, means for generating a first array including a plurality of
weighted values by
calculating the plurality of weighted values based on the numbers of
occurrences of the local
pattern value and corresponding sizes of the one or more macroblocks, means
for assigning a
unique index to each of the plurality of weighted values, generating a second
array of the
unique index by ranking the plurality of weighted values, and means for
generating a third
array including a plurality of ranking distances.
[10] Aspects of the present disclosure include an infrastructure having an
access-controlled
entry point, a ECD configured to emit an incident non-visible light onto a
face of a requester,
detect a detected non-visible light from the face of the requester, wherein
the detected non-
visible light includes a reflected non-visible light and a radiated non-
visible light, generate a
biometric template of the requester by generating a sampled profile including
a plurality of
sampling points having a plurality of characteristic values associated with
the detected non-
visible light, identifying one or more macroblocks each includes a subset of
the plurality of
sampling points, selecting a local pattern value, calculating a number of
occurrences of the
local pattern value within each subset of the plurality of the sampling points
for each of the
one or more macroblocks, generating a first array including a plurality of
weighted values by
calculating the plurality of weighted values based on the numbers of
occurrences of the local
pattern value and corresponding sizes of the one or more macroblocks,
assigning a unique
index to each of the plurality of weighted values, generating a second array
of the unique
index by ranking the plurality of weighted values, and generating a third
array including a
plurality of ranking distances, store a plurality of biometric templates of
authorized
personnel, compare the biometric template of the requester with the plurality
of biometric
templates of authorized personnel, generate a positive match signal in
response to identifying
a match between the biometric template of the requester and one of the
plurality of biometric
templates of authorized personnel, and transmit the positive match signal to a
gateway to
grant the requester access to the entry point.
BRIEF DESCRIPTION OF THE DRAWINGS
[11] The features believed to be characteristic of aspects of the
disclosure are set forth in
the appended claims. In the description that follows, like parts are marked
throughout the
4
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
specification and drawings with the same numerals, respectively. The drawing
figures are
not necessarily drawn to scale and certain figures may be shown in exaggerated
or
generalized form in the interest of clarity and conciseness. The disclosure
itself, however, as
well as a preferred mode of use, further objects and advantages thereof, will
be best
understood by reference to the following detailed description of illustrative
aspects of the
disclosure when read in conjunction with the accompanying drawings, wherein:
[12] FIG. 1 is an example concurrent real-time identity verification and
authentication
system, in accordance with some aspects of the present disclosure;
[13] FIG. 2 shows a perspective view of an example of a concurrent real-time
identity
verification and authentication device, in accordance with some aspects of the
present
disclosure;
[14] FIG. 3 shows a frontal view of an example of a concurrent real-time
identity
verification and authentication device, in accordance with some aspects of the
present
disclosure;
[15] FIG. 4 shows another perspective view of an example of a concurrent real-
time
identity verification and authentication device, in accordance with some
aspects of the
present disclosure;
[16] FIG. 5 is a block diagram of an example processing component of a
concurrent real-
time identity verification and authentication device, in accordance with some
aspects of the
present disclosure;
[17] FIG. 6 shows a flow diagram of a facial recognition method, in accordance
with some
aspects of the present disclosure;
[18] FIG. 7(a) shows a facial image for a person, in accordance with some
aspects of the
present disclosure;
[19] FIG. 7(b) shows a different facial image for the same person, in
accordance with some
aspects of the present disclosure;
[20] FIG. 8 shows an example process for calculating local binary pattern
(LBP) feature, in
accordance with some aspects of the present disclosure;
[21] FIG. 9 shows an example process for calculating local ternary pattern
(LTP) feature,
in accordance with some aspects of the present disclosure;
[22] FIG. 10 shows positions of three example key features selected among one
or more
face images, in accordance with some aspects of the present disclosure;
[23] FIG. 11 shows an example of a receiver operating characteristic (ROC)
curve for
testing a face database, in accordance with some aspects of the present
disclosure;
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[24] FIG. 12 illustrates an example of biometric, asymmetric encryption for
confidentiality, in accordance with some aspects of the present disclosure;
[25] FIG. 13 illustrates another example of biometric, asymmetric encryption
for
authentication, in accordance with some aspects of the present disclosure;
[26] FIG. 14 illustrates a schematic view of an example of an environment for
implementing one or more gateways for access control;
[27] FIG. 15 illustrates an example of a computer system for implementing a
method of
managing data in accordance with aspects of the present disclosure;
[28] FIG. 16 illustrates a block diagram of various exemplary system
components, in
accordance with aspects of the present disclosure;
[29] FIG. 17 illustrates an example of a ECD for identifying biometric
templates, in
accordance with aspects of the present disclosure;
[30] FIG. 18 illustrates an example of the components of the ECD of FIG. 17,
in
accordance with aspects of the present disclosure;
[31] FIG. 19 illustrates another example of the components of the ECD of FIG.
17, in
accordance with aspects of the present disclosure;
[32] FIG. 20 illustrates an example of a sampled profile, in accordance with
aspects of the
present disclosure;
[33] FIG. 21 illustrates an example of LBP operation on measurement points of
the
sampled profile of FIG. 20, in accordance with aspects of the present
disclosure;;
[34] FIG. 22 illustrates examples of sub-matrices;
[35] FIG. 23 illustrates an example of a table of results for sequence
conversion, in
accordance with aspects of the present disclosure;
[36] FIG. 24 illustrates an example of a flow chart for converting a sequence,
in
accordance with aspects of the present disclosure;
[37] FIG. 25 illustrates an example of a table of verification
calculations, in accordance
with aspects of the present disclosure;
[38] FIG. 26 illustrates an example of deep learning;
[39] FIG. 27 illustrates another example of deep learning; and
[40] FIG. 28 illustrates a flow chart of a method for identifying biometric
templates.
DETAILED DESCRIPTION
6
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[41] The following includes definitions of selected terms employed herein. The
definitions
include various examples and/or forms of components that fall within the scope
of a term and
that may be used for implementation. The examples are not intended to be
limiting.
[42] A "processor," as used herein, processes signals and performs general
computing and
arithmetic functions. Signals processed by the processor may include digital
signals, data
signals, computer instructions, processor instructions, messages, a bit, a bit
stream, or other
computing that may be received, transmitted and/or detected.
[43] A "bus," as used herein, refers to an interconnected architecture that is
communicatively coupled to transfer data between computer components within a
singular or
multiple systems. The bus may be a memory bus, a memory controller, a
peripheral bus, an
external bus, a crossbar switch, and/or a local bus, among others. The bus may
also be a
vehicle bus that interconnects components inside a vehicle using protocols,
such as Controller
Area network (CAN), Local Interconnect Network (LIN), among others.
[44] A "memory," as used herein may include volatile memory and/or non-
volatile
memory. Non-volatile memory may include, for example, ROM (read only memory),
PROM
(programmable read only memory), EPROM (erasable PROM) and EEPROM
(electrically
erasable PROM). Volatile memory may include, for example, RAM (random access
memory), synchronous RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM
(SDRAM), double data rate SDRAM (DDR SDRAM), and/or direct RAM bus RAM
(DRRAM).
[45] As used in the specification and the appended claims, the singular forms
"a," "an" and
"the" include plural referents unless the context clearly dictates otherwise.
[46] Ranges may be expressed herein as from "about," "substantially," or
"approximately"
one particular value and/or to "about," "substantially," or "approximately"
another particular
value. When such a range is expressed, another implementation includes from
the one
particular value and/or to the other particular value.
[47] By "comprising" or "containing" or "including" is meant that at least the
named
compound, element, particle, or method step is present in the composition or
article or
method, but does not exclude the presence of other compounds, materials,
particles, method
steps, even if the other such compounds, material, particles, method steps
have the same
function as what is named.
[48] It is also to be understood that the mention of one or more method steps
does not
preclude the presence of additional method steps or intervening method steps
between those
steps expressly identified. Similarly, it is also to be understood that the
mention of one or
7
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
more components in a device or system does not preclude the presence of
additional
components or intervening components between those components expressly
identified.
[49] Biometric identification techniques generally refer to pattern
recognition techniques
that perform a requester identification process by determining the
authenticity of a specific
physiological or behavioral characteristic possessed by the requester. In some
instances,
biometric identification may be preferred over traditional methods involving
passwords and
personal identification numbers (PINs) for various reasons. For example, with
biometric
identification, the person (e.g., requester) to be identified is typically
required to be
physically present at the point-of-identification. Additionally,
identification based on
biometric techniques obviates the need to remember a password or carry a token
(i.e., a
security device used to gain access to an access controlled entry point).
[50] One kind of texture based local binary pattern ("LBP") feature describes
facial
information that produces desirable recognition results. The improved local
ternary pattern
("LTP") feature may be a further improvement over conventional LBP methods.
LBP and
LTP features may not be sensitive to light and expression variations and are
computationally
efficient, but they also have shortcomings, such as information redundancy due
to correlation
between the positive histogram and the negative histogram.
[51] It is therefore desirable to contemplate concurrent real-time identity
verification and
authentication techniques to create biometric signature data for providing
keyless access to
authorized requesters to a vehicle, building, or the like with varying degrees
of security by
utilizing various types of biometric data of authorized requesters. As
discussed above, in
some implementations of the present disclosure, the biometric signature data
may be
interchangeable across a wide variety of applications. Accordingly, in some
examples of the
present disclosure, the same biometric signature data for a person may be used
to authenticate
that person at one or more locations and for one or more applications.
Additionally, an
example of a biometric system in the present disclosure allows the biometric
signature data to
be altered based on a desired security level. Thus, the type of biometric
signature data that
may be used for a particular application and/or relating to a particular
requester may vary
depending on the security level desired for that particular application and/or
requester. While
some implementations discussed herein are discussed in the context of facial
biometric data,
those skilled in the art would understand that various implementations of the
present
disclosure may employ many types of biometric data, including, but not limited
to,
fingerprint data, iris and retinal scan data, speech data, facial thermograms,
hand geometry
data, and the like.
8
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[52] In some implementations of the present disclosure, the biometric data
associated with
the intended recipient (e.g., a biometric template) may be obtained via a
biometric sensor of a
biometric-based access control system. As will be discussed below, variations
in light,
temperature, distance of the biometric sensor from a target may impact the
quantity and
quality of the biometric data obtained via the biometric sensor. For example,
variations in
light intensity and angle may create shadows on the face of a requester,
making facial
recognition more difficult. If the biometric data for identifying a requester
is obscured, more
templates may be needed to properly authenticate the requester, thus
increasing the quantity
of the biometric data necessary. To reduce the undesirable impact of these
environmental
factors, the biometric sensor may utilize either near infrared (IR) or
ultraviolet (UV) light or a
combination of both IR and UV at desired intensities. In an implementation,
the method uses
near IR light. An Infrared light emitting diode (LED) array may be utilized in
the facial
recognition device or biometric sensor to minimize the impact of the
surrounding lighting on
capturing the facial uniqueness. The camera and the LED array are packaged
into a dedicated
edge device (e.g., an ECD or a faceplate) mounted at a location requiring
verification and/or
identification/analysis, such as a door requiring access control.
[53] In some implementations, an access control system may utilize IR or new
IR
illumination and detection to identify facial features. IR or new IR lighting
may penetrate
into the dermis of the face. The IR or new IR lighting may penetrate into the
dermis by 10
micrometers, 20 micrometers, 50 micrometers, 100 micrometers, 200 micrometers,
500
micrometers, 1 millimeters, 2 millimeters, 5 millimeters, and/or 10
millimeters. Other
penetration depths are possible. The penetration depths may depend on the
location of the
body, wavelength of the infrared lighting, and/or intensity of the infrared
lighting. The
penetration may expose characteristics of the skin that may be difficult to
see in visible light
including (age spots, spider veins, hyperpigmentation, rosacea, acne, and
porphyrins). The
identification of these subdermal features may be used to adjust/supplement
the unique
identification of the requester. These features on the face of the requester
may be unique
because they are based on the requesters exposure to nature and the sun over
the life of the
requester. Facial recognition based on subdermal features may identify the
uniqueness of the
face at the time of capture to provide opportunities for identification
analysis. The number of
subdermal features may increase over time with exposure to the sun and on a
daily basis.
[54] In another example, an access control system may utilize ultraviolet
illumination and
detection to identify facial features. Ultraviolet lighting may penetrate into
the dermis of the
face. The UV lighting may penetrate into the dermis by 10 micrometers, 20
micrometers, 50
9
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
micrometers, 100 micrometers, 200 micrometers, 500 micrometers, 1 millimeters,
2
millimeters, 5 millimeters, and/or 10 millimeters. Other penetration depths
are possible. The
penetration depths may depend on the location of the body, wavelength of the
ultraviolet
lighting, and/or intensity of the ultraviolet lighting. The
penetration may expose
characteristics of the skin that may be difficult to see in visible light
including (age spots,
spider veins, hyperpigmentation, rosacea, acne, and porphyrins). The
identification of these
subdermal features may be used to adjust/supplement the unique identification
of the
requester. These features on the face of the requester may be unique because
they are based
on the requesters exposure to nature and the sun over the life of the
requester. Facial
recognition based on subdermal features may identify the uniqueness of the
face at the time
of capture to provide opportunities for identification analysis. The number of
subdermal
features may increase over time with exposure to the sun and on a daily basis.
The facial
recognition system of the present disclosure may estimate the age of a person
based on the
quantity and nature of the subdermal features. The access control system may
also track the
change in these features over time to confirm the individual's identity and
establish lifestyle
and daily routines based on interpretations of the subdermal features.
Subdermal facial
recognition may also increase the difficulty of creating a duplicate (e.g.,
duplicate of a
biometric template) of the face due to its elimination of dependency on facial
features capable
of being captured by standard visible wavelength photography and camera
technology. The
access control system may also further obfuscate the content of the
ultraviolet capture by
introducing time-sequenced cross-polarization filters to the capturing process
that further
eliminates the ability to present an artificial duplicate of the face to the
access control system.
[55] A benefit of the system in the present disclosure includes allowing a
single credential
system replacing PINs, passwords, and multi-factor authentication that is
seamless to the
requester. With this architecture in place, the requester(s) of the system may
rely on a single
credential management solution. The system of the present disclosure may
support both
logical and physical gateways. In some implementations of the present
disclosure, the system
may provide protection at home and at work.
[56] Aspects of the present disclosure may include a method referred to as
"layered
reinforcement." The method comprises of taking the image of face from the
biometric sensor
and overlaying several layers of different size pixel boxes on the image. This
layering of
pixel boxes of different sizes has an amplifying impact on the analysis of the
uniqueness of
the face. Areas that are more unique to the face are amplified. Areas that are
more common
among faces are deemphasized. As a result, layered reinforcement may improve
the
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
algorithm performance while allowing the method to handle a large number of
users at
multiple sites where the biometric sensor ECD is deployed. The "layer
reinforcement" of the
method may allow for the processing of the same number of requesters on a
local Advanced
Reduced Instruction Set Computing Machine (ARM) processor at the biometric
sensor ECD
where the image is first captured, thus reducing hardware and processing
requirements and
contributing to the accuracy and reliability of the method as a network
failure cannot prevent
the biometric sensor ECD from processing a face verification.
[57] Some aspects of this embodiment of the invention cover the use of a
gateway
(described below) to manage the data analyzed by the various algorithms to
increase
performance by decreasing false negative and false positive results through
the following
processes: pixel box hierarchical analysis to create binary tree of dominant
features (i.e.,
determining what is the most distinctive feature); pixel box time domain
analysis with heat
maps (i.e., determining over time features that are problematic due to overlap
among
subjects); and binary tree collision (flagging overlap of biometric signature
data for two
subjects that may cause a false positive and addressing in a proactive
fashion).
[58] Benefits to the system of the present disclosure include improved
performance when
accuracy requires reduction in false negative and false positive results. The
improvement
also allows for the benefits of 1:1 comparison in a 1:N environment as a
potential
replacement to video surveillance and comparison thereby opening up the
massive
surveillance market to significantly increased accuracy.
[59] Referring to FIG. 1, an example of an identification system 100 for
concurrent real-
time identity verification and authentication for use in, e.g., allowing
access by an authorized
requester to a vehicle, building, or the like is illustrated in accordance
with aspects of the
present disclosure.
[60] It should be appreciated that FIG. 1 is intended to describe aspects
of the disclosure to
enable those skilled in the art. Other implementations may be utilized and
changes may be
made without departing from the scope of the present disclosure.
[61] The identification system 100 comprises a concurrent real-time
identity verification
and authentication device 102 including at least one biometric sensor 104, a
processor 106,
memory 108, a display 110, and input/output mechanism 112. The identification
system 100
may be used to secure or control access to a secured area, device, or
information, such as an
airport boarding area, building, stadium, database, locked door, vehicle, or
other access
controlled assets/infrastructure.
11
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[62] The biometric sensor(s) 104 may include a camera, a fingerprint reader,
retinal
scanner, facial recognition scanner, weight sensor, height sensor, body
temperature sensor,
gait sensor, heartbeat sensor, or any other sensor or device capable of
sensing a biometric
characteristic of a person. As shown in FIGS. 2-4, in an exemplary
implementation of the
present disclosure, the biometric sensor(s) 104 may be an optical sensor, such
as a camera.
[63] In some aspects, the biometric sensor(s) 104 may include an optical
sensor that
captures visual data. For example, the biometric sensor(s) 104 may be a camera
that senses
visual information of a requester, such as the facial features of the person.
The facial features
of the person may include the textures, complexions, bone structures, moles,
birthmarks,
contours, coloring, of the face of the person. The biometric sensor(s) 104 may
capture the
facial features of the person and convert the visual information into digital
sensed
information as discussed below).
[64] The processor 106 may be configured for comparing the sensed information
via
biometric sensor(s) 104 with known characteristics of a person in an attempt
to identify the
person via biometric signature data. The processor 106 may include any number
of
processors, controllers, integrated circuits, programmable logic devices, or
other computing
devices. The processor 106 may be communicatively coupled with the biometric
sensor(s)
104 and other components of the system 100 through wired or wireless
connections to enable
information to be exchanged between the device 102 and external devices 114 or
systems
(e.g., network 116) to allow for comparison of the stored biometric signature
data with the
sensed information obtained from the biometric sensor(s) 104.
[65] The processor 106 may implement a computer program and/or code segments
stored
on memory 108 to perform some of the functions described herein. The computer
program
may include an ordered listing of executable instructions for implementing
logical functions
in the device 102. The computer program can be embodied in any computer-
readable
medium (e.g., memory 108) for use by or in connection with an instruction
execution system,
apparatus, or device, and execute the instructions. The memory 108 may
contain, store,
communicate, propagate or transport the program for use by or in connection
with the
instruction execution system, apparatus, or device. Examples of memory 108 may
include an
electrical connection having one or more wires, a random access memory (RAM),
a read-
only memory (ROM), an erasable, programmable, read-only memory (EPROM or Flash
memory), a portable computer diskette, or a portable compact disk read-only
memory
(CDROM). The memory 108 may be integral with the device 102, a stand-alone
memory, or
a combination of both. The memory 108 may include, for example, removable and
non-
12
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
removable memory elements such as RAM, ROM, Flash, magnetic, optical, USB
memory
devices, and/or other conventional memory elements.
[66] In some aspects, the memory 108 may store the known characteristics of a
number of
people and various other data associated with operation of the system 100,
such as the
computer program and code segments mentioned above, or other data for
instructing the
device 102 and other device elements to perform the aspects described herein.
The various
data stored within the memory 108 may be associated within one or more
databases (not
shown) to facilitate retrieval of the information, e.g., via the external
devices 114 or the
network 116. Although the memory 108 as shown in FIG. 1 is integrated into the
device 102,
it should be appreciated that memory 108 may be stand-alone memory positioned
in the same
enclosure as the device 102, or may be external memory accessible by the
device 102.
[67] In an aspect, the display 110 may be configured to display various
information
relating to the system 100 and its underlying operations. For example, a
notification device
may be included (not shown) for indicating the sensed biometric characteristic
or the sensed
signal fail to match the known characteristics of the person and may include
an audible alarm,
a visual alarm, and/or any other notification device.
[68] In an aspect, the device 102 may also include input/output mechanism 112
to facilitate
exchanging data and other information among different components within the
device 102, or
with various the external devices 114 or systems via the network 116.
[69] For example, various I/O ports may be contemplated including a Secure
Disk Digital
(SD) card slot, Mini SD Card slot, Micro SD Card slot or the like for
receiving removable SD
cards, Mini SD Cards, Micro SD Cards, or the like, and a USB port for coupling
with a USB
cable communicatively coupled with another computing device such as a personal
computer.
In some aspects, the input/output mechanism 112 may include an input device
(not shown)
for receiving identification information about a person-to-be-identified. The
input device
may include a ticket reader, a credit card reader, an identification reader, a
keypad, a touch-
screen display, or any other device. In some other aspects, as described
above, the
input/output mechanism 112 may be configured to enable the device 102 to
communicate
with other electronic devices through the network 116, such as the Internet, a
local area
network, a wide area network, an ad hoc or peer to peer network, or a direct
connection such
as a USB, Firewire, or BluetoothTM connection, etc. In one example, known
characteristics
about persons may be stored and retrievable in remote databases or memory via
the network
116. The input/output mechanism 112 may thus communicate with the network 116
utilizing
wired data transfer methods or wireless data transfer methods such as WiFi
(802.11), Wi-
13
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
Max, BluetoothTM, ANT , ultra-wideband, infrared, cellular telephony, radio
frequency, etc.
In an aspect, the input/output mechanism 112 may include a cellular
transceiver for
transmitting and receiving communications over a communications network
operable with
GSM (Global System for Mobile communications), CDMA (Code Division Multiple
Access), or any other known standards.
[70] The device 102 may also include a power source (not shown) for providing
electrical
power to the various components contained therein. The power source may
include batteries,
battery packs, power conduits, connectors, and receptacles operable to receive
batteries,
battery connectors, or power cables.
[71] In an aspect, the device 102 may be installed and positioned on an access
control
entry point (not shown) such as a gate, locked door, etc. for preventing
persons from
accessing certain areas until the device 102 determines that the sensed
biometric
characteristic and/or signal match the known characteristics. In some other
aspects, as shown
in FIGS. 2-4, the device 102 may be a stand-alone, compact, handheld, and
portable device.
In one example, one may use such a stand-alone, compact, handheld, and
portable device to
protect sensitive documents or information that are electronically stored and
accessed on the
Internet and/or an intranet. In some aspects, a concurrent realtime identity
verification facility
access unit may use biometric signature data to create interchangeable
authentication for a
variety of uses (e.g., office, home, smart phone, computer, facilities).
[72] Referring to FIG. 5, the processor 106 in FIG. 1 may be configured to
include, among
other features, a detection module 502 and a recognition module 508 for
providing concurrent
real-time or near real-time identity verification and authentication with
keyless access to
authorized requesters to secured facilities or information. The detection
module 502 may
include a face detection module 504 for detecting facial features of a
requester. The detection
module 502 may include an eye detection module 506 for identifying the
locations of the
eyes of a requester. In some implementations, the detection module 520 may
include one or
both the face detection module 504 and/or the eye detection module 506. In
some aspects,
the processor 106 may receive inputs (digital or analog) from the sensor(s)
104.
[73] FIG. 6 describes an example procedure of selecting key features from a
database with
a large number of facial information and building one classifier which can
distinguish
different faces accordingly. LBP and LTP may be used to provide a full
description of face
information, and then with the use of an adaptive boosting ("adaboost")
learning algorithm,
one may select key features and build a classifier to distinguish different
faces by creating
biometric signature data. This biometric signature data may be used to create
universal
14
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
verification and authentication that can be used for a variety of applications
(e.g., computer,
building access, smartphone, automobile, data encryption) with varying degrees
of access and
security (e.g., access to network, but heightened security for requester
computer). At block
602, create face sample database. For example, the processor 106 and the
recognition
module 508 may create a face sample database using unrecognized face samples.
In one
implementation, the processor 106 and the recognition module 508 may store,
into the
memory 108, 1000, different persons with each person showing 10 different
postures and/or
expressions.
[74] At block 604, extract LBP and LTP features. For example, the detection
module 502
and/or the face detection module 504 may extract LBP and LTP features from
different
blocks in different positions of each face sample.
[75] At block 606, calculate positive sample and negative sample. For example,
at least
one of the face detection module 502, the face detection module 504, and the
eye detection
module 506 may calculate the feature absolute value distance for the same
position of any
two different images from one person and set this distance as positive sample
feature
database. Further, the face detection module 502 and the face detection module
504 may
jointly or separately calculate the feature absolute value distance for the
same position of any
two different images from different person and set this distance as negative
sample feature
database.
[76] At block 608, build adaboost classifier. For example, the face detection
module 502
and the face detection module 504 may select the most distinguishable key
feature from the
candidate feature database with adaboost and create a face classifier.
[77] At block 610, generate recognition result. For example, the recognition
module 508
may generate recognition result. Once there is a fixed dataset of macro blocks
and the
specific LPB ranging from 1 to 255 is determined, a value is assigned to that
unit of the
dataset based upon the number of pixels within the block that satisfy that
specific LBP. For
example, assuming a 10 x 10 macro block in unit number 1 of 255 and LBP of 20,
the
method 600 determines the number of pixels in the histogram that fall within
that LBP of 20
on scalar value. The method 600 calculates scalar value and then normalize
value in a second
array to address the problem of determining value within various sized macro-
blocks. The
scalar value based upon the known method was based on size of macro-block
where the
maximum value could be from 100 to 1600 depending upon the size of the macro-
block. The
scalar value in this second array may now a percentage of the total pixels
available in that
macro block to normalize the data for the subsequent assessment. Normalization
causes the
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
data to not be skewed based on the size of the macro block. After
normalization under the
improved method, each unit of the data set in this second array has the same
weight. This
normalized data may be then sorted to establish and assign a value from 1 ¨
2165 where the
scale reflects the highest normalized value going to the top of the sort. For
example if dataset
2000 had the highest value in the array it would be assigned a value of 1 with
descending
value reflecting the datasets that have lower values. The second normalized
array may then
be converted to a third simulated dna sequencing array where the position is
established
within this third array based upon its value in previous sort. The third array
assesses the
position and calculates the differences between where the data set appears in
the sequence
(e.g., ranking distance). This improved method analyzes traits as opposed to
scalar value
based upon the uniqueness of traits within the face and not merely on scalar
values.
[78] At block 620, test face sample. For example, the detection module 502 may
optionally test face samples.
[79] At block 622, extract LBP and LTP features. For example, the detection
module 502
and/or the face detection module 504 may extract LBP and LTP features from
different
blocks in different positions of each face sample.
[80] Further, online recognition may include the following steps:
[81] (1) Calculate the offline stage extracted key feature of different
blocks in different
positions for face sample to be identified.
[82] (2) Calculate the key feature selected from step (1) with that of each
face sample in
database and determine whether they belong to the same person or not. If
calculated distance
is less than the set threshold, it may be determined that they are the same
person, otherwise it
may be determined that they are not.
[83] As shown in FIG. 8, an example process starts with creating a face
database with
different postures and different expressions. For example, one may include the
images of,
e.g., 1000, different persons and each person shows, e.g., 10, images
differently. FIG. 7(a)
shows the different face image of one same person, and FIG. 7(b) refers to the
different face
image of different person.
[84] LBP and LTP may be used to describe face. FIG. 8 shows a calculating
process of
LBP features, and FIG. 9 shows a calculating process of LTP features. In order
to obtain as
many features to describe face information, different block size may be
divided on different
positions of face sample. For example, face size can be 100 x 100, block size
may be w x h,
w and h values can range from 2 to 100, and 7837 blocks may be selected as a
result. The
16
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
identification system 100 may select the bin features of LBP and LTP on
different block size
and make it as the final candidate feature database.
[85] The next step is to calculate positive samples and negative samples. The
bin feature
absolute value distance of the same position for different images from a same
person may be
calculated and set as the positive sample. Additionally, the bin feature
absolute value
distance of same position for different persons may be calculated and set as
the negative
sample. For example, the result may involve calculating 32356 positive samples
and
58747698 negative samples.
[86] Thereafter, the key bin feature that can distinguish all positive and
negative samples
among the large number of feature database may be selected with a learning
algorithm. For
example, one may choose the learning algorithm of discrete adaboost to select
feature and
build a classifier.
[87] An example algorithm of using adaboost to classify may include the
following
computational steps:
[88] 1. Given f as the maximum negative sample error rate, d as the minimum
positive
sample correct rate, Ft., as the target of negative sample error rate, and
Dtar as the target of
positive sample correct rate that cascade classifier has to achieve. P, N are
the positive and
negative database, respectively.
[89] 2. Set Fo=1.0, Do=1.0, and i=0;
[90] 3. When Fi>Ftar, i=i+1, ni=0, Fi=Fi_i; when Fi>fx Fi_i, ni=ni+1.
[91] 4. Compute the strong classifier with n features via adaboost in database
P and N;
calculate F, and Di for current cascade classifier, adjust the threshold value
of current strong
classifier until the rate is no less than dxDi_i, N is nonempty set.
[92] 5. If Fi>Ftar, classify the currently obtained cascade classifier in
other negative sample
image and determine, put wrongly determined image into N.
[93] 1) Given n computing sample (xi, yi), . . . , (xn, yn), yi=0, 1, xi
presents negative
sample label and positive sample label, respectively.
[94] 2) Initialize weight
_ 1 1
¨ 2m '
where the number of positive samples is 1 and the number of negative samples
is m.
[95] 3) Try t from 1 to T and run below steps repeatedly:
[96] a) Normalize weight wt =
17
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[97] b) Compute a weak classifier 111 for each feature f, and mark the error
rate of this
classifier Ej=Iiwt,j111Axi-Y)1
[98] c) Find out classifier ht with lowest error rate cot among all weak
classifier computed
from last step,
[99] d) Update weight wt+i,t=wt,13t1-e among which r3t=e 41- Et), If x, is
correctly classified,
e1=0. Otherwise e1=1.
[100] Get the strong classifier lastly: if It=iTatht(xP1/2It=iTat
then h(x)=1, otherwise h(x)=0. There, at=log 1/f3t
[101] FIG. 10 shows the position of the first three key features selected
among face image
by taking online testing for face database of 100 persons based on offline
selected features
and classifier.
[102] FIG. 11 shows recognition results for 100 persons, wherein X axis
represents false
accept rate, which means the wrongly identified rate of face samples. Y axis
represents
verification rate, which means the rate of face samples correctly recognized.
As shown in
FIG. 11, when the false accept rate is below 10-4, it may achieve 95%
recognition rate. The
face recognition in this example not only improves the robustness of face
sample, but also
reduces its computational complexity thus improves the face recognition
significantly.
[103] Referring back to FIG. 5, in some aspects, the detection module 502 may
be
configured to use, among other features, a face detection module 504 and an
eyes detection
module 506 for processing the acquired image of the person-to-be-identified as
follows.
[104] Face Detection Module 504
[105] Inputs: Acquired frontal face image (grey image), face classifier
[106] Outputs: Face frame positions, and the number of faces
[107] Flow:
[108] a. Reduce the acquire frontal face image to user-defined size
[109] b. Calculate an integral image of the reduced image
[110] c. Initialize a traverse window based on the size defined by the face
classifier, e.g.,
20x20
[111] d. Move the traverse window on the integral image from left to right and
then from
top to bottom with each move distance corresponding to a user-defined
distance. However, if
the user-defined distance is zero, set the move distance as 1/20 of the width
of the traverse
window.
18
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[112] e. Use the face classifier to determine whether the current position of
the traverse
window defines a valid portion of a face. If so, save the current rectangular
frame position of
the traverse window as results.
[113] f. After traversing the entire integral image, increase the width and
the length of the
traverse window by 1.1 times and repeat step e until the size of the traverse
window exceeds
the size of the image, or the buffer allocated for saving the results is used
up.
[114] g. Return to face frame position and faces
[115] Eyes Detection Module 506
[116] Inputs: Acquired frontal face image (grey image), face frame positions,
classifier for
both left and right eyes, left eye classifier, right eye classifier, left eye
coarse detection
classifier, right eye coarse detection classifier
[117] Outputs: frame position for both eyes, frame position of left eye, and
frame position
of right eye
[118] Flow:
[119] a. Obtain face image from the acquired frontal face image
[120] b. If user-defined classifier for both left and right eyes is available,
use
correspondingly defined face detection function to detect both the left and
right eyes of the
obtained face image. If not, estimate the positions of both the left and right
eyes based on
experience.
[121] c. If user-defined left/right eye course detection classifier for the
left/right eye is
available, detect the left/right eye on the corresponding half of the obtained
face image.
Further, based on the coarse detection result, determine whether the detected
human subject
is wearing glasses or not. If glasses are present, detect the obtained face
image and return
with results. If no glasses are present, continue to detect the obtained face
image based on
the coarse detection result and return the detection result without
considering the presence of
glasses. (If user-defined classifier for glasses-wearing subject is not
available, detect the
obtained face image without considering the presence of glasses.)
[122] d. If user-defined course detection classifiers are not available,
determine whether
glasses are present by directly detecting the left/right half of the obtained
face image. If
glasses are present, detect the obtained face image and return with results.
If no glasses are
present, continue to detect the obtained face image based on the coarse
detection result and
return the detection result without considering the presence of glasses. (If
user-defined
classifier for glasses-wearing subject is not available, detect the obtained
face image without
considering the presence of glasses.)
19
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[123] e. Return
[124] In some aspects, the processor 106 may further use, e.g., a recognition
module 508, to
extract pertinent facial features obtained from the detection module 502 for
comparing
against known characteristics and/or information of a number of authorized
people as
follows.
[125] Recognition Module 508
[126] Normalization
[127] Inputs: to-be-normalized image (grey image), the coordinates of the
centers of both
the left and right eyes on the image axis (the origin is located at the left
top corner of the
image). The meanings of parameters: lx refers to the x coordinate of the
center point of the
left eye (horizontal direction) in the output image divided by the width of
the output image,
and 1 y refers to x coordinate of the center point of the left eye (vertical
direction) in the
output image divided by the height of the output image.
[128] Output: output image
[129] Feature Extraction
[130] Inputs: Normalized image (grey image) and feature types
[131] Outputs: If output buffer is NULL, return feature dimensional degrees.
Otherwise,
assume the size of the output buffer equals the feature dimensional degrees,
write the features
of the image into the buffer, and return feature dimensional degrees. Certain
features are
associated with certain image size. For example, #6 feature may require the
image size of 100
by 100. Therefore, when the input image fails corresponding defined image size
requirement,
a result of zero can be returned.
[132] Feature Comparison
[133] Inputs: Two features to be compared and the comparison method
[134] Output: The smaller the comparison result (a floating point), the higher
the similarity.
[135] Obtaining Algorithm Information
[136] Function: instruct the requester to correctly assign parameters for the
algorithm
[137] Input: algorithm type based on the usage context
[138] Outputs: parameters information of the algorithm including feature type,
feature
dimensional degrees, normalized image size, the minimum distance, suggested
range, and
distance type.
[139] Many of the systems and methods described above can be used to create
Biometric
Signature Data ("BSD") files that allow a system to identify and distinguish
requesters with a
high degree of accuracy. Various implementations of the present disclosure may
employ the
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
BSD files to create an encryption/decryption key, thus increasing the security
of such keys.
Examples of the present disclosure can generate asymmetric keys based on one
or more BSD
files in such a way that by utilizing a biometric sensor, a person's biometric
measurement can
act as the person's private key. Implementations of the present disclosure may
also
incorporate BSD files into digital rights management (DRM) security in such a
way that files
cannot be decrypted or accessed by anyone other than the requester or group of
requesters
intended, or encrypted in a way that the original owners, such as a business,
can no longer
access the files. Accordingly, by using implementations of the present
disclosure employing
BSD files, when a file is accessed, there can be assurance of the identity of
the requester who
accessed the file.
[140] BSD files can be generated by the algorithmic analysis of data from an
AID IR and/or
UV sensor. Accordingly, many of these elements can be considered when
constructing the
private key of the asymmetrical pair (i.e., analog and/or digital values).
Thus, in some
implementations of the present disclosure, multiple elements of a sensor can
contribute real-
time data or real-time analog data related to a recognition event in order to
de-encrypt, thus
ensuring a real-time event (i.e., the actual measurement of the intended
person) has triggered
the authentication.
[141] As shown in FIGS. 12-13, in accordance with some implementations of the
present
disclosure, messages can be sent as follows. A requester can register, e.g.,
on a computer,
and create a public key for the requester. The requester than then publish the
public key so
that the key is publicly known. Other people, systems, or entities, can use
the requester's
public key to encrypt messages for the requester and send those messages to
the requester.
The requester can decrypt the message using her private key created by one or
more live BSD
files associated with the requester. Accordingly, the sender of the message is
ensured that the
requester is actually the person decrypting the message because the private
key used to
decrypt the message can be generated by the requester's live biometric data.
These systems
and methods for encryption provide substantial advantages over conventional
systems and
methods. For example, instead, of simply matching anonymous asymmetrical
codes, by using
BSD filed in the encryption process, authentication because inherent in the
key itself
[142] Various implementations of the present disclosure can also improve DRM.
For
example, DRM rules can allow for additional content to be added to a file and
additional
rules to be required. DRM rules can be expressed in many rights management
languages
known in the art, including but not limited to, XrML (extensible rights markup
language),
XMCL (extensible media commerce language), ODRL (open digital rights
language), and the
21
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
like. Rules can specify the actions that are permitted (e.g., decrypting,
encrypting,
transferring, copying, editing, etc.). The rules can also specify the people
authorized to
perform actions and the conditions under which these actions are permitted.
BSD files can be
used to authenticate a requester to determine whether the requester is one of
the people
specified in the rules.
[143] Various systems and methods for biometric encryption and authentication
can also
find application in corporate settings where, e.g., employees use corporate
devices for
personal use as well as business, or when, e.g., an employee uses a personal
device and the
corporate digital assets are transferred to and from the personal device. By
applying rules to
documents that have certain digital signatures, there may be controllable
segmentation
between private and business concerns. Both parties may have access to the
parts they are
entitled to access but can be prevented from accessing parts that are not
entitled to access.
For example, possible applications include, but are not limited to, providing
remote access,
making purchases, and conditional security.
[144] In the case of remote access, various implementations of the present
disclosure can
generate BSD files used to authenticate a requester, thus providing secure
access for any
remote network connection, i.e., VPN server, secure access to network email,
and/or
company proprietary information, from a remote device.
[145] Additionally, biometric authentication techniques of the present
invention can be used
to make authenticated online purchases/transactions. For example, spending
limits can be
based on requester or group profile for an account. In order for a requester
to make a
purchase, a system can use the biometric authentication techniques of the
present disclosure
to authenticate the true identity of that requester to verify the requester is
entitled to make the
desired purchase.
[146] Biometric authentication techniques can also be used to provide
conditional security
to various digital files. For example, files that contain sensitive
information can only be
accessed by authorized requesters, which can be authenticated using the
requester's live BSD
files.
[147] Biometric Encryption and Authentication Application to Digital Cinema
[148] The biometric encryption and authentication techniques described herein
find many
applications in the digital cinema industry. Movies are popular commodities,
especially pre-
DVD release. In order to maximize both production efficiencies and
distribution
opportunities, movies need to be accessed and handled by many different strata
of requesters.
Persons skilled in the art appreciate that techniques capable of protecting
digital assets in the
22
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
digital cinema industry can be used to protect digital assets in almost any
industry.
Accordingly, the principles described herein are not limited to application in
the digital
cinema industry, but may instead be applied to any industry for a similar
purpose.
[149] Digital cinema security views itself as an end-to-end process from
production via
distribution to consumption. SMPTE DC28, the body responsible for digital
cinema
standards, has identified five separate areas of digital cinema: (1) capture;
(2) production; (3)
Master (cinema, home, video, trailers, test screenings); (4) distribution
(satellite, fiber,
packaged); and (5) exhibition (digital projector security). In each area
identified by SMPTE
DC28, a movie is vulnerable to theft. In order to discourage theft, movies can
be encrypted
prior to distribution. Movies are then typically stored in their encrypted
state in the theater
until showtime. At showtime, the movie is decrypted and decompressed. This
decryption/decompression may take place in a server or in a projector.
[150] In an exemplary SMPTE DC28 process, DC28.4 can represent the conditional
access
portions of the cinema delivery system. Modem DRM encryption methods have
proven
sufficient to withstand unwarranted deciphering attempts, but securing the
keys has become a
problem. From capture to exhibition to distribution, a movie is encrypted and
decrypted
multiple times. Accordingly, various biometric encryption and authentication
techniques
discussed herein can be applied to one of more of the encryption, decryption,
and
authentication steps, in accordance with various implementations of the
present disclosure.
[151] Referring to FIG. 14, an example of an environment 1400 for managing
data may rely
on a first gateway 1402a, a second gateway 1402b, and a third gateway 1402c to
route data
via wired and/or wireless communication links. The first gateway 1402a may be
implemented as a software-based gateway virtualized in a computer system. The
second
gateway 1402b may be a standalone device that routes data as described below.
The third
gateway 1402c may be a cloud-based gateway. Other architectures are possible.
[152] In certain implementations, the gateways 1402 may perform several
functions
including managing the movement of data to and from the biometric sensor as
described
below, providing a networked solution that efficiently moves binary facial
data between
devices, and when clustered together (physical and virtual), providing a high
availability
solution for security designs. The gateway 1402 may receive credentialing data
through an
XML file structure. By monitoring and actively consuming the XML files, the
gateway 1402
may be able to utilize a standardized universal interface agnostic to the
programming
language, operating system and type of connectivity of the data source,
support physical and
logical access control requirements within the same method, offer an interface
that supports
23
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
simultaneous connectivity from different system types, and/or support either
live or batch
processing of credentials, including the immediate recovery of a credential
system through
file replacement.
[153] In certain aspects, data stored in the gateways 1402 may be stored in
files based on
Javascript Object Notation format.
[154] The environment 1400 may include one or more ECD ECDs 1404, e.g., ECD-a
1404a, ECD-b 1404b, ECD-c 1404c, ECD-d 1404d, ECD-e 1404e, and ECD-f 1404f.
The
ECDs 1404 may also be referred to as edge capture devices. The ECD ECDs 1404
may
transmit data to the gateways 1402 to be routed to another device within the
environment
1400. For example, the ECD-a 1404a may send a biometric template and/or
identity
information of a person to the third gateway 1402c. The third gateway 1402c
may send one
or more biometric templates to the ECD-a 1404a for performing matching
operations.
[155] Still referring to FIG. 14, in certain implementations, the environment
1400 may
include an access control server 1406a, an enterprise server 1406b, and a
third party server
1406c. The access control server 1406a may be communicatively coupled to one
or more
access controlled entry points 1408 via a wired or wireless communication
link. The
enterprise server 1406b may be communicatively coupled to a storage device
1410. The
storage device 1410 may be a network drive, local hard drive, flash drive,
tape drive, or other
suitable storage media.
[156] Still referring to FIG. 14, during operations for example, the ECD-b
1404b may
receive a biometric template of a first requester 1450a. The first requester
1450a may attempt
to access one or more access controlled entry points 1408 controlled by the
access control
server 1406a. The one or more access controlled entry points 1408 may include
a vault, lock,
secure door, secure gate, equipment or machinery, computing device, digital
storage device,
database, or file, for example. In some examples, the one or more access
controlled entry
points 1408 may be an access controlled door or gate of an infrastructure,
such as a
warehouse, office building, restricted area, etc. The biometric template of
the first requester
1450a may include one or more of the fingerprints, voice patterns, iris
patterns, facial
features, signature patterns, shapes of the ears, retinal patterns, gait, hand
geometry of the
first requester 1450a.
[157] In certain implementations, the ECD-b 1404b may extract the facial
features of the
first requester 1450a, and compare the facial features with the facial
features of authorized
personnel. If the facial feature of the first requester 1450a matches one of
the facial features
of the authorized personnel, the ECD-b 1404b may send a first positive match
signal to the
24
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
third gateway 1402c. The third gateway 1402c may route the first positive
match signal to
the access control server 1406a. Upon receiving the first positive match
signal, the access
control server 1406a may unlock one of the one or more access controlled entry
points 1408
associated with the ECD-b 1404b to allow the requester 1450 access. If the
facial feature of
the first requester 1450a does not match one of the facial features of the
authorized personnel,
the first requester 1450a may not gain access to the entry point 1408
associated with the
ECD-b 1404b.
[158] Still referring to FIG. 14, in certain examples, the ECD-d 1404d may
transmit a
second positive match signal of a second requester 1450b at a first time and a
third positive
match signal at a second time of the second requester 1450b to the second
gateway 1402b.
The second gateway 1402b may route the second positive match signal and the
third positive
match signal to the first gateway1402a, which may route the second and third
positive match
signals to the enterprise server 1406b. The enterprise server 1406b may use
the second
positive match signal and the third positive match signal to log access
information associated
with the second requester 1450b. For example, the enterprise server 1406b may
record, into
the storage device 1410, the first time as the arrival time of the second
requester 1450b and
the second time as the departure time on a work day. In another example, the
enterprise
server 1406b may record the first and second time as the number of accesses to
the one or
more access controlled entry points 1408 by the second requester 1450b. The
enterprise
server 1406b may also log, based on the information in the second and third
positive match
signals, the premises, equipment, files, locations, and information accessed
by the second
requester 1450b.
[159] Referring still to FIG. 14, in certain implementations, the ECD-f 1404f
may transmit a
fourth positive match signal of the third requester 1450c to the first gateway
1402a. The first
gateway 1402a may route the fourth positive match signal to the third party
server 1406c
through a firewall 1412. The firewall 1412 may filter information transmitted
through the
firewall 1412 and prevent malicious requesters from gaining unauthorized
access. The fourth
positive match signal may indicate to the third party server 1406c that the
third requester
1450c gained access to the one or more access controlled entry points 1408.
For example, the
fourth positive match signal may indicate to the third part server 1406c that
the third
requester 1450c accessed a software that requires payment to the owner of the
third party
server 1406c.
[160] Still referring to FIG. 14, in certain examples, a network administrator
1452 may
install, manage, update, maintain, and/or control the software in the ECD ECDs
1404, the
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
gateways 1402, the access control server 1406a, the one or more access
controlled entry
points 1408, the enterprise server 1406b, the storage device 1410, and/or the
firewall 1412
via a workstation 1414. The workstation 1414 may be a desktop computer, laptop
computer,
tablet computer, handheld computer, smartphone, or other suitable computer
devices
communicatively coupled via a wired or wireless connection to the third
gateway 1402c. The
network administrator 1452 may transmit software commands from the workstation
1414 to
the third gateway 1402c to be routed to a destination. In some examples, the
network
administrator 1452 may upgrade the firmware in the ECD ECDs 1404. In other
examples,
the network administrator 1452 may install new software onto the access
control server
1406a. In another example, the network administrator 1452 may perform
maintenance
operations, such as disk error check and defragmentation, on the storage
device 1410. In yet
another example, the network administrator 1452 may lock down or open the one
or more
access controlled entry points 1408 in an emergency.
[161] Referring to FIG. 14, in some implementations, an employee 1454, such as
a
supervisor, may utilize a requester terminal 1416 to access information
through the second
gateway 1402b. The requester terminal 1416 may be a desktop computer, laptop
computer,
tablet computer, handheld computer, smartphone, or other suitable computer
devices
communicatively coupled via a wired or wireless connection to the second
gateway 1402b.
In some examples, the employee 1454 may download information, such as work
hours,
arrival time, access history, utilization frequencies, using the requester
terminal 1416.
[162] In certain examples, data exchanges within the environment 1400 may be
encrypted.
Data transmissions between the gateways 1402 and the ECD ECDs 1404 may use
advanced
TLS v1.2 communications with a proprietary key management framework. Data
transmitted
via TLS v1.2 communications may be fully encrypted to remove the threat of
exposure of the
data to unwanted parties. The encryption of the data may be further protected
by protecting
the generation of the encrypted keys through the use of the biometric data as
the seed for the
generation of the keys. As such, data exchanged within the environment may be
difficult to
access by unauthorized requesters.
[163] Further, the gateways 1402 may be used to manage the data and the
creation of a
blockchain for the requesters. The first gateway 1402a, the second gateway
1402b, and the
third gateway 1402c may each include a blockchain wallet. The wallets will
contain the
requester personal credentials required to authenticate to any device or
application. The
wallets may be tied to the gateways 1402 to provide cybersecurity monitoring
and to provide
the interaction between the wallet and the facial recognition devices. The ECD
ECDs 1404
26
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
linked to the personal blockchain may be able to enable a transaction in the
blockchain.
Communications between blockchain wallets, a ledger for the blockchain
transactions, and
the gateways 1402 may use the blockchain protocol. The personal blockchain
will also
extend to devices that verify more than one requester, like a bank ATM (not
shown). The
gateways 1402 will utilize location tracking to move the link binary facial
data and the link to
the blockchain ledger between shared devices to improve the security of the
transactions and
to manage the number of requesters held within each ECD 1404.
[164] Aspects of the present disclosure may be implemented using hardware,
software, a
cloud network, or a combination thereof and may be implemented in one or more
computer
systems or other processing systems. In an aspect of the present disclosure,
features are
directed toward one or more computer systems capable of carrying out the
functionality
described herein. An example of such the computer system 1500 is shown in FIG.
15. One
or more of the gateways 1402, the servers 1406, the firewall 1412, the
workstation 1414,
and/or the requester terminal 1416 may be implemented based on the computer
system 1500.
[165] Referring now to FIG. 15, the computer system 1500 includes one or more
processors,
such as the processor 1504. The processor 1504 is communicatively coupled to a
communication infrastructure 1506 (e.g., a communications bus, cross-over bar,
or network).
Various software aspects are described in terms of this example computer
system. After
reading this description, it will become apparent to a person skilled in the
relevant art(s) how
to implement aspects of the disclosure using other computer systems and/or
architectures.
[166] The computer system 1500 may include a display interface 1502 that
forwards
graphics, text, and other data from the communication infrastructure 1506 (or
from a frame
buffer not shown) for display on a display unit 1530. Computer system 1500
also includes a
main memory 208, preferably random access memory (RAM), and may also include a
secondary memory 1510. The secondary memory 1510 may include, for example, a
hard
disk drive 1512, and/or a removable storage drive 1514, representing a floppy
disk drive,
magnetic tape drive, optical disk drive, universal serial bus (USB) flash
drive, etc. The
removable storage drive 1514 reads from and/or writes to a removable storage
unit 1518 in a
well-known manner. Removable storage unit 1518 represents a floppy disk,
magnetic tape,
optical disk, USB flash drive etc., which is read by and written to removable
storage drive
1514. As will be appreciated, the removable storage unit 1518 includes a
computer usable
storage medium having stored therein computer software and/or data.
[167] Alternative aspects of the present disclosure may include secondary
memory 1510
and may include other similar devices for allowing computer programs or other
instructions
27
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
to be loaded into computer system 1500. Such devices may include, for example,
a
removable storage unit 1522 and an interface 1520. Examples of such may
include a
program cartridge and cartridge interface (such as that found in video game
devices), a
removable memory chip (such as an erasable programmable read only memory
(EPROM), or
programmable read only memory (PROM)) and associated socket, and other
removable
storage units 1522 and interfaces 1520, which allow software and data to be
transferred from
the removable storage unit 1522 to computer system 1500.
[168] Computer system 1500 may also include a communications interface 1524.
Communications interface 1524 allows software and data to be transferred
between computer
system 1500 and external devices. Examples of communications interface 1524
may include
a modem, a network interface (such as an Ethernet card), a communications
port, a Personal
Computer Memory Card International Association (PCMCIA) slot and card, etc.
Software
and data transferred via communications interface 1524 are in the form of
signals 1528,
which may be electronic, electromagnetic, optical or other signals capable of
being received
by communications interface 1524. These signals 1528 are provided to
communications
interface 1524 via a communications path (e.g., channel) 1526. This path 1526
carries signals
1528 and may be implemented using one or more of a wire or cable, fiber
optics, telephone
line, cellular link, RF link and/or other communications channels. In this
document, the
terms "computer program medium" and "computer usable medium" are used to refer
generally to media such as a removable storage drive 1518, a hard disk
installed in hard disk
drive 1512, and signals 1528. These computer program products provide software
to the
computer system 1500. Aspects of the present disclosure are directed to such
computer
program products.
[169] Computer programs (also referred to as computer control logic) are
stored in main
memory 1508 and/or secondary memory 1510. Computer programs may also be
received via
communications interface 1524. Such computer programs, when executed, enable
the
computer system 1500 to perform the features in accordance with aspects of the
present
disclosure, as discussed herein. In particular, the computer programs, when
executed, enable
the processor 1504 to perform the features in accordance with aspects of the
present
disclosure. Accordingly, such computer programs represent controllers of the
computer
system 1500.
[170] In an aspect of the present disclosure where the method is implemented
using
software, the software may be stored in a computer program product and loaded
into
computer system 1500 using removable storage drive 1514, hard drive 1512, or
28
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
communications interface 1520. The control logic (software), when executed by
the
processor 1504, causes the processor 1504 to perform the functions described
herein. In
another aspect of the present disclosure, the system is implemented primarily
in hardware
using, for example, hardware components, such as application specific
integrated circuits
(ASICs). Implementation of the hardware state machine so as to perform the
functions
described herein will be apparent to persons skilled in the relevant art(s).
[171] FIG. 16 illustrates a block diagram of various example system
components, in
accordance with an aspect of the present disclosure. Fig. 16 shows a
communication system
1600 usable in accordance with aspects of the present disclosure. The
communication system
1600 includes one or more accessors 1660, 1662 and one or more terminals 1642,
1666. In
one aspect, data for use in accordance with aspects of the present disclosure
is, for example,
input and/or accessed by accessors 1660, 1662 via terminals 1642, 1666, such
as personal
computers (PCs), minicomputers, mainframe computers, microcomputers,
telephonic devices,
or wireless devices, such as personal digital assistants ("PDAs") or a hand-
held wireless
devices coupled to a server 1643, such as a PC, minicomputer, mainframe
computer,
microcomputer, or other device having a processor and a repository for data
and/or
connection to a repository for data, via, for example, a network 1644, such as
the Internet or
an intranet, and couplings 1645, 1646, 1664. The couplings 1645, 1646, 1664
include, for
example, wired, wireless, and/or fiberoptic links. In another example
variation, the method
and system in accordance with aspects of the present disclosure operate in a
stand-alone
environment, such as on a single terminal.
[172] Turning now to FIG. 17, an example of a ECD 1404 that is configured to
perform
access control may analyze a biometric template of a requester 1450 to
determine whether the
requester 1450 is authorized to gain access to an entry point (not shown)
associated with the
ECD 1404. The ECD 1404 may include an optical sensor 1702, an illumination
source 1704,
a display 1706, a keypad 1708, and a scanner 1710.
[173] In some implementations, the optical sensor 1702 may be configured to
capture still
or moving images. For example, the optical sensor 1702 may capture the
fingerprints, the iris
patterns, the facial features, the signature patterns, the shapes of the ears,
the retinal patterns,
the gait, and/or the hand geometry of the requester 1450. In some examples,
the optical
sensor 1702 may be a broadband camera configured to detect electromagnetic
radiation
having wavelengths ranging from 200 nanometers (e.g., soft UV) to 2000
nanometers (e.g.,
near infra-red (NIR)). In a particular example, the optical sensor 1702 is
configured to detect
29
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
radiation between 700 to 900 nanometers. In certain implementations, the
optical sensor
1702 may include a motion sensor configured to detect people approaching the
ECD 1404.
[174] In certain examples, the optical sensor 1702 may include a wide angle
lens (e.g., such
as a fisheye lens) used to provide both vertical area coverage to capture
faces across the full
range of human heights (as well as addressing American's With Disabilities Act
requirements) and to provide horizontal coverage of the complete area of an
access point to
address more than one person accessing the secure area on one authentication.
In other
examples, the optical sensor 1702 may employ a high resolution (e.g.,
megapixel per square
inch) charge coupled device (CCD) array. The high resolution array may provide
the ability
to identify faces at a greater distance from the sensor. The ECD 1404 may take
advantage of
the increased distance of identification to pre-identify requesters in queuing
situations.
Potential requesters may be identified as they enter a queuing area and a
placed in a priority
list to accelerate the confirmation at the access controlled entry point and
discharge of the
queue. The pre-identification and prioritizing of individuals in the
identification process at
the access point enables high volume throughput at the entry point by reducing
the
identification time during the identification and authentication process.
[175] In certain implementations, the illumination source 1704 may emit
electromagnetic
radiation having wavelengths ranging from 200 nanometers to 2000 nanometers.
In certain
examples, the illumination source 1704 may emit non-visible radiation between
700 to 900
nanometers and/or 200-300 nanometers. The illumination source 1704 may emit
radiation to
illuminate bodily features and patterns of the requester 1450 used for
biometric analysis
(analyzing biometric template to determine access rights). The emitted
radiation may
impinge on a portion of a body of the requester 1450, and reflect off of the
portion of the
body. The reflected radiation may be captured by the optical sensor 1702 for
biometric
analysis.
[176] In some implementations, the display 1706 may present useful information
to the
requester 1450. For example, the display 1706 may show one or more images of a
face 1730
of the requester 1450, captured by the optical sensor 1702, to assist the
requester 1450 in
aligning the face 1730 during biometric analysis. In another example, the
display 1706 may
notify the requester 1450 a status of the entry point associated with the ECD
1404 (e.g.,
locked down, temporarily unavailable, normal operations, under maintenance).
In yet
another example, the display 1706 may display information such as time, date,
weather,
current location, etc.
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[177] Still referring to FIG. 17, the keypad 1708 may allow the requester 1450
to enter
numbers, symbols, and alphabets into the ECD 1404. In an example, the
requester 1450 may
enter a password in addition to the biometric analysis to gain access to the
entry point.
[178] In some implementations, the scanner 1710 may be a radio frequency
identification
(RFID) scanner, a proximity card scanner (e.g., HIDTM card scanner), a contact
card scanner,
or a magnetic card scanner. In an example, the scanner 1710 may send an
interrogatory
signal to a proximity card (not shown) having a coil and an integrated circuit
with a
programmable or non-programmable identification sequence. The interrogatory
signal may
be "absorbed" by the coil and may energize the integrated circuit. In
response, the energized
integrated circuit sends a response signal including the identification
sequence back to the
scanner 1710 via the coil. The scanner 1710, in turn, analyzes the
identification sequence to
determine whether or not to grant access. The identification sequence may be
one or more
numbers, alphabets, symbols, and/or a combination thereof.
[179] Still referring to FIG. 17, in an implementation, the requester 1450 may
approach the
ECD 1404 during operations. The optical sensor 1702 may detect the requester
1450. In
response to the detection, the illumination source 1704 may emit incident NIR
radiation 1760
toward the requester 1450. The incident NIR radiation 1760 may impinge on the
face 1730
of the requester 1450, and reflect off of the face 1730 of the requester 1450.
In some
implementations, the optical sensor 1702 may detect detected NIR radiation
1762 originating
from the surface of the face 1730. The detected NIR radiation 1762 may include
reflected
incident NIR radiation 1760 and/or NIR radiation emitted from the requestor
1450 due to
thermal heating (i.e., black body radiation). The intensity and distribution
of the detected
NIR radiation 1762 may depend on the intensity and angle of the incident NIR
radiation
1760, the contour of the face 1730, angle of detection by the optical sensor
1702, and other
factors. The ECD 1404 may use the detected NIR radiation 1762 to construct a
facial
template (the "NIR sampled profile") of the requester 1450. The ECD 1404 may
compare the
constructed NIR sampled profile with existing templates stored therein
(details described
below). If the ECD 1404 detects a match, the ECD 1404 may allow the requester
1450
access to the entry point (as described above).
[180] In some examples, a NIR sampled profile generated via NIR radiation
detection may
be resistant to changes in ambient lighting. As ambient lighting fluctuates
(e.g., changes in
luminance, color, color temperature, lighting angle), a NIR sampled profile
constructed using
NIR radiation detection may remain sufficiently constant to prevent a false
acceptance or a
false rejection. For example, the NIR sampled profile of the requester 1450
constructed via
31
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
NIR radiation detection under a "bright" condition (e.g., 1000 lux) may be
substantially
identical to the NIR sampled profile constructed via NIR radiation detection
under a "dark"
condition (e.g., 100 lux). In another example, the NIR sampled profile of the
requester 1450
constructed via NIR radiation detection under substantially blue light (e.g.,
CIE coordinates
x=0.153, y=0.100) may be substantially identical to the NIR sampled profile
constructed via
NIR radiation detection under substantially white light (e.g., CIE coordinates
x=0.30,
y=0.33).
[181] Still referring to FIGs. 14 and 17, in other implementations, a NIR
sampled profile
generated via NIR radiation detection may improve the privacy of the owner.
For example,
the storage device 1410 may store NIR sampled profiles of employees and
associated
confidential information (e.g., birthdates, email account passwords). If an
unauthorized
person gains access to the NIR sampled profiles and the associated
confidential information,
the unauthorized person may not be able to exploit the stolen confidential
information
because it may be difficult to identify the employees based on the NIR sampled
profiles.
Given that the NIR sampled profiles are constructed using, for example,
detected NIR
radiation, they may be unrecognizable because the NIR sampled profile of a
person may be
drastically different from the visual image of the face of the same person.
[182] In another implementation, the requester 1450 may approach the ECD 1404
during
operations. The optical sensor 1702 may detect the requester 1450. In response
to the
detection, the illumination source 1704 may emit incident UV radiation 1760
(e.g.,
electromagnetic radiations having wavelengths between 315-390 nanometers)
toward the
requester 1450. The incident UV radiation 1760 may impinge on the face 1730 of
the
requester 1450, and reflect off of the face 1730 of the requester 1450. In
certain examples,
the incident UV radiation 1760 may penetrate the surface of the face 1730 and
reflect off of
subdermal features of the face 1730 (e.g., dermis, subcutaneous tissue,
muscles,
imperfections). In some implementations, the optical sensor 1702 may detect
detected UV
radiation 1762 originating from the surface and/or subdermal features of the
face 1730. The
detected UV radiation 1762 may include reflected incident UV radiation 1760.
The intensity
and distribution of the detected UV radiation 1762 may depend on the intensity
and angle of
the incident UV radiation 1760, the contour of the face 1730, angle of
detection by the optical
sensor 1702, and other factors. The ECD 1404 may use the detected UV radiation
1762 to
construct a facial template (the "UV sampled profile") of the requester 1450.
The ECD 1404
may compare the constructed UV sampled profile with existing templates stored
therein
32
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
(details described below). If the ECD 1404 detects a match, the ECD 1404 may
allow the
requester 1450 access to the entry point (as described above).
[183] In some examples, a UV sampled profile generated via UV radiation
detection may be
resistant to changes in ambient lighting. As ambient lighting fluctuates
(e.g., changes in
luminance, color, color temperature, lighting angle), a UV sampled profile
constructed using
UV radiation detection may remain sufficiently constant to prevent a false
acceptance or a
false rejection. For example, the UV sampled profile of the requester 1450
constructed via
UV radiation detection under a "bright" condition (e.g., 1000 lux) may be
substantially
identical to the UV sampled profile constructed via UV radiation detection
under a "dark"
condition (e.g., 100 lux). In another example, the UV sampled profile of the
requester 1450
constructed via UV radiation detection under substantially blue light (e.g.,
CIE coordinates
x=0.153, y=0.100) may be substantially identical to the UV sampled profile
constructed via
NIR radiation detection under substantially white light (e.g., CIE coordinates
x=0.30,
y=0.33).
[184] Still referring to FIGs. 14 and 17, in other implementations, a UV
sampled profile
generated via UV radiation detection may improve the privacy of the owner. For
example,
the storage device 1410 may store UV sampled profiles of employees and
associated
confidential information (e.g., birthdates, email account passwords). If an
unauthorized
person gains access to the UV sampled profiles and the associated confidential
information,
the unauthorized person may not be able to exploit the stolen confidential
information
because it may be difficult to identify the employees based on the UV sampled
profiles.
Given that the UV sampled profiles are constructed using, for example,
detected UV
radiation, they may be unrecognizable because the UV sampled profile of a
person may be
drastically different from the visual image of the face of the same person. A
sampled profile
may be a rendering of the dataset that visualizes the consistency of position
within the three
arrays. If the sampled profile is compromised, it may be more difficult to
obtain biometric
information used to identify the individual.
[185] In some examples, the ECD 1404 may generate a visible-light sampled
profile based
on detected visible-light reflected from the face 1730 of the requester 1450.
[186] In certain implementations, the requestor 1450 may be asked to provide a
password, a
personal identification number (PIN), and/or a valid HIDTM card to be used in
conjunction
with the constructed sampled profile to gain access to the entry point
associated with the
ECD 1404.
33
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[187] In other implementations, the ECD 1404 may include a microphone (not
shown in
FIG. 17) to perform voice recognition.
[188] In some examples, the display 1706 may show the face 1730 of the
requester 1450 as
imaged by the optical sensor 1702. The display 1706 may include alignment
marks (not
shown) to assist the requester 1450 in aligning the face 1730 with respect to
the optical sensor
1702. This alignment process may minimize false acceptances and false
rejections due to
misalignment. In certain implementations, the ECD 1404 may be a pocket-sized
mobile
device powered by rechargeable batteries.
[189] Referring to FIG. 18, in some implementations, the ECD 1404 may include
a
processor 1802 having a communication module 1852 configured to communicate
with the
gateways 1402 and other ECD ECDs 1404 as described in this disclosure. The
communication module 1852 may be implemented as hardware in the processor 1802
for
example, as software code executed by the processor 1802, or a combination
thereof. The
processor 1802 may also include a security module 1854 configured to encrypt
and/or
decrypt data. The security module 1854 may be implemented as hardware in the
processor
1802 for example, as software code executed by the processor 1802, or a
combination
thereof The processor 1802 further includes an algorithm module 1856 for
constructing and
comparing biometric templates as described throughout this disclosure.
Alternatively, the
communications with the Gateway may be facilitated by a processor in the
control panel.
The algorithm module 1856 may be implemented as hardware in the processor 1802
for
example, as software code executed by the processor 1802, or a combination
thereof. The
processor 1802 may further include a parallel computation module 1858 for
performing
distributed processing. The parallel computation module 1858 may be
implemented as
hardware in the processor 1802 for example, as software code executed by the
processor
1802, or a combination thereof The processor 1802 may include one or more
processors or
cores, and may be implemented as a semiconductor processor, graphical
processing unit, a
field programmable gate array, a programmable logic device, a processing
cluster, an
application specific integrated circuit, or other suitable architectures.
[190] The ECD 1404 includes a memory 1804. The memory may be static or dynamic
memory such as flash memory, random access memory, magnetic memory, or
semiconductor
memory. The memory 1804 may include external memory such as a cloud storage.
The
memory 1804 may include or store applications and/or computer executable code.
The ECD
1404 further includes a modem 1808 for communicating with the gateways 1402
and other
ECD ECDs 1404, and may operate in cooperation with the communication module
1852.
34
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
The ECD 1404 also includes a RAM 1806, such as static or dynamic random access
memory
(RAM). The ECD 1404 may also include an Input/Output (I/O) device 1810
communicatively coupled to the display 1706, the optical sensor 1702, the
keypad 1708, and
the scanner 1710. The components within the ECD 1404 may be interconnected by
an
internal bus 1812a. The processor 1802, the memory 1804, the RAM 1806, and the
internal
bus 1812a may be disposed on a processing board 1820.
[191] Referring now to FIG. 19, another example of the ECD 1404 may include a
number
of processing boards 1820 interconnected by an external bus 1812b. Each
processing board
1820 may include the processor 1802, the memory 1804, the RAM 1806, and the
internal bus
1812a. Data distributed among the processing boards 1820 may be distributed by
a controller
1814 via the external bus 1812b. In a non-limiting example, the ECD 1404 may
download
(from the gateways 1402) 240,000 biometric templates of potential requestors.
The controller
1814 may distribute the 240,000 biometric templates evenly or unevenly among
the
processing boards 1820 such that each processing board 1820 may store, in the
respective
memory 1804, 30,000 biometric templates. In some implementations, each
processing board
1820 may store the same or different number of biometric templates.
[192] Still referring to FIG. 19, in an example, during operations, the ECD
1404 may
construct a sampled profile (e.g., UV or NIR) of the requestor 1450 based on
the detected
NIR radiation 1762. The controller 1814 may distribute copies of the
constructed sampled
profile to each of the processing boards 1820. In some implementations, the
processing
boards 1820 may simultaneously compare the duplicated sampled profiles with
the locally
stored biometric templates (e.g., 30,000 stored in each processing board
1820). While the
current example of the ECD 1404 shown in FIG 19 includes eight processing
boards 1820,
other numbers of processing boards 1820 may also be used. For example, the ECD
1404 may
include 2, 4, 6, 8, 12, 16, 32 or 64 processing boards 1820.
[193] In an implementation, the ECD 1404 may rely on remote processing boards
(not
shown) to perform the distributed computing described above. For example, the
ECD 1404
may send the duplicated sampled profiles to the remote processing boards to
jointly and
simultaneously implement the algorithm (described below) for matching the
duplicated
sampled profiles (or the numerical representation of the duplicated sampled
profiles) to
known biometric templates. The processing boards may be within other ECD ECDs
within
the network. In certain examples, the ECD 1404 may send the duplicated sampled
profiles to
a Beowulf cluster. The clustering design may employ inter-process and inter-
processor
protocols to share processing tasks of the same application between both
processors and the
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
processing cores within those processors. In the case of facial recognition,
processor-
intensive operations like facial verification may utilize multiple processors
or multiple cores
to complete the operation within an acceptable period of time. In some cases,
large multi-
core processors may be used for this type of operation. In other cases, the
operation may be
spread across several advanced RIISC machine ("ARM") processors to accomplish
the same
performance as the single multi-core processor without the high hardware cost
and potential
for a single point of failure.
[194] Referring back to FIG. 14, in a non-limiting example of distributed
parallel
comparison, the first gateway 1402a may download 600,000 biometric templates
of potential
requesters (e.g., employees and contractors) from the enterprise server 1406b.
The first
gateway 1402a may distribute 100,000 biometric templates to each of the ECD-e
1404e and
the ECD-f 1404f, and 200,000 biometric templates to each of the second gateway
1402b and
the third gateway 1402c. Next, the second gateway 1402b may distribute 100,000
biometric
templates to each of the ECD-c 1404c and the ECD-d 1404d, and the third
gateway 1402c
may distribute 100,000 biometric templates to each of the ECD-a 1404a and the
ECD-b
1404b. Consequently, the 600,000 biometric templates downloaded from the
enterprise
server 1406b may be evenly distributed among the ECD ECDs 1404 (i.e., 100,000
non-
overlapping templates each). In some implementations, when the ECD-b 1404b
constructs a
sampled profile (UV or NIR) of the first requester 1450a, the ECD-b 1404b may
duplicate the
constructed sampled profile and distribute the duplicated sampled profiles to
the ECD-a
1404a, ECD-c 1404c, ECD-d 1404d, ECD-e 1404e, and ECD-f 1404f (via one or more
gateways 1402). The ECD ECDs 1404 may compare the duplicated sampled profile
of the
first requester 1450a with the 100,000 templates stored locally to determine a
match. The
ECD-a 1404a, ECD-c 1404c, ECD-d 1404d, ECD-e 1404e, and ECD-f 1404f may send
(via
one or more gateways 1402) the result of the comparison back to the ECD-b
1404b. The
ECD-b 1404b may gather the results and determine whether to grant access to
the first
requester 1450a. In other implementations, the biometric templates may be
unevenly
distributed among the ECD ECDs 1404.
[195] Referring now to FIG. 20, the ECD 1404 may generate a sampled profile
2000 based
on measuring the intensity of the detected radiation 1762 from the face 1730
of the requester
1450. The sampled profile 2000 may include a base matrix 2010 of one or more
measurement points 2002. The measurement points 2002 may include a value
indicative of
the intensity of the detected radiation 1762 at a location of the particular
measurement point.
For example, the measurement point-a 2002a may indicate an intensity scale
value of 2
36
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
corresponding to a region 2004a (e.g., black hair) around the face 1730 of the
requester 1450.
The measurement point-b 2002b may indicate an intensity scale value of 50 that
corresponds
to a background color. The measurement point-c 2002c may indicate an intensity
scale value
of 21 that corresponds to a region 2004c (e.g., cheeks) on the face 1730 of
the requester 1450.
In certain examples, the intensity scale may range from 0 (no reflection) to
100 (maximum
reflection able to be detected by the optical sensor 1702 of the ECD 1404. In
an
implementation, the intensity scale may measure an absolute intensity (e.g.,
brightness) of the
measurement points 2002. In other examples, the sampled profile 2000 may
include
measurement points 2002 for detected radiation 1762 of different wavelengths
(e.g., UV,
NIR, red, green, blue).
[196] While the sampled profile 2000 in FIG. 20 shows the base matrix 2010 of
10 x 12
measurement points 2002 across the face 1730 of the requester 1450, other
measurement
points density may be possible for the base matrix 2010. For example, the ECD
1404 may
generate another sampled profile using 100 x 100 measurement points across the
face 1730 of
the requester 1450. In another example, the ECD 1404 may generate a sampled
profile using
500 x 500 measurement points. Other measurement points density are also
possible, and may
depend on desired accuracy, computational constraints, amount of data storage,
etc.
[197] In some implementations, the ECD 1404 may remove measurement points 2002
indicating the background.
[198] Turning now to FIG. 21, in one implementation, the ECD 1404 (or the
processor
1802) may apply the LBP operation to one or more of the measurement points
2002 within
the base matrix 2010. For example, the ECD 1404 may perform LBP on the
measurement
point-d 2002d, which includes an intensity value of 41. The LBP string for the
measurement
point-d 2002d is 01110100, based on the span of 1 (i.e., performing LBP using
the immediate
neighbors, having distance of 1 cell, of the measurement point-d 2002). The
ECD 1404 may
track the numbers of LBP strings for the remaining measurement points 2002r as
shown in a
table 2100. In some examples, the ECD 1404 may compute 24 of the measurement
points
2002 as having the LBP string of 00000000, 3 having the LBP string of
00000001, 11 having
the LBP string of 00000010, and 0 having the LBP string of 00000011, etc., as
indicated in a
table 2100. The table 2100 may include 256 entries for the possible strings
(i.e., 8 bits). In
some examples, the table 2100 may include 255 entries by eliminating the entry
for the "all
white" string (11111111) or the "all black" string (00000000). In some
implementations, the
data in the table 2100 may be plotted as a histogram indicating the number of
occurrences
(e.g., measurement points 2002) for the possible LBP strings.
37
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[199] In some implementations, based on the distribution of the LBP strings in
the table
2100, the ECD 1404 may determine one or more unique features. The one or more
unique
features may be non-zero LBP strings occurring fewer than a threshold
frequency (e.g., 5),
such as LBP strings 00000001, 01110100, and 11111100. In another
implementations, the
one or more unique features may be the non-zero LBP strings occurring the
least, such as
LBP strings 01110100. In some examples, the ECD 1404 may track the location
the
measurement points 2002 having the one or more unique features. For example,
the ECD
1404 may track the coordinate of the LBP strings. In some implementations, the
one or more
unique features may be the non-zero LBP strings occurring the most.
[200] Turning now FIG. 22, in certain implementations, the ECD 1404 may divide
the base
matrix 2010 into one or more sub-matrices. Each sub-matrix may include a
macroblock of
measurement points. For example, the ECD 1404 may divide the base matrix 2010
into a 5 x
sub-matrix 2202, a 3 x 3 sub-matrix 2204, and a 6 x 6 sub-matrix 2206. When
computing
the LBP strings for the measurement points 2002 within the 5 x 5 sub-matrix
2202, the ECD
1402 may calculate the LBP strings of the measurement points 2002 within the 5
x 5 sub-
matrix 2202 using a span of 3 (i.e., calculating LBP strings using neighbors 3
cells away).
Similarly, when computing the LBP strings for the measurement points 2002
within the 3 x 3
sub-matrix 2204, the ECD 1402 may calculate the LBP strings of the measurement
points
2002 within the 3 x 3 sub-matrix 2204 using a span of 2. When computing the
LBP strings
for the measurement points 2002 within the 6 x 6 sub-matrix 2206, the ECD 1402
may
calculate the LBP strings of the measurement points 2002 within the 6 x 6 sub-
matrix 2206
using a span of 4. Other sizes for the base matrix, the sub-matrices, and
spans are possible, as
determined by the ECD 1402.
[201] In one example, a 100 x 100 base matrix may be divided into sub-matrices
of six
different sizes: 10 x 10, 15 x 15, 20 x 20, 30 x 30, 35 x 35 and 40 x 40.
Within each of these
sub-matrices, the LBP string may be calculated with a different span around
the measurement
point being calculated to characterize the texture/slope of the area
surrounding the cell at
different coverage areas. The span in pixels from the measurement point 2002
to be
calculated to each of the neighboring cells for each of the sub-matrix sizes
may be 3, 9, 15,
21, 27 and 33. In another example, a 200 x 200 base matrix may include 6
different sub-
matrices having sizes of 10 x 10, 20 x 20, 35 x 35, 50 x 50, 65 x 65, 80 x 80,
and 100 x 100.
The span in pixels from the measurement point 2002 to be calculated to each of
the
neighboring cells for each of the sub-matrix sizes may be 3, 5, 7, 10, 25 and
40. The sub-
38
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
matrices may overlap in some instances. In certain implementations, the ECD
1404 may
apply LTP computations onto the measurement points 2002.
[202] Referring now to FIG. 23, in certain examples, the ECD 1404 may convert
the list of
binary features derived from the sampled profile into a data sequence that is
extensible and
capable of being tailored to the unique traits of each requester 1450 while
still providing a
methodology of object comparison and matching/verification. A binary feature
may contain
three characteristics: the size of the sub-matrix/macroblock, the location of
the sub-
matrix/macroblock on the sampled profile, and the Uniform Local Binary Pattern
(ULBP)
assigned to binary feature. A macroblock may be a "sub-image" of the image of
the face
1730 of the requester 1450. For instance, a macroblock of size 10 is a 10 x 10
(i.e., 100
measurement points 2002) sub-image of the image of the face 1730. In a non-
limiting
example, the location of the macroblock on the image may be defined by the
position of the
first pixel in the macroblock located in the top, left corner of the
macroblock. The position of
this pixel is defined by two values, its distance from the left boundary of
the face image (x-
dimension) and the distance from the top boundary of the face image (y-
dimension). The
final characteristic is the ULBP, which is a mathematical methodology for
establishing a
scalar value for the edge and texture characteristics of a pixel starting with
the top left of the
face 1730. For each pixel within the macroblock, a ULBP calculation is
performed. If the
calculated ULBP matches the defined ULBP for that binary feature, the value of
the binary
feature is incremented by one. Therefore, the maximum value of a binary
feature is the size
of the defined macroblock (10x10 is 100) and the minimum value is zero. A
macroblock may
be used in more than one binary feature with a different ULBP definitions. The
location of a
macroblock may also be used in more than one binary feature with a different
macroblock
size and ULBP definition.
[203] Still referring to FIG. 23, a table 2300 may illustrate an example of
the results of the
sequence conversion process for macroblocks having sizes of 10 x 10, 35 x 35,
20 x 20, 30 x
30, 15 x 15, 40 x 40, and 15 x 15. Three arrays may be used to generate a
biometric template.
First array may be a scalar value in array normalized. Second array may be
sorted so that
binary feature 2000, which in this example has the highest value, goes to the
top of the
second array in position 1. Third array compares the difference in the
position of binary
feature 2000 in the sort instead of scalar value. If the position within the
array for binary
feature 2000 remains the same, the results will reflect a value of 0
representing the least
possible change in position and therefore the highest value for significance
in the
authentication method.
39
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[204] Turning now to FIG. 24, a method 2400 of converting a sequence may
identify the
most unique features used for identification.
[205] At block 2402, obtain the sampled profile of the face. For example, the
processor
1802 and/or the communication module 1825 may obtain the sampled profile of
the face
1730.
[206] At block 2404, assign a unique index to each macroblock/ULBP
combination. For
example, the algorithm module 1856 may assign a unique index (e.g., 4) to each
macroblock/ULBP combination (e.g., macroblock 30 x 30) and analyzing based
upon
uniqueness of traits within the face and not on scalar values.
[207] At block 2406, construct a first array of scalar values for each
macroblock/ULBP
combination in the master schema referenced by the schema. For example, the
algorithm
module 1856 may construct a first array of scalar values for each
macroblock/ULBP
combination in the master schema referenced by the schema. First array
includes number of
pixels that fall within that ULBP within that macro-block. When analyzing the
LBP of a
pixel, the ECD 1404 may perform normal LBP calculation and get a histogram
that is the
value from 0-255, and if the result is the highest value (standard LBP formula
that takes pixel
and takes 8 pixels around it and then each pixel has a different binary value)
then the result
receives a value of 1. If some other LBP is higher than that pixel has a 0
value.
[208] At block 2408, weigh each scalar value by the size of the macroblock.
For example,
the algorithm module 1856 may weigh each scalar value (e.g., 380) by the size
of the
macroblock (e.g., 30 x 30).
[209] At block 2410, sort the first array of scalar values in descending
order. For example,
the algorithm module 1856 may sort the elements of the first array of scalar
values in
descending order from, for example, 1 to 2165.
[210] At block 2412, convert the associated indices into a sequence. For
example, the
algorithm module 1856 may convert the associated indices into a sequence or
the sort order
of the data set instead of the scalar value.
[211] At block 2414, convert the sequence into a second array with a scalar
value for each
unique index that is the distance (difference in position in array rather than
measurement in
mm) of the primary index from the beginning of the sequence array. For
example, the
algorithm module 1856 may convert the sequence into a second array with a
scalar value for
each unique index that is the position of the primary index from the beginning
of the
sequence array with 2165 elements currently but may be extensible depending
upon data for
each individual.
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[212] Referring to FIG. 25, in certain implementations, a deep learning may be
performed
on the full data set to establish the minimum data set required to accurately
perform facial
recognition. The result of the deep learning may be a reduced data set of
binary features that
is a fraction of the total data set (e.g., 1%, 2%, 5%, 10%, 20%, 50%,). Each
element of data
in this data set may be defined by a sub-matrix of a specific size and
location and a single
value within the full LBP base matrix. Each data element may be assigned a
unique numeric
identifier. For the enrollment process, this static data set is used for each
requester. As each
requester continues to verify their faces, the verification data may be
collected and through
deep learning a new set of binary features may be created for each requester.
The
introduction of the sequence to the verification algorithm may allow the
introduction of this
individualized set of binary features and still allow for comparison between
different sets.
The matching of two sequences may be achieved by calculating the difference
(e.g., the
difference in position of a data set) of the two arrays. The greater the sum
the lower the
quality of the match. A perfect object match may have a sum of substantially
zero.
[213] Still referring to FIG. 25, a table 2500 illustrates an example of the
verification
calculations. The first step in creating the sequence may be the sorting of
the scalar array by
the magnitude of the scalar value. The sort may be performed in descending
order. The
binary feature unique identification with the largest value is at the top of
the sort. By sorting
in this manner, the binary features providing with the most uniqueness
(greatest scalar
magnitude) are at the top of the array. After the sort, the new sort position
is transferred to
that third array and the location on the new array is based upon the binary
feature unique
identification. Binary feature unique identifier is 2000 when it gets sorted
it goes to position
1. In third array the 2000 position will receive a value of 1. This array may
be a unique
sequence for the object and is the new basis for object verification and/or
matching.
[214] The sequence matching algorithm determines the quality of the match by
the distance
of each binary feature unique identification from the beginning of the
sequence. The distance
is the index value of the binary feature unique identification within the
sequence array where
the index values are sequentially assigned. To efficiently complete this
matching calculation,
another array is created that contains the distance of each binary feature
from the beginning
of the sequence. As an example, after the sort, the binary feature with the
unique
identification of one is in position ten of the sorted sequence. In the new
array, which is zero-
indexed, position one will have a value of ten. By structuring the array in
this manner, the
algorithm to perform a match between two of these arrays is the same as the
original
algorithm. The absolute value of the difference of the values with the
matching index is
41
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
aggregated to create a single value for the array match. A lower value equates
to a closer
match. A perfect match may be substantially a zero value.
[215] By introducing the sequence array, the matching algorithm may be
extensible.
Sequence array is second array and scalar array is the first array (normalized
for block size).
Extensible may indicate that the fixed list of binary features may now be
expanded and
contracted as required to optimize the process of matching objects while also
improving the
matching process integrity. In this implementation, the sequence array may be
dynamically
adjustable based on two separate deep learning functions based upon the
environment of the
edge device and the specific individual face being authorized. The first
function may seek to
continuously optimize the default list of binary features applied to some or
all object
matching attempts prior to the development of an individual list of binary
features. The
second function may aggregate image data on individual objects and over time
develop an
optimized list of binary features for the face. Once the individualized list
of binary features
exceeds the matching performance of the default list of binary features
through parallel object
matching trials on incoming object image data, the default list of binary
features may be
replaced with the individualized list of binary features for that object in
the live object
matching functions.
[216] In certain implementations, the deep learning engine for both processes
may receive
sample data in parallel with the live object/face matching functions. As image
data is
received, two lists of binary features may be applied to it. The first list
may either be the
default or individualized list of reduced binary features. (e.g., three most
important blocks out
of 100 in 10 x 10 macroblock). The output of the application of this list may
be used in the
live matching process. The second list may be the full list of binary
features, which
comprises all possible macroblock size (e.g., all 100 within the 10 x 10
macroblock), applied
to possible image positions for each ULBP. The complete list of binary
features may be, for
example, twenty times the size of the default or individualized list of binary
features. The
deep learning engine may receive a sequence derived from the complete list of
binary
features for each object image received into the system. In the case of the
default list of
binary features deep learning process, the images may be categorized into the
training and
validating sets. An independent default test set may be created including
objects different
than the object being learned. In the case of the individualized list of
binary features, each
object identified by the default or existing individualized list of binary
features will be placed
in the training, validation and test sets. All other object's list may be
placed in the
verification and test sets.
42
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
[217] A deep learning training session for both the default object list of
binary features and
for each individual list of binary features may be executed for each new live
entry from the
object detection system.
[218] Once the eyes detection algorithm locates the eyes, the face matrix of
data is
transferred to the facial recognition algorithm. If the deep learning
algorithm has developed
specific list of binary features, then an aggregate list of these binary
features for all users in
the system may be created and used to generate the sequence. Otherwise, the
sequence may
be generated using the default set of binary features. The sequence may be
transferred to the
verification algorithm where the matching process may determine the identity
of the
sequence. With the identity established, a full list of binary features (all
sub-matrices with
full LBP histograms) may be generated and transferred to that identities data
set in the deep
learning algorithm. The deep learning algorithm may process the available data
set for that
identity (the set may continue to grow with each verification) and generate a
revised optimum
data set for that identity. The optimized data set may be converted to a
sequence and used in
the next verification process for that user. At the same time the aggregate
set of binary
features may be updated if necessary for converting the next face matrix.
While the
verification process will be within the locale cluster of devices, the deep
learning algorithm
may occur within both the locale and global clusters as a background task.
[219] As the results of the training session refine the list of binary
features, the respective
list may be updated in the live verification/matching process. The result may
be a matching
system capable of automatically adjusting to both the overall object
population and the
specific, unique traits of each object. The evolution of the object data will
allow for large
scale object matching solutions in excess of 100,000 objects capable of the
same precision as
a small solution (<1,000).
[220] Referring now to FIG. 26, an example of the deep learning process 2600
may rely on
feedback loops and machine learning to refine the identification process.
[221] At block 2602, obtain an image matrix. For example, the processor 1802
and/or the
communication module 1825 may obtain the sampled profile of the face 1730.
[222] At block 2604, determine if the face specific macroblock/ULBP is
available. For
example, the algorithm module 1856 may determine if the face specific
macroblock/ULBP is
available.
[223] At block 2606, if the custom list of binary features is not available,
convert image to
face generic detection list of binary features. For example, the algorithm
module 1856 may
43
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
convert image to face generic/default detection list of binary features if the
custom list is not
available.
[224] At block 2608, obtain face detection default sequence. For example, the
algorithm
module 1856 may obtain face detection default sequence if the macroblock is
not available.
[225] At block 2610, if the custom list of binary features is available,
convert image to face
detection specific list of binary features. For example, the algorithm module
1856 may
convert image to face detection specific list of binary features if the custom
list of binary
features (could be in original 2165 or could be all new ones that based upon
that individual
(e.g., 10 additional 10 x 10s that are more distinctive for that particular
individual) is
available.
[226] At block 2612, perform face detection. For example, the algorithm module
1856 may
perform face detection.
[227] At block 2614, perform eye detection and location. For example, the
algorithm
module 1856 may perform eye detection and location.
[228] At block 2616, convert image to full list of binary features. For
example, the
algorithm module 1856 may convert image to full list of binary features. (not
just 57, but all
100 of 10 x 10 of macroblocks and include all values). Tensor flow is the deep
learning
engine framework provided by GoogleTM. Equation and algorithm for tensor flow
is well
known in the art.
[229] At block 2618, develop face detection list of binary features using
tensor flow deep
learning. For example, the algorithm module 1856 may develop face detection
list of binary
features using tensor flow deep learning.
[230] At block 2620, feedback face detection refined sequence. For example,
the algorithm
module 1856 may feedback face detection refined sequence. To refine the list
of features, the
ECD 1404 may take entire feature set and collect all data for face and derive
a new refined
set that replaces the default set. Distance calculation between the position
of the binary
feature within two sets may be used as metrics. Every feature would be in the
same position
every time. As face data get better face to face comparison will get smaller
and the
difference between this face and all others will get larger.
[231] Turning now to FIG. 27, the example of the deep learning process 2600
shown in
FIG. 26 may rely on feedback loops and machine learning to refine the
identification process.
Rather than centralizing process of deep learning data that is subject to
hacking if stored in a
central server (tensor flow single application operating on a single central
server) particularly
on the cloud, the present disclosure may use tensor flow in a distributed
architecture to
44
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
evaluate at each reader that is privacy protected on all individual devices
rather than stored in
one central server.
[232] At block 2614, perform eye detection and location. For example, the
algorithm
module 1856 may perform eye detection and location.
[233] At block 2702, obtain face detection specific list of binary features.
For example, the
algorithm module 1856 may obtain face detection specific list of binary
features if the
macroblock is available.
[234] At block 2704, obtain facial recognition default list of binary
features. For example,
the algorithm module 1856 may convert image to face generic detection list of
binary features
if the specific list of binary features for that individual is not available.
[235] At block 2706, determine if the face specific list of binary features is
available. For
example, the algorithm module 1856 may determine if the face specific list of
binary features
is available. Custom list of binary features may include the original 2165 or
could be all new
ones that are based upon that individual (e.g., 10 additional 10 x lOs that
are more distinctive
for that particular individual).
[236] At block 2708, if the face specific list of binary features is not
available, convert
image to face generic list of binary features. For example, the algorithm
module 1856 may
convert image to face generic list of binary features if the face specific
list of binary features
is not available.
[237] At block 2710, if the face specific list of binary features is
available, convert image to
face specific list of binary features. For example, the algorithm module 1856
may convert
image to face specific list of binary features if the face specific list of
binary features is
available.
[238] At block 2712, convert list of binary features to sequences. For
example, the
algorithm module 1856 may convert list of binary features to sequences.
[239] At block 2714, perform verification. For example, the algorithm module
1856 may
perform verification.
[240] At block 2716, convert image to full list of binary features. For
example, the
algorithm module 1856 may convert image to full list of binary features.
[241] At block 2718, develop face specific list of binary features using
tensor flow deep
learning. For example, the algorithm module 1856 may develop face specific
list of binary
features using tensor flow deep learning. (not just 57 as is typical with LBP,
but all 100 of 10
x 10 of macroblocks and include all values). Tensor flow is the deep learning
engine
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
framework provided by GoogleTM. Equation and algorithm for tensor flow is well
known in
the art.
[242] In some examples, the gateway 1402 and/or the ECD 1404 may include
proactive
algorithms to identify and reduce "consistency-collisions" during the
verification process. A
"consistency-collisions" may occur when one or more requesters have unique
facial data that
causes a mistake (i.e., false acceptance or false reject) in the verification.
The gateway 1402
and/or the ECD 1404 may seek out potential 'consistency-collisions'
proactively using two
methods of the present disclosure described below.
[243] The 'Layered Reinforcement' algorithm may allow the requester to create
a binary
tree structure for the facial data. The binary data may create a hierarchy of
data points (e.g.,
LBP string) based on the uniqueness of the data points. The greater the
uniqueness the higher
up on the tree. Every verification transaction may result in the transfer of
the binary facial
data from the ECD 1404 to the gateway 1402. When the gateway 1402 receives the
binary
facial data, it may begin checking the uniqueness of the new data's binary
tree structure
against the data of the other requesters 1450 stored. When it identifies a
potential collision,
the gateway 1402 may notify the ECD 1404 to escalate verification transactions
between the
two identified requesters 1450 to the gateway 1404. When the gateway 1404
receives an
escalated transaction it may perform two advanced algorithms (i.e., the
Proactive Collision
Identification Algorithm and the Time-Domain Trending Algorithm) on the data.
Binary tree
may take most distinctive features and establish hierarchy based upon most and
least
distinctive. Prioritized tree may reduced analysis based on 20 rather than
2165 saving time
and computing/processing. Based upon the branches of the tree off the most
distinctive
features may allow the ECD 1404 to determine if this is or is not the person
rather than going
thru all 2165 features. Threshholds may be empirical and derived from testing.
[244] The first advanced algorithm, the Proactive Collision Identification
Algorithm, may
take the binary tree data and analyze the facial data based on its location in
the binary tree.
The binary tree data may be weighted. If the weighting of the binary data does
not yield a
sufficient differentiation of the data, the gateway 1402 may extend the
verification process to
find adequate differentiation data.
[245] In some implementations, the facial characteristics of requesters may
continuously
change. If the biometric data of a face, such as the face 1730 of the
requester 1450, remains
static, the identification of the face may ultimately result in false
rejections, requiring the re-
enrollment of the face 1730. Dynamic adjustments to the biometric data may be
continuously
applied and algorithms may determine whether the proposed change is not the
introduction of
46
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
a different face into the biometric data resulting in a false acceptance on
the face 17303.
With every iteration of the biometric data, the algorithm must determine the
amount of
biometric data to retain to insure the next successful identification of the
requester 1450 and
the amount to change to insure the long-term identification of the requester
1450. The
determination may be made by a combination of a time-domain analysis where the
changes
are regressed to insure linearity and facial regional analysis to determine if
the area of change
is rational.
[246] The use of this dynamic data capability leads to the second algorithm,
the Time-
Domain Trending Algorithm. The gateway 1402 may maintain the history of the
binary data
of the requester 1450 and perform a time-domain based analysis of the data to
assess what
features are changing when the history of the binary data of requester 1450 is
compared with
face 1730 and the speed with which these features are changing. The gateway
1402 may
track the evolution of the face and uses this data to extrapolate and
reinforce the unique
differences identified in the face 1730 of the requester 1450 to establish and
later emphasize
core features or baseline features and assess the rate of change in these core
features for the
face 1730. The gateway 1402 then determines using time-domain based analysis
whether
these changes are taking place over a time period to suggest natural changes
in the face 1730,
artificial changes in face 1730 warranting further analysis, or a mismatch of
the history of
binary data of the requester 1450 with the face 1730. The gateway 1402 may
identify the
unique differences and verify that the differences identified in the
verification request are
consistent with the trending over time. In some implementations, the gateway
1402 may
factor in lifestyle and daily routines of the requester 1450 in the Time-
Domain Trending
Algorithm. For example, if the requester 1450 enjoys outdoor activities, the
Algorithm may
factor in increased tanning during the summer season. The Proactive Collision
Identification
Algorithm combined with the Time-Domain Trending Algorithm allows the method
of this
embodiment of the invention to maintain its high performance in a large
population '1:N'
solution.
[247] In some implementations, the Time-Domain Trending Algorithm may reduce
the
change over time in the biometric data into one or more equations that
characterize the
change over time (e.g., curve fitting algorithms). The linearity of the one or
more equations
over time may determine the integrity of the changes. If a particular area of
the change is
represented as a discrete function, the change may be flagged as a potential
threat (e.g.,
disguise, incorrect match). The changes may also be evaluated based on the
physical location
of the pixel box on the face. Locations may be weighted based on the
probability of change
47
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
in that area of the face. Significant changes in low probability change areas
will be flagged
as a potential threat.
[248] Turning now to FIG. 28, referencing figures above, in some
implementations, a
method 2800 of constructing a biometric template may be performed by the ECD
1404.
[249] In optional implementations, the method 2800 may perform an enrollment
process.
The enrollment process may include the ECD 1404 capturing a plurality of
images (e.g., 5,
10, 15, 20, 30, 50) of the face 1730. The ECD 1404 may convert the plurality
of images to
biometric data as described above.
[250] At block 2802, emitting an incident non-visible light. For example, the
illumination
source 1704 may emit an incident non-visible light. The non-visible light may
include near
IR or UV. In some implementations, the illumination source 1704 may emit a
visible light.
[251] At block 2804, detecting a detected non-visible light, wherein the
detected non-visible
light includes a reflected non-visible light and a radiated non-visible light.
For example, the
optical sensor 1702 may detect a detected non-visible light, wherein the
detected non-visible
light includes a reflected non-visible light and a radiated non-visible light.
In some
implementations, the optical sensor 1702 may detect IR light reflected off of
the face 1730 of
the requester 1450 and/or IR light radiated due to the heat of the requester
1450.
[252] At block 2806, generating a sampled profile including a plurality of
sampling points
having a plurality of characteristic values associated with the detected non-
visible light. For
example, the algorithm module 1856 may generate a sampled profile, such as the
sampled
profile 2000, including a plurality of sampling points having a plurality of
characteristic
values associated with the detected non-visible light.
[253] At block 2808, identifying one or more macroblocks each includes a
subset of the
plurality of sampling points. For example, the algorithm module 1856 may
identify one or
more macroblocks each includes a subset of the plurality of sampling points.
The one or
more macroblocks may be a 10 x 10 macroblock, a 15 x 15 macroblock, a 20 x 20
macroblock, a 30 x 30 macroblock, a 35 x 35 macroblock, and a 40 x 40
macroblock. Other
sizes are possible. In some implementations, the ECD 1404 may identify 2165
macroblocks
having 2165 associated dimensions.
[254] At block 2810, selecting a local pattern value. For example, the
algorithm module
1856 may select a local binary pattern value of 20. In other examples, the
algorithm module
1856 may select a local ternary pattern value.
[255] At block 2812, calculating a number of occurrences of the local pattern
value within
each subset of the plurality of the sampling points for each of the one or
more macroblocks.
48
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
For example, the algorithm module 1856 may calculate a number of occurrences
of the local
pattern value within each subset of the plurality of the sampling points for
each of the one or
more macroblocks.
[256] At block 2814, generating a first array including a plurality of
weighted values by
calculating the plurality of weighted values based on the numbers of
occurrences of the local
pattern value and corresponding sizes of the one or more macroblocks. For
example, the
algorithm module 1856 may generate a first array including the weighted values
shown in the
table 2300 (i.e., [0.30 0.54 0.15 0.42 0.62 0.46 0.53]).
[257] At block 2816, assigning a unique index to each of the plurality of
weighted values.
For example, the algorithm module 1856 may assign unique indices (i.e., 1, 2,
3, 4, 5, 6, and
7) to each of the plurality of weighted values shown in the table 2300. The
unique index 1 is
assigned to the weighted value of 0.30, 2 to 0.54, 3 to 0.15 , etc.
[258] At block 2818, generating a second array of the unique index by ranking
the plurality
of weighted values. For example, the algorithm module 1856 may generate a
second array of
the unique index by ranking the plurality of weighted values, such as the
sequence 5, 2, 7, 6,
4, 1, 3 in the table 2300. In some implementations, the second array/sequence
may indicate a
ranking of the weighted values from the highest to the lowest. The first
number in the
sequence is 5 because the weighted value (i.e., 0.62) associated with the
unique index of 5 is
the highest among the elements of the first array.
[259] At block 2820, generating a third array including a plurality of ranking
distances. For
example, the algorithm module 1856 may generate a third array including a
plurality of
ranking distances, such as the stored array [5 1 6 4 0 3 2] in the table 2300.
A ranking
distance may indicate a numerical difference between the ranks of the highest
weighted value
(e.g., 0.62 - rank 1) and the current weighted value (e.g., 0.30 ¨ rank 6).
Therefore, the
ranking distance between 0.62 and 0.30 may be 5 (i.e., 6-1).
[260] In some optional implementations, during the verification process, the
biometric data
based on the sampled profile (the "requestor biometric data") may be compared
to the
biometric data of the plurality of images captured during the enrollment
process (the
"enrollment biometric data"). If the matching percentage exceeds a threshold
percentage
(e.g., 20, 30, 40, 50, 60, 70, or 80), the ECD 1404 and/or one of the gateways
1402 may
determine that the requestor biometric data is a positive match and the
verification is
successful.
[261] In certain aspects, the ECD 1404 and/or one of the gateways 1402 may
adjust the
enrollment biometric data over time to accommodate any changes to the face
1703 due to, for
49
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
example, sun tan, aging, injuries, mood change, weight change, facial hair
change, cosmetics
usage, accessories, or other causes. During each verification, the ECD 1404
and/or one of the
gateways 1402 may adjust a portion (e.g., 20%, 30%, 40%, 50%) of the
enrollment biometric
data. Another portion of the biometric data may remain unchanged.
[262] It will be appreciated that various implementations of the above-
disclosed and other
features and functions, or alternatives or varieties thereof, may be desirably
combined into
many other different systems or applications by one ordinarily skilled in the
art. Also that
various presently unforeseen or unanticipated alternatives, modifications,
variations, or
improvements therein may be subsequently made by those skilled in the art
which are also
intended to be encompassed by the following claims.
[263] The above detailed description set forth above in connection with the
appended
drawings describes examples and does not represent the only examples that may
be
implemented or that are within the scope of the claims. The term "example,"
when used in
this description, means "serving as an example, instance, or illustration,"
and not "preferred"
or "advantageous over other examples." The detailed description includes
specific details for
the purpose of providing an understanding of the described techniques. These
techniques,
however, may be practiced without these specific details. For example, changes
may be
made in the function and arrangement of elements discussed without departing
from the
scope of the disclosure. Also, various examples may omit, substitute, or add
various
procedures or components as appropriate. For instance, the methods described
may be
performed in an order different from that described, and various steps may be
added, omitted,
or combined. Also, features described with respect to some examples may be
combined in
other examples. In some instances, well-known structures and apparatuses are
shown in block
diagram form in order to avoid obscuring the concepts of the described
examples.
[264] Information and signals may be represented using any of a variety of
different
technologies and techniques. For example, data, instructions, commands,
information,
signals, bits, symbols, and chips that may be referenced throughout the above
description
may be represented by voltages, currents, electromagnetic waves, magnetic
fields or particles,
optical fields or particles, computer-executable code or instructions stored
on a computer-
readable medium, or any combination thereof
[265] The various illustrative blocks and components described in connection
with the
disclosure herein may be implemented or performed with a specially-programmed
device,
such as but not limited to a processor, a digital signal processor (DSP), an
ASIC, a FPGA or
other programmable logic device, a discrete gate or transistor logic, a
discrete hardware
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
component, or any combination thereof designed to perform the functions
described herein.
A specially-programmed processor may be a microprocessor, but in the
alternative, the
processor may be any conventional processor, controller, microcontroller, or
state machine.
A specially-programmed processor may also be implemented as a combination of
computing
devices, e.g., a combination of a DSP and a microprocessor, multiple
microprocessors, one or
more microprocessors in conjunction with a DSP core, or any other such
configuration.
[266] The functions described herein may be implemented in hardware, software
executed
by a processor, firmware, or any combination thereof. If implemented in
software executed
by a processor, the functions may be stored on or transmitted over as one or
more instructions
or code on a non-transitory computer-readable medium.
Other examples and
implementations are within the scope and spirit of the disclosure and appended
claims. For
example, due to the nature of software, functions described above can be
implemented using
software executed by a specially programmed processor, hardware, firmware,
hardwiring, or
combinations of any of these. Features implementing functions may also be
physically
located at various positions, including being distributed such that portions
of functions are
implemented at different physical locations. Also, as used herein, including
in the claims,
"or" as used in a list of items prefaced by "at least one of' indicates a
disjunctive list such
that, for example, a list of "at least one of A, B, or C" means A or B or C or
AB or AC or BC
or ABC (i.e., A and B and C).
[267] Computer-readable media includes both computer storage media and
communication
media including any medium that facilitates transfer of a computer program
from one place to
another. A storage medium may be any available medium that can be accessed by
a general
purpose or special purpose computer. By way of example, and not limitation,
computer-
readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk
storage, magnetic disk storage or other magnetic storage devices, or any other
medium that
can be used to carry or store desired program code means in the form of
instructions or data
structures and that can be accessed by a general-purpose or special-purpose
computer, or a
general-purpose or special-purpose processor. Also, any connection is properly
termed a
computer-readable medium. For example, if the software is transmitted from a
website,
server, or other remote source using a coaxial cable, fiber optic cable,
twisted pair, digital
subscriber line (DSL), or wireless technologies such as infrared, radio, and
microwave, then
the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless
technologies such as
infrared, radio, and microwave are included in the definition of medium. Disk
and disc, as
used herein, include compact disc (CD), laser disc, optical disc, digital
versatile disc (DVD),
51
CA 03132721 2021-09-07
WO 2020/185525 PCT/US2020/021256
floppy disk and Blu-ray disc where disks usually reproduce data magnetically,
while discs
reproduce data optically with lasers. Combinations of the above are also
included within the
scope of computer-readable media.
52
