KR20210135315A - Method and device for facial recognition - Google Patents

Abstract

Aspects of the present disclosure include generating a sampled profile comprising a plurality of sampling points having a plurality of feature values associated with detected invisible light, identifying one or more macroblocks each comprising a subset of the plurality of sampling points, and , calculate the number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks, and calculate the number of occurrences of the local pattern value based on the number of occurrences of the local pattern value and the corresponding size of the one or more macroblocks. generating a first array comprising a plurality of weights by calculating the weights, assigning each of the plurality of weights a unique index, ranking the plurality of weights to create a second array of unique indices, and and a method for generating a third array comprising a rank distance.

Description

Method and device for facial recognition

This application is filed on September 9, 2013, claiming the benefit of priority to U.S. Provisional Application No. 61/792,922, filed March 15, 2013, and U.S. Provisional Application No. 61/698,347, filed September 7, 2012 Portion of U.S. Patent Application Serial No. 15/649,144, filed July 13, 2017 Serial Applicant, Partial Serial Applicant of U.S. Patent Application Serial No. 16/104,826, filed on August 17, 2018, claims the benefit of priority from U.S. Patent Application Serial No. 16/297,366, filed March 8, 2019, the content of which is expressly incorporated herein by reference in its entirety.

There is a growing need for stronger identity verification to protect both physical and electronic personal property. For example, it is important to control access to premises, vehicles and personal property so that only authorized requesters are allowed access. The requestor may be a user/person requesting access to access controlled assets and/or infrastructure. In a traditional example, the requestor may carry and use a key, which is designed to fit the lock to allow the requestor of the key to open and enter the lock. However, loss or damage to the key may render it inaccessible. In another example, the requestor uses an electronic key fob to generate an infrared (“IR”) or radio frequency (“RF”) signal that is detected by, for example, a sensor in a vehicle that controls a door. You can also remotely lock or unlock your vehicle's doors at the push of a button. Such vehicle keyless access systems may require the requestor to activate the ignition system. Other similar keyless access implementations may involve inserting and presenting a magnetic card or the like into a slot or card reader/detector, or enabling an authorized requestor to enter a numeric or alphanumeric code into a provided keypad. However, in each of these conventional techniques, it is very difficult to determine whether the person holding the key/card is the actual authorized requestor. A crook can steal or duplicate a valid key to gain unauthorized access to premises, vehicles and/or personal property.

Traditional biometrics access control systems may alleviate some of the shortcomings of key/card-based access control systems, but limitations may also exist. Traditional biometric sensors, such as iris detection sensors, can be limited in certain lighting conditions, significantly reducing the effectiveness of the biometric sensor as well as the environment in which it can be applied. The performance of biometric sensors may also degrade in direct sunlight due to glare, shadows and other artifacts. Even with the advent of megapixel 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 influences, lighting, background and camera angles collide to blur subject details, making it difficult to recognize heterogeneous faces (matching video and other probe images to large databases of frontal photos) .

Other factors may also increase the false acceptance rate and/or false recognition rate of traditional biometric sensors. For example, biometric sensors also have difficulty obtaining the necessary data in the absence of light. Light source shadowing and other intensity variations may create contrasts in the face that may be misinterpreted as facial features and/or may slightly distort measurements of actual facial features. Another major cause of inaccuracy is the increased increase in similar measured features between faces in a growing population. Further, the problem of capturing individual facial features can be exacerbated by the need for low maintenance and/or low complexity facial recognition systems. Thus, improvements in access control may be desired.

The following presents a brief summary of one or more aspects in order to provide a basic understanding of them. This summary is not an exhaustive overview of all contemplated aspects, and is intended to neither identify key or key 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.

Some aspects of the present disclosure generate a sampled profile comprising a plurality of sampling points having a plurality of characteristic values associated with detected non-visible light, each of the plurality of identify one or more macroblocks comprising a subset of the sampling points, select a local pattern value, count the number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks, and , generating a first array including a plurality of weights by calculating a plurality of weighted values based on the number of occurrences of the local pattern value and corresponding sizes of one or more macroblocks, and a unique index for each of the plurality of weights A method for allocating a unique index, creating a second array of unique indices by ranking a plurality of weights, and creating a third array comprising a plurality of ranking distances. include

Certain aspects of the present disclosure are operably to an illumination source configured to emit incident invisible light, an optical sensor configured to detect detected invisible light including reflected invisible light and emitted invisible light, an illumination source, and an optical sensor an edge capture device (ECD) having one or more processes coupled, the one or more processes generating a sampled profile comprising a plurality of sampling points having a plurality of feature values associated with detected invisible light, each identify one or more macroblocks comprising a subset of the plurality of sampling points, select a local pattern value, and number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks , creating a first array including a plurality of weights by calculating a plurality of weights based on the number of occurrences of the local pattern value and corresponding sizes of one or more macroblocks, and assigning a unique index to each of the plurality of weights biometric identification of a requestor requesting access to an entry point by generating a second array of unique indices by ranking the plurality of weights, and creating a third array including the plurality of ranking distances. configured to constitute a biometric template.

Aspects of the present disclosure provide code for, when executed by one or more processes, the one or more processes to generate a sampled profile comprising a plurality of sampling points having a plurality of feature values associated with detected invisible light, each of the plurality of code for identifying one or more macroblocks comprising a subset of sampling points of Code for calculating the number of occurrences, code for generating a first array including the plurality of weights by calculating the plurality of weights based on the number of occurrences of the local pattern value and corresponding sizes of the one or more macroblocks, the plurality of weights code for assigning a unique index to each, code for creating a second array of unique indices by ranking a plurality of weights, and code for creating a third array comprising a plurality of rank distances , including a computer readable medium having code thereon.

Aspects of the present disclosure provide means for generating a sampled profile comprising a plurality of sampling points having a plurality of feature values associated with detected invisible light, each one or more macroblocks comprising a subset of the plurality of sampling points. means for identifying, means for selecting a local pattern value, means for counting the number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks, the occurrence of the local pattern value means for generating a first array comprising a plurality of weights by calculating a plurality of weights based on a number of times and a corresponding size of the one or more macroblocks; means for assigning a unique index to each of the plurality of weights; A system comprising: means for generating a second array of unique indices by ranking the s; and means for generating a third array comprising a plurality of rank distances.

Aspects of the present disclosure include an infrastructure having an access-controlled entry point and an ECD, wherein the ECD emits invisible light incident on a face of a requestor; detect detected invisible light including invisible light reflected from the requestor's face and emitted invisible light; generate a sampled profile comprising a plurality of sampling points having a plurality of feature values associated with the detected invisible light, identify one or more macroblocks each comprising a subset of the plurality of sampling points, and determine the local pattern values select, calculate the number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks, and calculate a plurality of occurrences of the local pattern value based on the number of occurrences of the local pattern value and the corresponding size of the one or more macroblocks. generating a first array comprising a plurality of weights by calculating the weights of generate the requestor's biometric template by generating a third array comprising the rank distance; store a plurality of biometric templates of the authorized person; compare the authorized person's plurality of biometric templates with the requestor's biometric templates; generate a positive match signal in response to identifying a match between the requestor's biometric template and one of the authorized person's plurality of biometric templates; and send a positive match signal to the gateway to grant the requestor access to the entry point.

Features believed to be characteristic of aspects of the present disclosure are set forth in the appended claims. In the following description, like parts are denoted by like reference numerals, respectively, throughout the specification and drawings. Figures in the drawings are not necessarily drawn to scale, and specific features may be shown in exaggerated or generalized form for clarity and brevity. However, the present disclosure itself, as well as preferred modes of use, additional objects and advantages, will be best understood when read in conjunction with the accompanying drawings in conjunction with the following detailed description of exemplary aspects of the disclosure.
1 is an exemplary simultaneous real-time identification and authentication system, in accordance with some aspects of the present disclosure.
2 illustrates a perspective view of an example of a simultaneous real-time identification and authentication apparatus, in accordance with some aspects of the present disclosure.
3 illustrates a front view of an example of a simultaneous real-time identification and authentication apparatus, in accordance with some aspects of the present disclosure.
4 depicts another perspective view of an example of an exemplary simultaneous real-time identification and authentication apparatus, in accordance with some aspects of the present disclosure.
5 is a block diagram of exemplary processing components of an apparatus for simultaneous real-time identification and authentication, in accordance with some aspects of the present disclosure.
6 illustrates a flow diagram of a method for facial recognition, in accordance with some aspects of the present disclosure.
7A illustrates a facial image of a person, in accordance with some aspects of the present disclosure.
7B shows different facial images for the same person, in accordance with some aspects of the present disclosure.
8 depicts an example process for calculating a local binary pattern (LBP) feature, in accordance with some aspects of the present disclosure.
9 illustrates an example process for calculating a local ternary pattern (LTP) feature, in accordance with some aspects of the present disclosure.
10 illustrates positions of three example key features selected from one or more facial images, in accordance with some aspects of the present disclosure.
11 shows an example of a receiver operating characteristic (ROC) curve for testing a facial database, in accordance with some aspects of the present disclosure.
12 illustrates an example of biometric asymmetric encryption for confidentiality, in accordance with some aspects of the present disclosure.
13 illustrates another example of biometric asymmetric encryption for authentication, in accordance with some aspects of the present disclosure.
14 illustrates a schematic diagram of an example of an environment for implementing one or more gateways for access control.
15 illustrates an example of a computer system for implementing a method for managing data in accordance with aspects of the present disclosure.
16 illustrates a block diagram of various example system components, in accordance with aspects of the present disclosure.
17 illustrates an example of an ECD for identifying a biometric template, in accordance with aspects of the present disclosure.
18 illustrates examples of components of the ECD of FIG. 17 , in accordance with aspects of the present disclosure.
19 illustrates another example of components of the ECD of FIG. 17 , in accordance with aspects of the present disclosure.
20 illustrates an example of a sampled profile, in accordance with aspects of the present disclosure.
21 illustrates an example of LBP operation for a measurement point of the sampled profile of FIG. 20 , in accordance with aspects of the present disclosure.
22 illustrates an example of sub-matrices.
23 illustrates an example of a table of results of sequence transformation, according to an aspect of the present disclosure.
24 illustrates an example of a flowchart for transforming a sequence, in accordance with aspects of the present disclosure.
25 illustrates an example of a table of validation calculations, in accordance with aspects of the present disclosure.
26 illustrates an example of deep learning.
27 illustrates another example of deep learning.
28 illustrates a flow diagram of a method for identifying a biometric template.

The following includes definitions of selected terms as used herein. The definitions include various examples and/or forms of elements that fall within the scope of the term and that may be used for implementation. The examples are not intended to be limiting.

As used herein, a “processor” 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, bits, bit streams, or other computing capable of being received, transmitted, and/or detected.

A “bus,” as used herein, refers to an interconnected architecture communicatively coupled to transfer data between computer components within a single or multiple systems. The bus may be, inter alia, a memory bus, a memory controller, a peripheral bus, an external bus, a crossbar switch and/or a local bus. The bus can also be a vehicle bus that interconnects components inside the vehicle using protocols, such as, inter alia, Controller Area Network (CAN), Local Interconnect Network (LIN).

As used herein, “memory” may include volatile memory and/or non-volatile memory. Non-volatile memory may include, for example, read only memory (ROM), programmable read only memory (PROM), erasable PROM (EPROM), and electrically erasable PROM (EEPROM). Volatile memory includes, for example, random access memory (RAM), synchronous RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), and/or direct RAM bus RAM ( DRRAM) may be included.

As used herein and in the appended claims, the singular forms include plural referents unless the context clearly dictates otherwise.

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 ranges are expressed, other implementations include from the one particular value and/or to the other particular value.

"comprising" or "containing" or "including" means that at least the named compound, element, particle or method step is present in the composition or article or method, but other such compounds; Even though a material, particle, or method step has the same function as named, it is not meant to exclude the presence of another compound, material, particle, or method step.

It should also be understood that reference to one or more method steps does not exclude the presence of additional method steps or intervening method steps between those steps explicitly identified. Likewise, it should also be understood that reference to one or more components within an apparatus or system does not exclude the presence of additional components or intervening components between the components explicitly identified.

Biometric identification technology generally refers to a pattern recognition technique that performs the requestor identification process by determining the authenticity of certain physiological or behavioral characteristics possessed by the requestor. In some instances, biometric identification may be desirable over traditional methods involving passwords and personal identification numbers (PINs) for a variety of reasons. For example, when using biometric identification, the person to be identified (eg, the requestor) is traditionally required to be physically at a point-of-identification. Additionally, identification based on biometric technology eliminates the need to remember passwords or carry tokens (ie, security devices used to gain access to access-controlled entry points).

One kind of texture-based local binary pattern (“LBP”) feature describes facial information that produces desirable recognition results. The improved local tenery pattern (“LTP”) feature may be a better improvement over conventional LBP methods. Although LBP and LTP features may not be sensitive to changes in light and expression and are computationally efficient, they also have disadvantages such as information redundancy due to correlation between positive and negative histograms.

Thus, by utilizing various types of biometric data of an authorized requestor, generating biometric signature data for providing keyless access to vehicles, buildings, etc. to authorized requestors, with varying degrees of security. It is desirable to consider simultaneous real-time identification and authentication techniques for As noted above, in some implementations of the present disclosure, biometric signature data may be interchanged across a variety of applications. Accordingly, in some examples of this 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, examples of biometric systems in this disclosure allow biometric signature data to be altered based on a desired level of security. Accordingly, the type of biometric signature data that may be used in connection with a particular application and/or requestor may vary depending on the level of security required for that particular application and/or requestor. While some implementations discussed herein are discussed in the context of facial biometric data, those skilled in the art will recognize that various implementations of the present disclosure can be used for fingerprint data, iris and retina scan data, speech data, facial thermograms, hand It will be appreciated that many types of biometric data may be used, including, but not limited to, hand geometry data and the like.

In some implementations of the present disclosure, biometric data (eg, biometric template) associated with an intended recipient may be obtained via a biometric sensor of a biometric-based access control system. As discussed below, changes in light, temperature, and distance of the biometric sensor from the target may affect the amount and quality of biometric data obtained through the biometric sensor. For example, changes in light intensity and angle can create shadows on the requester's face, making facial recognition more difficult. If the biometric data to identify the requestor is ambiguous, more templates may be needed to properly authenticate the requestor, thus increasing the amount of biometric data needed. To reduce the undesirable effects of these environmental factors, biometric sensors can utilize near infrared (IR) or ultraviolet (UV) light, or a combination of both IR and UV of the desired intensity. In an implementation, the method uses near IR. Infrared light emitting diode (LED) arrays can be utilized in facial recognition devices or biometric sensors to minimize the effects of ambient light when capturing facial uniqueness. The cameras and LED arrays are packaged in dedicated edge devices (eg EDCs or faceplates) mounted in locations that require identification and/or identification/analysis, such as doors requiring access control.

In some implementations, the access control system may utilize IR or new IR illumination and detection to identify facial features. IR or new IR illumination can penetrate the dermis of the face. IR or new IR illumination may penetrate the dermis by 10 micrometers, 20 micrometers, 50 micrometers, 100 micrometers, 200 micrometers, 500 micrometers, 1 millimeter, 2 millimeters, 5 millimeters, and/or 10 millimeters. can Other penetration depths are possible. The penetration depth may depend on the location of the body, the wavelength of the infrared illumination, and/or the intensity of the infrared illumination. Penetration can expose features of the skin that may be difficult to see with visible light, including age spots, spider veins, hyperpigmentation, rosacea, acne and porphyrins. have. The identification of these subdermal features can be used to adjust/complement the requestor's unique identification. These features on the requestor's face may be unique as they are based on the requestor's exposure to nature and the sun during the requestor's life. Facial recognition based on subcutaneous features can identify the uniqueness of a face upon capture to provide an opportunity for identification analysis. The number of subcutaneous features may increase over time and daily with exposure to the sun.

In another example, the access control system may utilize ultraviolet illumination and detection to identify facial features. Ultraviolet light can penetrate the dermis of the face. UV illumination may penetrate the dermis by 10 micrometers, 20 micrometers, 50 micrometers, 100 micrometers, 200 micrometers, 500 micrometers, 1 millimeter, 2 millimeters, 5 millimeters, and/or 10 millimeters. Other penetration depths are possible. The penetration depth may depend on the location of the body, the wavelength of the ultraviolet light and/or the intensity of the ultraviolet light. Penetration can expose features of the skin that may be difficult to see with visible light (including age spots, spider veins, hyperpigmentation, rosacea, acne, and porphyrins). The identification of these subcutaneous features can be used to adjust/complement the unique identification of the requestor. These features on the requestor's face may be unique as they are based on the requestor's exposure to nature and the sun during the requestor's life. Facial recognition based on subcutaneous features can identify the uniqueness of a face upon capture to provide an opportunity for identification analysis. The number of subcutaneous features may increase over time and daily with exposure to the sun. The facial recognition system of the present disclosure may estimate a person's age based on the amount and characteristics of the subcutaneous features. Access control systems can also track changes in these features over time to identify an individual, and characterize lifestyle and routine based on the interpretation of subcutaneous features. Subcutaneous facial recognition can also increase the difficulty of creating replicas of faces (e.g., replicas of biometric templates) because it removes the dependence on facial features that can be captured by standard visible wavelength photography and camera techniques. have. The access control system also controls the content of the UV capture by introducing a time-sequenced cross-polarization filter into the capture process that further removes the ability to provide the access control system with an artificial replica of a face. It can be further obfuscated.

Advantages of the system in this disclosure include allowing the requestor to have a single credential system that replaces persistent, PIN, password and multi-factor authentication. Using this architecture in place, the requestor(s) of the system can rely on a single qualification solution. The system of the present disclosure may support both logical and physical gateways. In some implementations of the present disclosure, the system can provide protection at home and at work.

Aspects of the present disclosure may include a method referred to as "layered reinforcement." The method involves taking an image of a face from a biometric sensor and overlaying several layers of pixel boxes of different sizes on the image. This layering of pixel boxes of different sizes has a great impact on the analysis of facial uniqueness. A more distinct area for the face is magnified. The more general areas of the face are less emphasized. Consequently, layered reinforcement can allow the method to handle a large number of users at multiple sites where biometric sensor ECDs are deployed, while improving algorithm performance. The "layer enrichment" of the method can allow processing of the same number of requestors to the local Advanced Reduced Instruction Set Computing Machine (ARM) in the biometric sensor ECD where the image is captured first, thus reducing hardware and processing requirements and contributes to the accuracy and reliability of the method, as network failures cannot prevent the biometric sensor ECD from processing face identification.

Some aspects of this embodiment of the invention include the following processes: pixel box hierarchical analysis to generate a binary tree of dominant features (ie, determining which features are the most distinct); pixel box time domain analysis using heatmaps (ie, determining over time the features in question due to overlap between subjects); and by reducing false negatives and false positives through binary tree collisions (overlap flagging and proactive addressing of biometric signature data for two subjects that could lead to false positives). Includes the use of gateways (discussed below) to manage data that is analyzed by various algorithms to increase performance.

Advantages for the systems of the present disclosure include improved performance when accuracy requires reduction of false negatives and false positives. The improvement also allows the benefit of 1:1 comparison in a 1:N environment as a potential replacement for video surveillance and comparison, greatly increasing accuracy in the large-scale surveillance market.

Referring to FIG. 1 , an example of an identification system 100 for simultaneous real-time identification and authentication for use, eg, to allow access to a vehicle, building, etc., by an authorized requestor is illustrated in accordance with aspects of the present disclosure.

It should be appreciated that FIG. 1 is intended to illustrate aspects of the present disclosure to make available to those skilled in the art. Other implementations may be utilized and changes may be made without departing from the scope of the present disclosure.

The identification system 100 includes at least one biometric sensor 104 , a processor 106 , a memory 108 , a display 110 , and an input/output mechanism 112 , including simultaneous real-time identification and authentication. device 102 . Identification system 100 is designed to protect or control access to secure areas, devices, or information, such as airport boarding areas, buildings, stadiums, databases, locked doors, vehicles, or other access-controlled assets/infrastructure. can be used for

The biometric sensor(s) 104 may be a camera, fingerprint reader, retina scanner, facial recognition scanner, weight sensor, height sensor, body temperature sensor, gait sensor, heart rate sensor, or any other capable of detecting a biometric characteristic of a person. may include other sensors or devices of 2-4 , in an exemplary implementation of the present disclosure, the biometric sensor(s) 104 may be an optical sensor, such as a camera.

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 detects the requestor's visual information, such as a person's facial features. Facial features of a person may include texture, complexion, bone structure, moles, birthmarks, contours, and coloring of the person's face. The biometric sensor(s) 104 may capture facial features of a person and convert visual information, as discussed below, into digitally detected information.

The processor 106 may be configured to compare information sensed via the biometric sensor(s) 104 with known characteristics of the person to identify the person via the biometric signature data. Processor 106 may include any number of processors, controllers, integrated circuits, programmable logic devices, or other computing devices. The processor 106 is configured to allow information exchanged between the device 102 and an external device 114 or system to allow comparison of stored biometric signature data with sensed information obtained from the biometric sensor(s) 104 . may be communicatively coupled with the biometric sensor(s) 104 and other components of the system 100 via a wired or wireless connection (eg, the network 116 ) that enables it.

Processor 106 may implement computer programs and/or code segments stored in memory 108 to perform some of the functions described herein. The computer program may include an ordered list of executable instructions for implementing logical functions in device 102 . A computer program may be embodied in any computer readable medium (eg, memory 108 ) for use by or in connection with an instruction execution system, apparatus, or apparatus to execute instructions. Memory 108 may contain, store, communicate, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. Examples of memory 108 include one or more wires, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), a portable computer diskette, or a portable compact disk read. - May contain electrical connections with dedicated memory (CDROM). Memory 108 may be integrated with device 102 , standalone memory, or a combination of both. Memory 108 may include, for example, removable and non-removable memory elements, such as RAM, ROM, flash, magnetic, optical, USB memory devices, and/or other conventional memory elements.

In some aspects, memory 108 may store a number of known characteristics of a person, and various other data associated with the operation of system 100, such as the computer programs and code segments mentioned above, or visible to device 102 and other device elements. It may store other data for instructing to perform aspects described herein. The various data stored in memory 108 may be associated in one or more databases (not shown), such as to facilitate retrieval of information via external device 114 or network 116 . Although the memory 108 is integrated into the device 102 as shown in FIG. 1 , the memory 108 may be a standalone memory located in the same enclosure as the device 102 , or the device 102 . ) may be an external memory accessible by

In one aspect, the display 110 may be configured to display various information regarding the system 100 and its basic operation. For example, a notification device (not shown) may be included to indicate that a sensed biometric feature or sensed signal does not match a known characteristic of a person and may include an audible alert, a visual alert and/or any other A notification device may be included.

In one aspect, device 102 also provides input to facilitate exchange of data and other information with various external devices 114 or systems, either between different components within device 102 , or via network 116 . /output mechanism 112 .

SD card slot, mini SD card slot, micro SD card slot, etc., for accommodating a removable Secure Disk Digital (SD) card, mini SD card, micro SD card, etc., or other computing device such as a personal computer. A variety of I/O ports are contemplated, including a USB port for coupling with a USB cable communicatively coupled to the <RTI ID=0.0> In some aspects, the input/output mechanism 112 may include an input device (not shown) for receiving identification information about the person to be identified. The input device may include a ticket reader, credit card reader, identification card reader, keypad, touch-screen display, or any other device. In some other aspects, as described above, the input/output mechanism 112 allows the device 102 to connect to a network 116 , such as the Internet, a local area network, a wide area network, an ad-hook or peer-to-peer network, or a direct connection; For example, it may be configured to enable communication with another electronic device via a USB, Firewire or Bluetooth ^{TM connection or the like.} In one example, known characteristics of a person may be stored and retrieved from a remote database or memory via network 116 . Accordingly, the input/output mechanism 112 may be configured to use a wired data transmission method or a wireless data transmission method, such as WiFi (802.11), Wi-Max, Bluetooth ^™ , ANT®, ultra-wideband, infrared, cellular phone, radio frequency, etc. It can be used to communicate with the network 116 . In one aspect, the input/output mechanism 112 is for sending and receiving communications over a communications network operable with Global System for Mobile Communications (GSM), Code Division Multiple Access (CDMA), or any other known standard. It may include a cellular transceiver.

Device 102 may also include a power source (not shown) to provide power to various components included therein. A power source may include a battery, a battery pack, a power conduit, a connector, and a receptacle operable to receive a battery, battery connector, or power cable.

In one aspect, the device 102 provides access-controlled access control to prevent a person from accessing a particular area until the device 102 determines that the sensed biometric characteristic and/or signal matches a known characteristic. It may be installed and positioned at an entry point (not shown), such as a gate, locked door, or the like. In some other aspects, as shown in FIGS. 2-4 , the device 102 may be a standalone, compact, handheld, and portable device. In one example, such stand-alone, small, handheld and portable devices may be used to protect sensitive documents or information stored and accessed electronically on the Internet and/or intranets. In some aspects, the simultaneous real-time identification facility access unit may use the biometric signature data to generate interchangeable authentication for a variety of uses (eg, office, home, smart phone, computer, facility).

Referring to FIG. 5 , the processor 106 of FIG. 1 is configured to provide, among other features, simultaneous real-time or near-real-time identification and authentication with an authorized requestor's keyless access to a secure facility or information, detecting, among other features. It may be configured to include a module 502 and a recognition module 508 . The detection module 502 may include a face detection module 504 for detecting a facial feature of the requestor. The detection module 502 may include an eye detection module 506 for identifying the location of the requestor's eye. In some implementations, the detection module 520 may include one or both of the face detection module 504 and/or the eye detection module 506 . In some aspects, the processor 106 may receive input (digital or analog) from the sensor(s) 104 .

6 illustrates an exemplary procedure for selecting key features from a database with a large number of facial information and building one classifier capable of distinguishing different faces accordingly. LBP and LTP can be used to provide a full description of facial information, and then, with the use of an adaptive boosting (“adaboost”) learning algorithm, can select key features and generate biometric signature data. By creating it, we can build a classifier that distinguishes different faces. This biometric signature data can be used in a variety of applications (eg computers, building access, smart phones, automobiles, data encryption) with variable degrees of access and security (eg access to the network, but enhanced security on the requestor computer). It can be used to create universal validations and certificates that can be used. At block 602, a face sample database is created. For example, the processor 106 and the recognition module 508 may use unrecognized face samples to generate a face sample database. In one implementation, the processor 106 and the recognition module 508 may store in the memory 108 1000 different people, each person showing 10 different postures and/or facial expressions.

At block 604 , the LBP and LTP features are extracted. For example, the detection module 502 and/or the face detection module 504 may extract LBP and LTP features at different positions in each face sample, from different blocks.

At block 606 , positive and negative samples are calculated. For example, at least one of the face detection module 502 , the face detection module 504 , and the eye detection module 506 calculates a feature absolute value distance for the same position in any two different images from a person. and set this distance as a positive sample feature database. Further, the face detection module 502 and the face detection module 504 jointly or separately calculate the absolute value distance of the feature to the same position in any two different images from different persons and calculate this distance as the negative sample feature. It can be set as a sub-database.

At block 608, an Adaboost classifier is built. For example, the face detection module 502 and the face detection module 504 may use Adaboost to select the most distinguishable key features from a database of candidate features and generate a face classifier.

At block 610, a recognition result is generated. For example, the recognition module 508 may generate a recognition result. Once there is a fixed data set of macroblocks and a specific LPB ranging from 1 to 255 has been determined, a value is assigned to that unit of the data set based on the number of pixels in the block that satisfy that specific LBP. For example, assuming a 10 x 10 macroblock with an LBP of 20 and a unit number of 1 of 255, the method 600 determines, for a scalar value, the number of pixels in the histogram belonging to an LBP of 20. Method 600 solves the problem of determining values within macroblocks of various sizes by computing a scalar value and then normalizing the values in the second array. The scalar value based on the known method is based on the size of the macroblock, where the maximum value may be 100 to 1600 depending on the size of the macroblock. The scalar value of this second array can now be a percentage of the total pixels available in that macroblock to normalize the data for subsequent evaluation. Normalization ensures that the data is not distorted based on the size of the macroblock. After normalization under the improved method, each unit of the data set in this second array has the same weight. This normalized data is then classified so that values from 1 to 2165 can be set and assigned, where the scale reflects the highest normalized value moving to the top of the classification. For example, if data set 2000 has the highest value in the array, it will be assigned a value of 1 with descending values reflecting the data set with lower values. The second normalized array can then be transformed into a third simulated dna sequencing array, wherein a position is established within this third array based on its value in a previous classification. A third array evaluates the position and calculates the difference (eg, rank distance) between where the data sets appear in the sequence. This improved method is based on the uniqueness of a trait within a face rather than simply a scalar value, and analyzes the characteristic as opposed to the scalar value.

At block 620, the face sample is tested. For example, the detection module 502 may optionally test a face sample.

At block 622 , the LBP and LTP features are extracted. For example, the detection module 502 and/or the face detection module 504 may extract LBP and LTP features at different positions in each face sample, from different blocks.

Also, online recognition includes the following steps:

(1) Calculate offline stage extracted key features of different blocks at different positions for face samples to be identified.

(2) Calculate the key features selected in step (1) as features of each face sample in the database, and determine whether they belong to the same person. If the calculated distance is less than a set threshold, it may be determined that they are the same person, otherwise it may be determined that they are not the same person.

As shown in FIG. 8 , the exemplary process begins with creating a database of faces having different postures and different facial expressions. For example, one may include, for example, 1000 images of different people, each person showing eg 10 images differently. Fig. 7a shows different face images of the same person, and Fig. 7b shows different face images of different people.

LBP and LTP can be used to describe a face. Fig. 8 shows the calculation process of the LBP feature, and Fig. 9 shows the calculation process of the LTP feature. In order to obtain many features that describe facial information, different block sizes can be divided at different positions of the face sample. For example, the face size may be 100 x 100, the block size may be w x h, the w and h values may range from 2 to 100, resulting in 7837 blocks selected. The identification system 100 may select LBPs and bin features of LTPs at different block sizes and make them into a database of final candidate features.

The next step is to count positive and negative samples. The bin feature absolute value distance of the same position for different images from the same person can be calculated and set to a positive sample. Additionally, bin feature absolute value distances of the same position relative to different persons may be calculated and set to negative samples. For example, the result may involve calculating 32356 positive samples and 58747698 negative samples.

Then, among the multiple feature databases, key bin features capable of distinguishing all positive and negative samples can be selected using a learning algorithm. For example, we can choose a separate Adaboost learning algorithm to select features and build a classifier.

An example algorithm for classifying using Adaboost may include the following computational steps:

1. Given f as the maximum negative sample error rate, d as the minimum positive sample correct rate, F _{tar as the target of the negative sample error rate, and D tar} as the target of the positive sample correct rate that the cascade classifier should achieve. P and N are positive and negative databases, respectively.

2. Set F ₀ = 1.0, D ₀ = 1.0, and i = 0;

3. If F _i > F _tar , then i = i + 1, n _i = 0, F _i = F _i-1 ; When F _i > fx F _i-1 , n _i = n _i+1 .

4. Compute a strong classifier with n features in databases P and N via adaboost; _{Calculate F i} and D _i for the current cascade classifier, adjust the threshold value of the current strong classifier until the rate is _{equal to or greater than dxD i-1, where N is a nonempty set.}

5. If F _i > F _tar , classify the currently acquired cascade classifier from other negative sample images, and put the erroneously determined image into N.

1) Given n computed samples (x _i , y _i ),...,(x _n , y _n ), y _i = 0, 1, x _i is, respectively, a negative sample label and a positive sample label. indicates.

2) Initialize weights

Here, the number of positive samples is l, and the number of negative samples is m.

3) Try t from 1 to T and repeat the steps below:

a) weight normalization

를 표시한다.b) compute a weak classifier h _j for each feature f _j , and the error rate of this classifier

to display

를 갖는 분류기 h_t를 찾는다.c) the lowest error rate among all weak classifiers computed from the last step

Find a classifier h _{t with}

중에서, 가중치

를 업데이트 하고, x가 올바르게 분류되면,

이다. 그렇지 않으면,

이다.d)

Among them, weight

update , and if x is classified correctly,

am. Otherwise,

am.

이면, h(x) = 1 이고, 그렇지 않으면, h(x) = 0 이다. 여기서

이다.Finally a strong classifier is obtained:

, then h(x) = 1, otherwise, h(x) = 0. here

am.

10 shows the positions of the first three key features selected in a facial image by performing an online test on a database of 100 human faces based on offline selected features and classifiers.

11 shows the recognition results for 100 people, and the X-axis represents a false accept rate, which means the percentage of face samples that are erroneously identified. The Y axis represents the verification rate, which means the percentage of correctly recognized face samples. 11 , when the false acceptance rate is ^{less than 10 -4} , a 95% recognition rate can be achieved. Face recognition in this example not only improves the robustness of the face sample, but also reduces its computational complexity, which greatly improves face recognition.

Referring back to FIG. 5 , in some aspects, the detection module 502 follows, among other features, the face detection module 504 and the eye detection module 506 to process the acquired image of the person to be identified. They can be configured to be used together.

face detection module (504)

Input: Acquired frontal face image (gray image), face classifier

Output: face frame position and number of faces

flow:

a. Reducing the acquired frontal face image to a user-defined size

b. Compute the integral image of the reduced image

c. Initialize the traverse window based on the size (eg 20 x 20) defined by the face classifier

d. Moves the traverse window of the integral image from left to right, and then moves from top to bottom using each move distance corresponding to a user-defined distance. However, if the user-defined distance is 0, set the travel distance to 1/20 the width of the traverse window.

e. A face classifier is used to determine whether the current position of the traverse window defines a valid part of the face. If so, save the current rectangular frame position of the traverse window as a result.

f. After traversing the entire integral image, increase the width and 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 to store the result is exhausted. Repeat.

g. Return to the face frame position and face.

Eye detection module (506)

Input: Acquired frontal face image (gray image), face frame position, classifier for both left eye and right eye, left eye classifier, right eye classifier, left eye coarse detection classifier, right coarse eye detection classifier

Output: frame position for both eyes, frame position for left eye, frame position for right eye

flow:

a. Acquiring a face image from the acquired frontal face image

b. If a user-defined classifier for both the left eye and the right eye is available, a correspondingly defined face detection function is used to detect both the left eye and the right eye in the obtained face image. Otherwise, it estimates the position of both the left eye and the right eye based on experience.

c. If a user-defined left/right eye coarse detection classifier is available for the left/right eye, it detects the left/right eye in the corresponding half of the acquired face image. Further, based on the rough detection result, it is determined whether the detected human subject is wearing glasses. If glasses are present, it detects the acquired face image and returns it with the result. In the case where the glasses are not present, the face image obtained based on the coarse detection result is continuously detected, and the detection result is returned without considering the presence of the glasses. (If a user-defined classifier for the glasses-wearing subject is not available, the acquired face image is detected without considering the presence of glasses.)

d. If a user-defined coarse detection classifier is not available, it determines whether glasses are present by directly detecting the left/right halves of the acquired facial image. If glasses are present, it detects the acquired face image and returns it with the result. If the glasses are not present, the face image obtained based on the coarse detection result is continuously detected, and the detection result is returned without considering the presence of the glasses. (If a user-defined classifier for the glasses-wearing subject is not available, the acquired face image is detected without considering the presence of glasses.)

e. return

In some aspects, the processor 106 may be configured to extract appropriate facial features obtained from the detection module 502 for comparison against known features and/or information of multiple authorized persons, such as, for example, a recognition module ( 508) can be additionally used.

Recognition module (508)

normalization

Input: image to be normalized (grey image), coordinates of the centers of both left and right eyes on the image axis (the origin is located in the upper left corner of the image). Meaning of the parameter: 1x indicates 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 1y is the left eye (vertical direction) of the output image divided by the height of the output image. Points to the x-coordinate of the center point.

Output: output image

Feature extraction

Input: normalized image (grey image) and feature type

Output: If the output buffer is NULL, return feature dimensional degrees. Otherwise, it is assumed that the size of the output buffer is equal to the feature dimension diagram, writes the features of the image to the buffer, and returns the feature dimension diagram. A particular feature is associated with a particular image size. For example, feature #6 may require an image size of 100 x 100. Thus, if the input image fails the corresponding defined image size requirement, a result of zero may be returned.

Feature Comparison

Input: 2 features to be compared and how to compare

Output: The smaller the comparison result (floating point), the higher the similarity.

Obtain Algorithm Information

Function: Instructs the requester to correctly assign parameters to the algorithm.

Input: Algorithm type based on usage context

Output: Parameter information of the algorithm including feature type, feature dimensionality diagram, normalized image size, minimum distance, suggested range and distance type.

Many of the systems and methods described above can be used to generate biometric signature data (“BSD”) files that allow the system to identify and differentiate a requestor with high fidelity. Various implementations of the present disclosure may use BSD files to generate encryption/decryption keys, thus increasing the security of such keys. Examples of this disclosure may generate an asymmetric key based on one or more BSD files in such a way that, by utilizing a biometric sensor, a person's biometric measurement may act as the person's private key. Additionally, implementations of the present disclosure may be implemented in such a way that the file is encrypted in such a way that it cannot be decrypted or accessed by anyone other than the requestor or group of the intended requestor, or in such a way that the original owner, such as an enterprise, can no longer access the file. BSD files can be integrated into digital rights management (DRM) security. Accordingly, by using an implementation of the present disclosure that uses a BSD file, the identity of the requestor who accessed the file can be ensured when the file is accessed.

BSD files can be generated by algorithmic analysis of data from A/D IR and/or UV sensors. Accordingly, many of these factors may be considered when constructing the private key of an asymmetric pair (ie, analog and/or digital values). Thus, in some implementations of the present disclosure, multiple elements of a sensor may contribute real-time data or real-time analog data related to a recognition event for decryption, such that the real-time event (ie, an actual measurement of the intended person) is authenticated. can be guaranteed to work.

12 and 13 , according to some implementations of the present disclosure, a message may be transmitted as follows. A requestor may, for example, register with a computer and generate a public key for the requestor. The requestor then publishes the public key so that the key is publicly known. Another person, system, or entity may use the requestor's public key to encrypt a message to the requestor and send the message to the requester. The requestor can decrypt the message using his or her private key generated by one or more live BSD files associated with the requestor. Accordingly, since the private key used to decrypt the message can be generated by the requester's live biometric data, the sender of the message confirms that the requestor is actually the person decrypting the message. These systems and methods for encryption provide significant advantages over conventional systems and methods. For example, by using a BSD file for the encryption process, instead of simply matching an anonymous asymmetric code, the authentication is inherent in the key itself.

Various implementations of the present disclosure may also improve DRM. For example, DRM rules may allow additional content to be added to a file and additional rules required. DRM rules may be expressed in many rights management languages known in the art, including, but not limited to, extensible rights markup language (XrML), extensible media commerce language (XMCL), open digital rights language (ODRL), and the like. Rules can specify allowed actions (eg, decrypt, encrypt, transmit, copy, edit, etc.). Rules can also specify who is authorized to perform actions and the conditions under which these actions are allowed. The BSD file can be used to authenticate the requestor to determine if the requestor is one of the people specified in the rule.

Various systems and methods for biometric encryption and authentication may also be used in a corporate environment, such as when an employee uses a company device for business as well as personal use, or, for example, when an employee uses a personal device and company digital assets are on the personal device. The application can be found in the case being sent and sent from this personal device. By applying rules to documents with specific digital signatures, there can be a controllable segmentation between personal and business interests. Both parties can access the part to which they have access, but access to the part to which they do not have access can be blocked. For example, possible applications include, but are not limited to, providing remote access, making a purchase, and conditional security.

For remote access, various implementations of the present disclosure may create a BSD file that is used to authenticate the requestor, thus providing secure access to any remote network connection, i.e., VPN server, network email, from a remote device; and/or provide secure access to company-owned information.

Additionally, the biometric authentication technology of the present invention can be used to perform authenticated online purchases/transactions. For example, the spending limit may be based on a requestor or group profile for the account. In order for a requestor to make a purchase, the system may use the biometric authentication techniques of the present disclosure to authenticate the true identity of the requestor to verify that the requestor is entitled to make the desired purchase.

Biometric authentication techniques can also be used to provide conditional security to various digital files. For example, a file containing sensitive information can only be accessed by an authorized requestor who can be authenticated using the requestor's live BSD file.

Biometric Encryption and Authentication Applications for Digital Cinema

The biometric encryption and authentication techniques described herein find many applications in the digital cinema industry. Movies, especially pre-DVD releases, are popular. In order to maximize both production efficiency and distribution opportunities, movies need to be accessed and handled by several different requestor classes. Those skilled in the art will recognize that technologies that can be used to secure digital assets in the digital cinema industry can be used to secure digital assets in virtually any industry. Accordingly, the principles described herein are not limited to applications in the digital cinema industry, but may instead be applied in any industry for similar purposes.

Digital cinema security sees itself as an end-to-end process from production through distribution to consumption.

SMPTE DC28, the body responsible for digital cinema standards, has identified five distinct areas of digital cinema: (1) capture; (2) production; (3) Master (Cinema, Home, Video, Trailer, Premiere); (4) distribution [satellite, fiber, packaging]; and (5) exhibits (digital projector security). In each area identified by SMPTE DC28, the film is vulnerable to theft. To prevent theft, movies may be encrypted prior to distribution. The movie is then stored in theaters, typically encrypted until run time. At run time, the movie is decrypted and decompressed. This decryption/decompression can happen on the server or on the projector.

In the example SMPTE DC28 process, DC28.4 may represent the conditional access portion of the cinema delivery system. Modem DRM encryption methods have proven to be sufficiently resistant to tampering with decryption attempts, but securing the key is problematic. From capture to exhibition to distribution, the film is encrypted and decrypted many times. Accordingly, the various various biometric encryption and authentication techniques discussed herein may be applied to one or more of the encryption, decryption, and authentication steps in accordance with various implementations of the present disclosure.

Referring to FIG. 14 , an example of an environment 1400 for managing data includes a first gateway 1402a, a second gateway 1402b, and a third gateway for routing data over wired and/or wireless communication links. 1402c). The first gateway 1402a may be implemented as a virtualized software-based gateway 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.

In certain implementations, gateway 1402 may perform several functions, including managing the movement of data to and from biometric sensors, as described below, and efficiently transfer binary facial data between devices. It provides a mobile network solution and, when clustered together (physical and virtual), a high availability solution for security design. The gateway 1402 may receive credentialing data through an XML file structure. By monitoring and actively consuming XML files, gateway 1402 utilizes a standardized universal interface that is independent of programming languages, operating systems, and connection types of data sources, and provides physical and logical access control requirements within the same way. support, provide an interface to support simultaneous connections from different system types, and/or support live or batch processing of credentialing, including instantaneous recovery of credential systems through file replacement.

In certain aspects, data stored in the gateway 1402 may be stored based on a JavaScript object notation format.

Environment 1400 may include one or more ECDs 1404 , such as ECD-a 1404a , ECD-b 1404b , ECD-c 1404c , ECD-d 1404d , ECD-e 1404e , and ECD -f (1404f). ECD 1404 may also be referred to as an edge capture device. ECD 1404 may send data to gateway 1402 to be routed to other devices in environment 1400 . For example, the ECD-a 1404a may send the biometric template and/or the person's identity information to the third gateway 1402c. The third gateway 1402c may send one or more biometric templates to the ECD-a 1404a to perform a matching operation.

With continued reference to FIG. 14 , in a particular implementation, 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 the storage device 1410 . The storage device 1410 may be a network drive, a local hard drive, a flash drive, a tape drive, or other suitable storage medium.

With continued reference to FIG. 14 , for example during operation, ECD-b 1404b may receive a biometric template of first requestor 1450a . The first requestor 1450a may attempt to access one or more access controlled entry points 1408 controlled by the access control server 1406a. One or more access-controlled entry points 1408 may include, for example, a safe, lock, security door, security gate, equipment or machine, computing device, digital storage device, database or file. In some examples, one or more access-controlled entry points 1408 may be access-controlled doors or gates of infrastructure, such as warehouses, office buildings, restricted areas, and the like. The biometric template of the first requestor 1450a may include one or more of a fingerprint, a voice pattern, an iris pattern, a facial feature, a signature pattern, an ear shape, a retina pattern, a gait, and a hand shape of the first requestor 1450a. can

In certain implementations, ECD-b 1404b may extract facial features of first requestor 1450a and compare the facial features to facial features of the authorized person. If the facial feature of the first requestor 1450a matches one of the facial features of the authorized person, the ECD-b 1404b may send a first positive match signal to the 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 unlocks one of the one or more access controlled entry points 1408 associated with the ECD-b 1404b to allow the requestor 1450 access. can If the facial feature of the first requestor 1450a does not match one of the facial features of the authorized person, the first requestor 1450a will not be able to gain access to the entry point 1408 associated with the ECD-b 1404b. can

With continued reference to FIG. 14 , in the specific example, ECD-d 1404d draws a second positive match signal of second requestor 1450b at a first time and a third positive at a second time of second requestor 1450b . A match signal may be sent 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 gateway 1402a, which may route the second and third positive match signals to the enterprise server 1406b. have. 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 requestor 1450b. For example, the enterprise server 1406b may record the first time as the arrival time of the second requestor 1450b on the workday and the second time as the departure time to the storage device 1410 . In another example, the enterprise server 1406b may record the first and second times as the number of accesses to the one or more access-controlled entry points 1408 by the second requestor 1450b. The enterprise server 1406b may also log the premises, equipment, files, locations, and information accessed by the second requestor 1450b based on the information in the second and third positive match signals.

With continued reference to FIG. 14 , in a particular implementation, ECD-f 1404f may send a fourth positive match signal of third requestor 1450c to first gateway 1402a . The first gateway 1402a may route the fourth positive match signal through the firewall 1412 to the third party server 1406c. Firewall 1412 may filter information transmitted through firewall 1412 to prevent malicious requesters from gaining unauthorized access. The fourth positive match signal may indicate to the third-party server 1406c that the third requestor 1450c has obtained access to the one or more access-controlled entry points 1408 . For example, the fourth positive match signal may indicate to the third-party server 1406c that the third-party requestor 1450c has accessed software for which payment is required from the owner of the third-party server 1406c.

With continued reference to FIG. 14 , in a particular example, network manager 1452 , via workstation 1414 , ECD 1404 , gateway 1402 , access control server 1406a , one or more access-controlled entry points Install, manage, update, maintain, and/or control software of 1408 , enterprise server 1406b , storage 1410 , and/or firewall 1412 . Workstation 1414 may be a desktop computer, laptop computer, tablet computer, handheld computer, smart phone, or other suitable computing device communicatively coupled to third gateway 1402c via a wired or wireless connection. The network manager 1452 may send a software command to be routed to the destination from the workstation 1414 to the third gateway 1402c. In some examples, network manager 1452 may upgrade firmware in ECD 1404 . In another example, the network administrator 1452 may install new software on the access control server 1406a. In another example, the network manager 1452 may perform maintenance operations such as disk error checking and defragmentation on the storage device 1410 . In another example, network manager 1452 can lock or open one or more access controlled entry points 1408 in an emergency.

Referring to FIG. 14 , in some implementations, an employee 1454 , such as a supervisor, may utilize the requestor terminal 1416 to access information via the second gateway 1402b . The requestor terminal 1416 may be a desktop computer, laptop computer, tablet computer, handheld computer, smart phone, or other suitable computing device communicatively coupled to the second gateway 1402b via a wired or wireless connection. In some examples, the employee 1454 may use the requestor terminal 1416 to download information such as office hours, arrival times, access history, and frequency of use.

In certain examples, data exchange within environment 1400 may be encrypted. Data transmission between gateway 1402 and ECD 1404 may use advanced TLS v1.2 communication with a proprietary key management framework. Data transmitted over TLS v1.2 communication can be fully encrypted to eliminate the risk of data exposure to unwanted parties. Encryption of data may be further protected by protecting the generation of an encrypted key through the use of biometric data as a seed for generation of the key. As such, data exchanged within the environment may be difficult to access by unauthorized requestors.

Gateway 1402 may also be used to manage data and create a blockchain for requestors. The first gateway 1402a, the second gateway 1402b, and the third gateway 1402c may each include a blockchain wallet. The wallet will contain the requestor's individual credentials required to authenticate any device or application. The wallet may be coupled to the gateway 1402 to provide cybersecurity monitoring and to provide interaction between the facial recognition device and the wallet. An ECD 1404 linked to a private blockchain may enable transactions on the blockchain. Communication between the blockchain wallet, the ledger for blockchain transactions, and the gateway 1402 may use a blockchain protocol. Private blockchains can also be extended to devices that verify more than one requestor, such as a bank ATM (not shown). Gateway 1402 location tracking to move links to the link binary facial data and blockchain ledger between shared devices to improve the security of transactions and manage the number of requestors maintained within each ECD 1404 will utilize

Aspects of the present disclosure may be implemented using hardware, software, cloud networks, or combinations thereof, and may be implemented in one or more computer systems or other processing systems. In aspects of the present disclosure, features relate to one or more computer systems capable of performing the functions described herein. An example of such a computer system 1500 is shown in FIG. 15 . One or more of gateway 1402 , server 1406 , firewall 1412 , workstation 1414 , and/or requestor terminal 1416 may be implemented based on computer system 1500 .

Referring now to FIG. 15 , computer system 1500 includes one or more processors, such as processor 1504 . The processor 1504 is communicatively coupled to a communications infrastructure 1506 (eg, a communications bus, cross-over bar, or network). Various software aspects are described in terms of this exemplary computer system. After reading this description, it will become apparent to those skilled in the relevant art(s) how to implement aspects of the present disclosure using other computer systems and/or architectures.

The computer system 1500 may include a display interface 1502 that communicates graphics, text, and other data from the communications infrastructure 1506 (or from a frame buffer not shown) for display on the display unit 1530 . have. Computer system 1500 also includes main memory 208 , preferably random access memory (RAM), and may also include secondary memory 1510 . Secondary memory 1510 may include, for example, a removable storage drive 1514 representing a floppy disk drive, a magnetic tape drive, an optical disk drive, a universal serial bus (USB) flash drive, and/or a hard disk drive 1512 . ) may be included. Removable storage drive 1514 reads from and/or writes to 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. that are read by and written to removable storage drive 1514 . As will be appreciated, removable storage unit 1518 includes computer usable storage media having computer software and/or data stored therein.

Alternative aspects of the present disclosure may include auxiliary memory 1510 and other similar devices for allowing computer programs or other instructions to be loaded into computer system 1500 . Such devices may include, for example, a removable storage unit 1522 and an interface 1520 . Examples of these include program cartridges and cartridge interfaces (such as those found in video game devices), removable memory chips (such as erasable programmable read-only memory (EPROM), or programmable read-only memory (PROM)) and associated sockets; and another removable storage unit 1522 and interface 1520 , which allow software and data to be transferred from the removable storage unit 1522 to the computer system 1500 .

Computer system 1500 may also include a communication interface 1524 . Communication interface 1524 allows software and data to be transferred between computer system 1500 and external devices. Examples of the communication interface 1524 may include a modem, a network interface (eg, an Ethernet card), a communication port, a Personal Computer Memory Card International Association (PCMCIA) slot and card, and the like. Software and data transmitted over communication interface 1524 are in the form of signals 1528 , which may be electronic, electromagnetic, optical, or other signals that may be received by communication interface 1524 . These signals 1528 are provided to the communication interface 1524 via a communication path (eg, a channel) 1526 . This path 1526 carries the signal 1528 and may be implemented using one or more of wires or cables, fiber optics, telephone lines, cellular links, RF links, and/or other communication channels. In this document, the terms "computer program medium" and "computer usable medium" are generally used to refer to media such as removable storage drive 1518 , hard disk installed in hard disk drive 1512 , and signal 1528 . do. This computer program product provides software to the computer system 1500 . Aspects of the present disclosure relate to such computer program products.

Computer programs (also referred to as computer control logic) are stored in main memory 1508 and/or secondary memory 1510 . The computer program may also be received via communication interface 1524 . Such computer programs, when executed, enable computer system 1500 to perform features according to aspects of the present disclosure, as described herein. In particular, the computer program, when executed, enables the processor 1504 to perform features according to aspects of the present disclosure. Accordingly, this computer program represents the controller of computer system 1500 .

In aspects of the present disclosure in which a method is implemented using software, the software may be stored in a computer program product, and may be stored in a computer system 1500 using a removable storage drive 1514 , a hard drive 1512 , or a communication interface 1520 . ) can be loaded. 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, a system is implemented primarily in hardware using hardware components such as, for example, application specific integrated circuits (ASICs). Implementation of a hardware state machine to perform the functions described herein will be apparent to those skilled in the relevant art(s).

16 illustrates a block diagram of various example system components, in accordance with aspects of the present disclosure. 16 illustrates a communication system 1600 usable in accordance with aspects of the present disclosure. Communication system 1600 includes one or more accessors 1660 and 1662 and one or more terminals 1642 and 1666 . In one aspect, data for use in accordance with aspects of the present disclosure can be, for example, a terminal 1642 , 1666 , such as a personal computer (PC), minicomputer, mainframe computer, microcomputer, telephone device, or wireless device. [For example, a PC, mini computer having a connection to a processor and storage for data and/or storage for data via a network 1644, such as the Internet or an intranet, and couplings 1645, 1646, 1664 , via a hand-held wireless device or personal digital assistant (“PDA”) coupled to server 1643, such as a mainframe computer, microcomputer, or other device], inputted and/or by accessors 1660, 1662 is accessed Couplings 1645 , 1646 , 1664 include, for example, wired, wireless and/or fiber optic links. In another exemplary variant, methods and systems according to aspects of the present disclosure operate in a standalone environment, such as a single terminal.

Referring now to FIG. 17 , an example of an ECD 1404 configured to perform access control is used to determine whether a requestor 1450 is authorized to obtain access to an entry point (not shown) associated with the ECD 1404 . , the biometric template of the requestor 1450 may be analyzed. The ECD 1404 may include an optical sensor 1702 , an illumination source 1704 , a display 1706 , a keypad 1708 , and a scanner 1710 .

In some implementations, the optical sensor 1702 may be configured to capture still images or moving images. For example, optical sensor 1702 may capture requestor 1450's fingerprint, iris pattern, facial features, signature pattern, ear shape, retina pattern, gait, and/or hand shape. In some examples, the optical sensor 1702 may be a broadband camera configured to detect electromagnetic radiation having a wavelength in the range of 200 nanometers (eg, soft UV) to 2000 nanometers (eg, near infrared (NIR)). In a particular example, the optical sensor 1702 is configured to detect radiation between 700 and 900 nanometers. In certain implementations, the optical sensor 1702 can include a motion sensor configured to detect a person approaching the ECD 1404 .

In a particular example, the optical sensor 1702 provides vertical area coverage for capturing faces across the full extent of a person's height (while addressing Americans with Disabilities Act requirements), as well as providing access to secure areas with one authentication. wide-angle lenses (eg, fisheye lenses) used to provide horizontal coverage of the entire area of the access point to handle more than one person. In another example, the optical sensor 1702 may use a high resolution (eg, megapixels per square inch) charge coupled device (CCD) array. High-resolution arrays could provide the ability to identify faces at a greater distance from the sensor. The ECD 1404 may use the increased identification distance to pre-identify the requestor in a queuing situation. Potential requesters can be identified as they enter the queue area and placed on a priority list to accelerate identification and discharge of queues at access-controlled entry points. Pre-identifying and prioritizing individuals in the identification process at the access point enables high throughput at the point of entry by reducing the identification time during the identification and authentication process.

In certain implementations, the illumination source 1704 can emit electromagnetic radiation having a wavelength in the range of 200 nanometers to 2000 nanometers. In certain examples, the illumination source 1704 may emit invisible radiation between 700 and 900 nanometers and/or between 200 and 300 nanometers. The illumination source 1704 may emit radiation to illuminate the physical features and patterns of the requestor 1450 used in the biometric analysis (analysis of the biometric template to determine access rights). The emitted radiation may strike a part of the requestor 1450's body and be reflected off the body part. The reflected radiation may be captured by the optical sensor 1702 for biometric analysis.

In some implementations, display 1706 can present useful information to requestor 1450 . For example, display 1706 shows one or more images of face 1730 of requestor 1450 , captured by optical sensor 1702 , so that requestor 1450 aligns face 1730 during biometric analysis. can help you do In another example, the display 1706 can notify the requestor 1450 of the status of the entry point associated with the ECD 1404 (eg, locked, temporarily unavailable, operating normally, under maintenance). In another example, display 1706 may display information such as time, date, weather, current location, and the like.

With continued reference to FIG. 17 , the keypad 1708 can allow the requestor 1450 to enter numbers, symbols, and alphabets into the ECD 1404 . In one example, the requestor 1450 may enter a password in addition to the biometric analysis to gain access to the entry point.

In some implementations, the scanner 1710 may be a radio frequency identification (RFID) scanner, a proximity card scanner (eg, a HID ^™ card scanner), a contact card scanner, or a magnetic card scanner. In one example, the scanner 1710 may send an interrogation signal to a proximity card (not shown) having an integrated circuit and coil with a programmable or non-programmable identification sequence. The interrogation signal may be “absorbed” by the coil and energize the integrated circuit. In response, the powered integrated circuit sends a response signal comprising the identification sequence back to the scanner 1710 via the coil. In turn, scanner 1710 analyzes the identification sequence to determine whether to grant access. The identification sequence may be one or more numbers, alphabets, symbols, and/or combinations thereof.

With continued reference to FIG. 17 , in an implementation, the requestor 1450 may access the ECD 1404 during operation. The optical sensor 1702 can detect the requestor 1450 . In response to the detection, the illumination source 1704 may emit incident NIR radiation 1760 towards the requestor 1450 . Incident NIR radiation 1760 may strike face 1730 of requestor 1450 and be reflected from face 1730 of requestor 1450 . In some implementations, the optical sensor 1702 can 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 (ie, blackbody 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 , the angle of detection by the optical sensor 1702 , and other factors. The ECD 1404 can use the detected NIR radiation 1762 to construct a facial template (“NIR sampled profile”) of the requestor 1450 . ECD 1404 may compare the constructed NIR sampled profile to existing templates stored therein (details are described below). If ECD 1404 detects a match, ECD 1404 may allow requestor 1450 access to the entry point (as described above).

In some examples, the NIR sampled profile generated via NIR radiation detection may be tolerant to changes in ambient illumination. As ambient illumination fluctuates (e.g., as luminance, color, color temperature, illumination angle change), the NIR sampled profile constructed using NIR radiation detection will remain sufficiently constant to prevent false acceptance or false rejection. can For example, the NIR sampled profile of the requestor 1450 constructed via NIR radiation detection under “bright” conditions (eg, 1000 lux) may be constructed via NIR radiation detection under “dark” conditions (eg, 100 lux). may be substantially identical to the NIR sampled profile. In another example, the NIR sampled profile of the requestor 1450 configured via NIR radiation detection under substantially blue light (eg, CIE coordinates x = 0.153, y = 0.100) is substantially white light (eg, CIE coordinates x = 0.30) , y = 0.33) may be substantially the same as the NIR sampled profile constructed through NIR radiation detection.

With continued reference to FIGS. 14 and 17 , in another implementation, an NIR sampled profile generated via NIR radiation detection may enhance the privacy of the owner. For example, storage 1410 may store an employee's NIR sampled profile and associated confidential information (eg, date of birth, email account, password). If an unauthorized person has access to the NIR sampled profile and associated confidential information, it may be difficult to identify an employee based on the NIR sampled profile, so that the unauthorized person cannot exploit the stolen confidential information. Given that NIR sampled profiles are constructed using, for example, detected NIR radiation, they may not be recognizable as a person's NIR sampled profile may differ significantly from a visual image of the same person's face. .

In another implementation, the requestor 1450 can access the ECD 1404 during operation. The optical sensor 1702 can detect the requestor 1450 . In response to the detection, the illumination source 1704 may emit incident UV radiation 1760 (eg, electromagnetic radiation having a wavelength between 315-390 nanometers) towards the requestor 1450 . Incident UV radiation 1760 may strike face 1730 of requestor 1450 and be reflected from face 1730 of requestor 1450 . In a particular example, incident UV radiation 1760 may penetrate the surface of face 1730 and be reflected from subcutaneous features (eg, dermis, subcutaneous tissue, muscles, blemishes) of face 1730 . In some implementations, optical sensor 1702 can detect detected UV radiation 1762 originating from subcutaneous features and/or surfaces of 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 , the angle of detection by the optical sensor 1702 , and other factors. The ECD 1404 can use the detected UV radiation 1762 to construct a face template (“UV Sampled Profile”) of the requestor 1450 . ECD 1404 may compare the constructed UV sampled profile to existing templates stored therein (details are described below). If ECD 1404 detects a match, ECD 1404 may allow requestor 1450 to access the entry point (as described above).

In some examples, a UV sampled profile generated through detection of UV radiation may be tolerant to changes in ambient lighting. As ambient lighting fluctuates (e.g., as luminance, color, color temperature, illumination angle change), the UV sampled profile constructed using UV radiation detection can be kept constant enough to prevent false acceptance or false rejection. can For example, a UV sampled profile of requestor 1450 constructed via UV radiation detection under “bright” conditions (eg, 1000 lux) may be constructed via UV radiation detection under “dark” conditions (eg, 100 lux). It may be substantially identical to the UV sampled profile. In another example, the UV sampled profile of the requestor 1450 configured through detection of UV radiation under substantially blue light (eg, CIE coordinates x = 0.153, y = 0.100) is a UV sampled profile of substantially white light (eg, CIE coordinates x = 0.30) , y = 0.33) can be substantially identical to the UV sampled profile constructed through NIR radiation detection under

With continued reference to FIGS. 14 and 17 , in another implementation, a UV sampled profile generated through detection of UV radiation may enhance the privacy of the owner. For example, storage 1410 may store an employee's UV sampled profile and associated confidential information (eg, date of birth, email account, password). If an unauthorized person gains access to the UV sampled profile and associated confidential information, the unauthorized person cannot exploit the stolen confidential information because it may be difficult to identify an employee based on the UV sampled profile. Given that the UV sampled profile is constructed using, for example, detected UV radiation, this may be unrecognizable as a person's UV sampled profile may differ significantly from a visual image of the same person's face. . The sampled profile may be a rendering of a data set that visualizes the consistency of positions within the three arrays. If the sampled profile is compromised, it can be more difficult to obtain biometric information used to identify an individual.

In some examples, ECD 1404 can generate a visible light sampled profile based on detected visible light reflected from face 1730 of requestor 1450 .

In a particular implementation, the requestor 1450 may provide a password, a personal identification number (PIN) and/or a valid HID ^TM card to be used with the configured sampled profile to gain access to the entry point associated with the ECD 1404 . can be requested

In another implementation, the ECD 1404 may include a microphone (not shown in FIG. 17 ) to perform voice recognition.

In some examples, display 1706 can show face 1730 of requestor 1450 being imaged by optical sensor 1702 . Display 1706 may include alignment marks (not shown) to assist requestor 1450 in aligning face 1730 with respect to optical sensor 1702 . This sorting process can minimize false accepts and false rejects due to misalignment. In certain implementations, the ECD 1404 may be a pocket-sized mobile device powered by a rechargeable battery.

Referring to FIG. 18 , in some implementations, an ECD 1404 includes a processor 1802 having a communication module 1852 configured to communicate with a gateway 1402 and another ECD 1404 , as described in this disclosure. may include The communication module 1852 may be implemented, for example, as hardware in the processor 1802 , 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, for example, as hardware in the processor 1802 , 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, communication with the gateway may be enabled by a processor in the control panel. The algorithm module 1856 may be implemented, for example, as hardware in the processor 1802 , 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, for example, as hardware in the processor 1802 , as software code executed by the processor 1802 , or a combination thereof. Processor 1802 may include one or more processors or cores, and may be implemented as a semiconductor processor, graphics processing unit, field programmable gate array, programmable logic device, processing cluster, application specific integrated circuit, or other suitable architecture.

ECD 1404 includes memory 1804 . The memory may be static or dynamic memory, such as flash memory, random access memory, magnetic memory or semiconductor memory. Memory 1804 may include external memory, such as cloud storage. Memory 1804 may contain or store applications and/or computer executable code. The ECD 1404 further includes a modem 1808 for communicating with the gateway 1402 and other ECDs 1404 , and may operate in concert with the communication module 1852 . ECD 1404 also includes 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 a display 1706 , an optical sensor 1702 , a keypad 1708 , and a scanner 1710 . Components within ECD 1404 may be interconnected by an internal bus 1812a. Processor 1802 , memory 1804 , RAM 1806 , and internal bus 1812a may be disposed on processing board 1820 .

Referring now to FIG. 19 , another example of an ECD 1404 may include multiple processing boards 1820 interconnected by an external bus 1812b. Each processing board 1820 may include a processor 1802 , memory 1804 , RAM 1806 , and an internal bus 1812a . Data distributed among processing board 1820 may be distributed by controller 1814 via external bus 1812b. In a non-limiting example, ECD 1404 may download 240,000 biometric templates of potential requestors (from gateway 1402 ). The controller 1814 distributes 240,000 biometric templates evenly or unevenly among the processing boards 1820 such that each processing board 1820 can store 30,000 biometric templates in individual memory 1804 . can do. In some implementations, each processing board 1820 may store the same or a different number of biometric templates.

19 , in an example, during operation, ECD 1404 can construct a sampled profile (eg, UV or NIR) of requestor 1450 based on detected NIR radiation 1762 . The controller 1814 may distribute a copy of the configured sampled profile to each processing board 1820 . In some implementations, the processing board 1820 may concurrently compare the replicated sampled profile with locally stored biometric templates (eg, 30,000 stored on each processing board 1820 ). The current example of the ECD 1404 shown in FIG. 19 includes eight processing boards 1820 , although 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 .

In implementations, the ECD 1404 may rely on a remote processing board (not shown) to perform the distributed computing described above. For example, ECD 1404 may jointly and concurrently implement an algorithm (described below) for matching a replicated sampled profile (or a numerical representation of a replicated sampled profile) to a known biometric template. For this purpose, the replicated sampled profile can be sent to a remote processing board. The processing board may be in another ECD in the network. In a particular example, the ECD 1404 may send the replicated sampled profile to a Beowulf cluster. The clustering design may use inter-process and inter-processor protocols to share the processing tasks of the same application between two processors and the processing cores inside the processor. In the case of facial recognition, processor-intensive operations, such as facial verification, may utilize multiple processors or multiple cores to complete the operation within an allowed time period. In some cases, a large multi-core processor may be used for this type of operation. In other cases, operations are distributed across multiple advanced RIISC machine (“ARM”) processors to achieve the same performance as a single multi-core processor, without high hardware costs and the possibility of a potential single point of failure. can be

Referring back to FIG. 14 , in a non-limiting example of a distributed parallel comparison, the first gateway 1402a may download 600,000 biometric templates of potential requestors (eg, employees and contractors) from the enterprise server 1406b. have. The first gateway 1402a distributes 100,000 biometric templates to ECD-e 1404e and ECD-f 1404f, respectively, and distributes 200,000 biometric templates to the second gateway 1402b and the third gateway 1402c ) can be distributed to each. Next, the second gateway 1402b may distribute 100,000 biometric templates to each of ECD-c 1404c and ECD-d 1404d, and the third gateway 1402c may distribute ECD-a 1404a and It is possible to dispense 100,000 biometric templates to each of the ECD-b 1404b. As a result, the 600,000 biometric templates downloaded from the enterprise server 1406b may be evenly distributed among the ECDs 1404 (ie, 100,000 non-overlapping templates each). In some implementations, when ECD-b 1404b constructs a sampled profile (UV or NIR) of first requestor 1450a, ECD-b 1404b duplicates the constructed sampled profile and replicates the replicated sampled profile. can be deployed (via one or more gateways 1402) to ECD-a 1404a, ECD-c 1404c, ECD-d 1404d, ECD-e 1404e, and ECD-f 1404f. . ECD 1404 may compare the replicated sampled profile of first requestor 1450a to 100,000 locally stored templates to determine a match. ECD-a 1404a, ECD-c 1404c, ECD-d 1404d, ECD-e 1404e, and ECD-f 1404f return the results of the comparison (via one or more gateways 1402) back. ECD-b 1404b may be transmitted. ECD-b 1404b may collect the results and determine whether to grant access to first requestor 1450a. In other implementations, the biometric template may be distributed unevenly among the ECDs 1404 .

Referring now to FIG. 20 , the ECD 1404 can generate a sampled profile 2000 based on measuring the intensity of radiation 1762 detected from the ECD 1730 of the requestor 1450 . The sampled profile 2000 may include a base matrix 2010 of one or more measurement points 2002 . The measurement point 2002 may include a value representing the intensity of the radiation 1762 detected at the location of the specific measurement point. For example, measurement point-a 2002a may represent an intensity scale value of 2 corresponding to region 2004a (eg, dark hair) around face 1730 of requestor 1450 . Measurement point-b 2002b may represent an intensity scale value of 50 corresponding to the background color. Measurement point-c 2002c may represent an intensity scale value of 21 corresponding to region 2004c (eg, cheeks) on face 1730 of requestor 1450 . In a particular example, the intensity scale may range from 0 (no reflection) to 100 (the maximum reflection that can be detected by the optical sensor 1702 of the ECD 1404 ). In one implementation, the intensity scale may measure the absolute intensity (eg, brightness) of the measurement point 2002 . In another example, sampled profile 2000 can include measurement points 2002 for detected radiation 1762 of different wavelengths (eg, UV, NIR, red, green, blue).

Although the sampled profile 2000 of FIG. 20 shows a base matrix 2010 of 10 x 12 measurement points 2002 across the face 1730 of the requestor 1450 , a different measurement point density for the base matrix 2010 . may be possible For example, the ECD 1404 may generate another sampled profile using 100×100 measurement points across the face 1730 of the requestor 1450 . In another example, the ECD 1404 can generate a sampled profile using 500 x 500 measurement points. Other measurement point densities are also possible and may depend on the desired accuracy, computational constraints, data storage, etc.

In some implementations, the ECD 1404 can remove the measurement point 2002 representing the background.

Referring now to FIG. 21 , in one implementation, ECD 1404 (or processor 1802 ) may apply an LBP operation to one or more of measurement points 2002 in base matrix 2010 . For example, ECD 1404 may perform LBP on measurement point-d 2002d comprising an intensity value of 41 . The LBP string for measurement point-d(2002d) is 01110100 based on a span of 1. to perform LBP]. ECD 1404 may track the number of LBP strings for the remaining measurement points 2002r as shown in table 2100 . In some examples, ECD 1404 indicates that 24 of measurement points 2002 have an LBP string of 00000000, 3 have an LBP string of 00000001, and 11 are LBP of 00000010, as indicated in table 2100 , as indicated in table 2100 . You can compute with a string, and 0 has an LBP string of 00000011, and so on. Table 2100 may contain 256 entries for possible strings (ie, 8 bits). In some examples, table 2100 may include 255 entries, by removing entries for "all white" string 11111111 or "all black" string (00000000). In some implementations, the data in table 2100 may be plotted as a histogram representing the number of occurrences (eg, measurement point 2002 ) for a possible LBP string.

In some implementations, based on the distribution of the LBP strings in table 2100 , ECD 1404 can determine one or more unique features. The one or more unique features may be non-zero LBP strings that occur less than a threshold frequency (eg, 5), such as the LBP strings 00000001, 01110100, and 11111100. In another implementation, the one or more unique features may be a minimally occurring non-zero LBP string, such as the LBP string 01110100. In some examples, the ECD 1404 can track the location of the measurement point 2002 with one or more unique features. For example, the ECD 1404 may track the coordinates of the LBP string. In some implementations, the one or more unique features may be the most-occurring non-zero LBP strings.

Referring now to FIG. 22 , in a particular implementation, the ECD 1404 may partition the base matrix 2010 into one or more sub-matrices. Each sub-matrix may contain a macroblock of measurement points. For example, the ECD 1404 can partition the base matrix 2010 into a 5×5 sub-matrix 2202 , a 3×3 sub-matrix 2204 , and a 6×6 sub-matrix 2206 . When computing the LBP string for the measurement points 2002 in the 5 x 5 sub-matrix 2202, the ECD 1402 uses a span of 3 to calculate the measurement points 2002 in the 5 x 5 sub-matrix 2202. We can compute the LBP string of Similarly, when computing the LBP string for measurement points 2002 in 3 x 3 sub-matrix 2204, ECD 1402 uses a span of 2 to measure points in 3 x 3 sub-matrix 2204 ( 2002) can be calculated. When computing the LBP string for the measurement points 2002 in the 6 x 6 sub-matrix 2206 , the ECD 1402 uses a span of 4 to determine the LBP of the measurement points 2002 in the 6 x 6 sub-matrix 2206 . Strings can be counted. Other sizes are possible for the base matrix, sub-matrix and span, as determined by ECD 1402 .

In one example, a 100 x 100 base matrix can be partitioned into six different sized sub-matrices: 10 x 10, 15 x 15, 20 x 20, 30 x 30, 35 x 35, and 40 x 40. Inside each sub-matrix, LBP strings can be computed with different spans around the calculated measurement points to characterize the texture/slope of the area surrounding the cell in different coverage areas. For each sub-matrix size, the span of pixels from the measurement point 2002 to be calculated to each neighboring cell 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 have. For each sub-matrix size, the span of pixels from the measurement point 2002 to be computed to each neighboring cell may be 3, 5, 7, 10, 25 and 40. In some cases, the sub-matrices may overlap. In a particular implementation, the ECD 1404 may apply an LTP computation on the measurement point 2002 .

Referring now to FIG. 23 , in a particular example, the ECD 1404 converts a list of binary features derived from a sampled profile into a data sequence that is extensible and can be tailored to the unique characteristics of each requestor 1450 . while still providing a methodology for object comparison and matching/verification. The binary feature may include three characteristics: the size of the sub-matrix/macroblock, the position of the sub-matrix/macroblock on the sampled profile, and a Uniform Local Binary Pattern (ULBP) assigned to the binary feature. The macroblock may be a “sub-image” of the image of the face 1730 of the requestor 1450 . For example, a macroblock of size 10 is a 10×10 (ie, 100 measurement points 2002) sub-image of the image of face 1730 . In a non-limiting example, the position of the macroblock on the image may be defined by the position of the first pixel within the macroblock, located at the top, left corner of the macroblock. The position of this pixel is defined by two values: the distance from the left border of the face image (x-dimension) and the distance from the top border of the face image (y-dimension). The final feature is ULBP, which is a mathematical methodology for setting scalar values for edge and texture features of pixels starting from the upper left corner of the face 1730 . For each pixel inside the macroblock, a ULBP calculation is performed. If the calculated ULBP matches the ULBP defined for that binary feature, the value of the binary feature is incremented by one. Thus, the maximum value of the binary feature is the size of the defined macroblock (10x10 is 100), and the minimum value is 0. A macroblock may be used in two or more binary features with different ULBP definitions. Also, the location of a macroblock may be used in two or more binary features with different macroblock sizes and ULBP definitions.

23 , a table 2300 shows a 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. Examples of the results of Three arrays can be used to create a biometric template. The first array may be a scalar value in a normalized array. The second array may be sorted such that, in this example, the binary feature 2000 with the highest value moves from position 1 to the top of the second array. A third array compares differences in positions of binary features 2000 in the classification instead of scalar values. If the position in the array for the binary feature 2000 remains the same, the result will reflect a value of zero representing the least likelihood of change in the position and thus the highest value for importance in the authentication method. .

Referring now to FIG. 24 , a method 2400 of transforming a sequence may identify a most unique feature used for identification.

At block 2402, a sampled profile of the face is obtained. For example, the processor 1802 and/or the communication module 1825 may obtain a sampled profile of the face 1730 .

At block 2404, each macroblock/ULBP combination is assigned a unique index. For example, the algorithm module 1856 may assign a unique index (eg, 4) to each macroblock/ULBP combination (eg, macroblock 30×30), the uniqueness of a feature within a face that is not a scalar value. can be analyzed based on

In 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. The first array contains the number of pixels belonging to that ULBP within the macroblock. When analyzing the LBP of a pixel, the ECD 1404 can perform a normal LBP calculation, obtain a histogram with values between 0 and 255, with the result being the highest value (taking the pixel and taking the 8 pixels around it). Then, if each pixel is a standard LBP formula (with a different binary value), the result is given a value of 1. It has a value of 0 if some other LBP is higher than that pixel.

At block 2408, each scalar value is weighted by the size of the macroblock. For example, the algorithm module 1856 may weight each scalar value (eg, 380) by the size (eg, 30×30) of the macroblock.

In block 2410, the first array of scalar values is sorted in descending order. For example, the algorithm module 1856 may sort the elements of the first array of scalar values in descending order, eg, from 1 to 2156.

At block 2412, transform the associated index into a sequence. For example, the algorithm module 1856 may convert the associated index into a sequence or classification order of the data set instead of a scalar value.

At block 2414, transform the sequence into a second array with a scalar value for each unique index, which is the distance (difference in array position, not in mm measurements) of the primary index from the start of the sequence array. For example, the algorithm module 1856 may transform the sequence into a second array with a scalar value for each unique index, which is the position of the primary index from the start of the sequence array, which currently has 2165 elements, but , can be extended depending on the data for each individual.

Referring to FIG. 25 , in a specific implementation, deep learning may be performed on the entire data set to establish a minimum data set required to accurately perform facial recognition. The result of deep learning may be a reduced data set of binary features that is a fraction of the entire data set (eg, 1 %, 2 %, 5 %, 10 %, 20 %, 50 %). Each data element of this data set may be defined by a single value within the entire LBP base matrix and a sub-matrix of a particular size and location. Each data element may be assigned a unique numeric identifier. For the registration process, this static data set is used for each requestor. As each requestor continues to verify his or her face, verification data is collected and, through deep learning, a new set of binary features can be generated for each requestor. The introduction of sequences to the validation algorithm may allow the introduction of this individualized set of binary features and still allow comparisons between different sets. Matching of two sequences can be accomplished by calculating the difference between the two arrays (eg, the difference in positions of the data sets). The higher the sum, the lower the match quality. A perfect object match can substantially have a sum of zero.

With continued reference to FIG. 25 , table 2500 illustrates an example of a validation calculation. The first step in generating the sequence may be to classify the scalar array according to the size of the scalar value. The classification may be performed in descending order. The binary feature unique identification with the largest value is at the top of the classification. By classifying in this way, the binary features that provide the greatest uniqueness (largest scalar size) are at the top of the array. After sorting, the new alignment position is transferred to that third array and the position on the new array is based on the binary feature unique identification. The binary feature unique identifier is 2000 when sorted, and moves to position 1. In the third array, 2000 positions will receive a value of 1. This array can be a unique sequence for an object, and is a new basis for object identification and/or matching.

The sequence matching algorithm determines the quality of a 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 assigned sequentially. To efficiently complete this matching calculation, another array is created containing the distance of each binary feature from the beginning of the sequence. As an example, after classification, a binary feature with a unique identification of one is at position ten of the sorted sequence. In a new array indexed by 0, position 1 will have a value of 10. By structuring the array in this way, the algorithm for performing a match between two of these arrays is identical to the original algorithm. The absolute value of the difference of values with matching indexes is aggregated to produce a single value for the array match. A lower value equals a closer match. A perfect match may be a substantially zero value.

By introducing a sequence array, the matching algorithm can be extended. The sequence array is the second array and the scalar array is the first array (normalized to block size). Extensible may indicate that the fixed list of binary features can now be expanded and contracted as desired to optimize the process of the matching object while also improving the matching process integrity. In this implementation, the sequence array can be dynamically adjusted based on two separate deep learning functions based on the environment of the edge device and the specific individual face applied. A first function may continually try to optimize a default list of binary features that is applied to some or all object matching attempts prior to deployment of the individual list of binary features. A second function may aggregate image data for individual objects and develop an optimized list of binary features for faces over time. If the personalized list of binary features exceeds the matching performance of the default list of binary features via the parallel object matching feature to incoming object image data, then the default list of binary features is personalized with the live object matching function for that object. can be replaced by a list.

In certain implementations, the deep learning engine for both processes may receive sample data in parallel with the live object/face matching function. When image data is received, two lists of binary features may be applied to it. The first list may be a default or customized list of reduced binary features. (eg, the 3 most significant blocks out of 100 of 10 x 10 macroblocks). The output of the applications in this list can be used in the live matching process. The second list may be a full list of binary features including all possible macroblock sizes (eg, all 100 in 10 x 10 macroblocks) that apply to possible image positions for each ULBP. The complete list of binary features may be, for example, 20 times the size of the default or customized list of binary features. The deep learning engine may receive a sequence derived from a complete list of binary features for each object image received into the system. For the default list of binary feature deep learning processes, images can be categorized into training and validation sets. An independent set of default tests can be created, including objects different from the object being trained. For a personalized list of binary features, each object identified by the default or existing personalized list of binary features will be placed into a training, validation and test set. A list of all other objects can be placed in the validation and test set.

A deep learning training session for a default object list of binary features and each individual list of binary features may be run for each new live entry from the object detection system.

Once the eye detection algorithm locates the eye, a face matrix of data is sent to the face recognition algorithm. Once the deep learning algorithm has developed a specific list of binary features, an aggregated list of these binary features for all users of the system can be generated and used to generate sequences. Otherwise, the sequence may be generated using a default set of binary features. The sequence may be sent to a verification algorithm where the matching process may determine the sequence's identity. Once the identity is established, a full list of binary features (all sub-matrices with full LBP histograms) can be generated and sent to that identity data set of the deep learning algorithm. A deep learning algorithm may process the available data sets for that identity (the set may continue to grow with each verification) and generate a modified optimal data set for that identity. The optimized data set can be converted into a sequence and used in the next verification process for that user. At the same time, the aggregated set of binary features can be updated as needed to transform the next face matrix. While the verification process is within a local cluster of devices, deep learning algorithms can occur as background tasks in both local and global clusters.

As the results of the training session refine the list of binary features, the individual lists can be updated in the live verification/matching process. The result may be a matching system that can automatically adjust both the entire population of objects and the specific unique characteristics of each object. The evolution of object data will allow large-scale object matching solutions of over 100,000 objects that can have the same precision as small solutions (<1,000).

Referring now to FIG. 26 , an example of a deep learning process 2600 may rely on a feedback loop and machine learning to refine the identification process.

At block 2602, an image matrix is obtained. For example, the processor 1802 and/or the communication module 1825 may obtain a sampled profile of the face 1730 .

At block 2604 , it is determined whether face specific macroblocks/ULBPs are available. For example, the algorithm module 1856 may determine whether a face specific macroblock/ULBP is available.

At block 2606 , if a custom list of binary features is not available, convert the image to a face generic detection list of binary features. For example, the algorithm module 1856 may convert the image to a face generic/default detection list of binary features if a custom list is not available.

At block 2608, a face detection default sequence is obtained. For example, the algorithm module 1856 can obtain a face detection default sequence if no macroblock is available.

At block 2610 , if a custom list of binary features is available, convert the image to a face detection specific list of binary features. For example, the algorithm module 1856 provides a custom list of binary features (which may be the original 2156 or all new based on that individual (eg, 10 additional 10 x 10 more distinct to that particular individual). If this is available, it can transform the image into a face detection specific list of binary features.

At block 2612, face detection is performed. For example, the algorithm module 1856 may perform face detection.

At block 2614 , eye detection and locating is performed. For example, the algorithm module 1856 may perform eye detection and localization.

At block 2616, transform the image into a full list of binary features. For example, the algorithm module 1856 can transform the image into a full list of binary features. (not just 57, but all 100 macroblocks of 10 x 10, plus all values). Tensor flow is a deep learning engine framework provided by ^{Google TM .} Equations and algorithms for TensorFlow are well known in the art.

At block 2618 , TensorFlow deep learning is used to develop a face detection list of binary features. For example, the algorithm module 1856 can develop a face detection list of binary features using TensorFlow deep learning.

At block 2620, the face detection refinement sequence is fed back. For example, the algorithm module 1856 may feed back a face detection refinement sequence. To refine the list of features, ECD 1404 may take the entire feature set, collect all data for the face, and derive a new refined set that replaces the default set. The calculation of the distance between the positions of the binary features within the two sets may be used as metrics. Each and every feature may be in the same position every time. The better the face data, the smaller the face-to-face comparison will be, and the bigger the difference between this face and all other faces will be.

Referring now to FIG. 27 , the example deep learning process 2600 shown in FIG. 26 may rely on a feedback loop and machine learning to refine the identification process. In particular, instead of the centralized process of deep learning data, which is a target of hacking when stored in a central server on the cloud (TensorFlow single application running on a single central server), the present disclosure uses TensorFlow in a distributed architecture, one central server Rather than being stored in , it can be evaluated at each reader, which protects privacy on every individual device.

At block 2614 , eye detection and locating is performed. For example, the algorithm module 1856 may perform eye detection and localization.

At block 2702 , a face detection specific list of binary features is obtained. For example, the algorithm module 1856 can obtain a face detection specific list of binary features when macroblocks are available.

At block 2704, obtain a facial recognition default list of binary features. For example, the algorithm module 1856 may convert the image to a face generic detection list of binary features if a specific list of binary features for that individual is not available.

At block 2706 , it is determined whether a face specific list of binary features is available. For example, the algorithm module 1856 can determine whether a face-specific list of binary features is available. The custom list of binary features may contain the original 2156 or be all new based on that individual (eg, 10 additional 10 x 10 more distinct for that particular individual).

At block 2708 , if the specific list of binary features is not available, convert the image to a face generic list of binary features. For example, the algorithm module 1856 can convert the image to a face generic list of binary features when a face specific list of binary features is not available.

At block 2710, if a face-specific list of binary features is available, convert the image to a face-specific list of binary features. For example, the algorithm module 1856 may convert the image to a face-specific list of binary features if a face-specific list of binary features is available.

At block 2712, transform the list of binary features into a sequence. For example, the algorithm module 1856 can convert a list of binary features into a sequence.

At block 2714, a verification is performed. For example, the algorithm module 1856 may perform the verification.

At block 2716, transform the image into a full list of binary features. For example, the algorithm module 1856 can transform the image into a full list of binary features.

At block 2718 , TensorFlow deep learning is used to develop a face-specific list of binary features. For example, the algorithm module 1856 can develop a face-specific list of binary features using TensorFlow deep learning. (Not the typical 57 in LBP, but all 100 macroblocks of 10 x 10, including all values). TensorFlow is a deep learning engine framework provided by ^{Google TM .} Equations and algorithms for TensorFlow are well known in the art.

In some examples, gateway 1402 and/or ECD 1404 may include proactive algorithms to identify and reduce “consistency-conflicts” during the validation process. A "consistency-conflict" can occur when more than one requestor has unique facial data that causes a mistake in the verification (ie, false accept or false reject). Gateway 1402 and/or ECD 1404 can proactively look for potential “consistency-conflicts” using the two methods of the present disclosure described below.

A 'layered enrichment' algorithm may allow the requestor to create a binary tree structure for facial data. Binary data may create a hierarchy of data points (eg, LBP strings) based on the uniqueness of the data points. The greater the uniqueness, the higher it is in the tree. Every verification transaction may result in the transfer of binary facial data from ECD 1404 to gateway 1402 . When the gateway 1402 receives the binary facial data, it may start checking again the uniqueness of the binary tree structure of the new data against the stored data of the other requestor 1450 . Upon identifying a potential conflict, gateway 1402 may notify ECD 1404 to escalate a confirmation transaction between two identified requestors 1450 to gateway 1404 . When gateway 1404 receives the escalated transaction, it may perform two advanced algorithms on the data (ie, a proactive conflict identification algorithm and a time-domain trending algorithm). A binary tree can take the most distinct features and establish a hierarchy based on the most distinct and the least distinct. A Prioritized tree can reduce analysis based on 20 instead of 2165, saving time and computing/processing. Based on the branching of the tree from the most distinct feature, it allows the ECD 1404 to determine whether it is the person or not, without going through all 2165 features. The threshold may be empirical and may be derived from testing.

A first advanced algorithm, a proactive conflict identification algorithm, may take binary tree data and analyze the facial data based on its location in the binary tree. Binary tree data may be weighted. If the weighting of the binary data does not yield sufficient differentiation of the data, the gateway 1402 may extend the verification process to find the appropriate differentiation data.

In some implementations, the requestor's facial characteristics may change continuously. If biometric data of a face, such as face 1730 of requestor 1450 , remains static, identification of the face may ultimately lead to false rejections and may require re-registration of face 1730 . Dynamic adjustments to the biometric data may be continuously applied, and the algorithm may determine if the proposed change does not introduce a different face in the biometric data resulting in false acceptance of face 17303 . Each iteration of the biometric data, the algorithm must determine the amount of biometric data to retain to ensure the next successful identification of the requestor 1450 and the amount to change to ensure long-term identification of the requestor 1450 . The determination may be made by a combination of time-domain analysis over which changes are regressed to ensure linearity and facial zone analysis to determine if the domain of change is reasonable.

The use of this dynamic data capability leads to the second algorithm, the time-domain trending algorithm. Gateway 1402 maintains a history of requestor 1450's binary data, and compares the history of requestor 1450's binary data to face 1730 and the rate at which these features change to evaluate which features are changing. time-domain based analysis of The gateway 1402 can track the evolution of the face and use this data to extrapolate and augment the unique differences identified in the face 1730 of the requestor 1450 to establish key or reference features and later highlight them. and evaluate the rate of change of these key features for face 1730 . The gateway 1402 then uses time-domain based analysis to determine whether these changes occur over a period of time to suggest natural changes in the face 1730 , assuring further analysis of the artificial intelligence of the face 1730 . It is determined whether it is a change or whether the history of the binary data of the face 1730 and the requestor 1450 does not match. Gateway 1402 may identify unique differences and verify that the differences identified in the confirmation request are consistent with trends over time. In some implementations, gateway 1402 may take into account the lifestyle and routine routines of requestor 1450 in a time-domain trending algorithm. For example, if requestor 1450 enjoys outdoor activities, the algorithm may consider increased tanning during the summer months. The proactive conflict identification algorithm combined with the time-domain trending algorithm allows the method of this embodiment of the present invention to maintain high performance in large population '1:N' solutions.

In some implementations, the time-domain trending algorithm may reduce the change over time of the biometric data to one or more equations (eg, a curve fitting algorithm) that characterize the change over time. The linearity of one or more equations over time may determine the integrity of the change. If a particular domain of change is expressed as a discrete function, then the change can be represented as a potential threat (eg, spurious, incorrect match). The change may also be evaluated based on the physical location of the pixel box on the face. Positions may be weighted based on the probability of a change in that region of the face. Significant changes in the area of low probability change will be marked as potential threats.

Referring now to FIG. 28 , with reference to the figure, in some implementations, a method 2800 of constructing a biometric template may be performed by the ECD 1404 .

In an optional implementation, method 2800 may perform a registration process. The registration process can include the ECD 1404 capturing a plurality of images (eg, 5, 10, 15, 20, 30, 50) of the face 1730 . The ECD 1404 may convert the plurality of images into biometric data as described above.

At block 2802, the incident invisible light is emitted. For example, the illumination source 1704 may emit incident invisible light. Invisible light may include near IR or UV. In some implementations, the illumination source 1704 can emit visible light.

At block 2804, detected invisible light is detected, wherein the detected invisible light includes reflected invisible light and emitted invisible light. For example, the optical sensor 1702 may detect detected invisible light, wherein the detected invisible light includes reflected invisible light and emitted invisible light. In some implementations, optical sensor 1702 can detect IR light reflected from face 1730 of requestor 1450 and/or IR light emitted due to heat of requestor 1450 .

At block 2806 , a sampled profile is generated that includes a plurality of sampling points having a plurality of feature values associated with the detected invisible light. For example, the algorithm module 1856 can generate a sampled profile, such as sampled profile 2000 , that includes a plurality of sampling points having a plurality of feature values associated with the detected invisible light.

At block 2808 , one or more macroblocks each comprising a subset of a plurality of sampling points are identified. For example, the algorithm module 1856 can identify one or more macroblocks each comprising a subset of a plurality of sampling points. The one or more macroblocks may be 10×10 macroblocks, 15×15 macroblocks, 20×20 macroblocks, 30×30 macroblocks, 35×35 macroblocks, and 40×40 macroblocks. Other sizes are also possible. In some implementations, ECD 1404 can identify 2165 macroblocks with 2165 associated dimensions.

At block 2810, a local pattern value is selected. For example, the algorithm module 1856 may select a local binary pattern value of 20. In another example, the algorithm module 1856 may select a local ternary pattern value.

At block 2812, count the number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks. For example, the algorithm module 1856 can calculate the number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks.

At block 2814, a first array comprising the plurality of weights is generated by calculating the plurality of weights based on the number of occurrences of the local pattern value and the corresponding sizes of the one or more macroblocks. For example, the algorithm module 1856 may generate a first array including the weights indicated in the table 2300 (ie, [0.30 0.54 0.15 0.42 0.62 0.46 0.53]).

At block 2816, a unique index is assigned to each of the plurality of weights. For example, the algorithm module 1856 may assign a unique index (ie, 1, 2, 3, 4, 5, 6, and 7) to each of the plurality of weights indicated in the table 2300 . Unique index 1 is 0.30, 2 to 0.54, 3 to 0.15... are assigned to weights such as

At block 2818, a second array of unique indices is generated by ranking the plurality of weights. For example, the algorithm module 1856 may rank the second array of unique indices by ranking the plurality of weights, such as sequence 5, 2, 7, 6, 4, 1, 3 in table 2300 . can create In some implementations, the second array/sequence may indicate a ranking of weights from highest to lowest. The first number in the sequence is 5 because the weight associated with the unique index of 5 (ie, 0.62) is the highest among the elements in the first array.

At block 2820 , a third array including a plurality of ranking distances is generated. For example, the algorithm module 1856 may generate a third array including a plurality of rank distances, such as the array [5 1 6 4 0 3 2] stored in the table 2300 . The rank distance may represent a numerical difference between the ranks of the highest weight (eg, 0.62 - rank 1) and the current weight (eg, 0.30 - rank 6). Thus, the ranking distance between 0.62 and 0.30 may be 5 (ie, 6 - 1).

In some optional implementations, during the verification process, biometric data based on the sampled profile (“requester biometric data”) may be compared with biometric data of a plurality of images captured during the enrollment process (“enrollment biometric data”). can If the match percentage exceeds a threshold percentage (eg, 20, 30, 40, 50, 60, 70, or 80), one of the ECD 1404 and/or gateway 1402 confirms that the requestor biometric data is a positive match It can be determined that this is a success.

In certain aspects, the ECD 1404 and/or one of the gateways 1402 may determine the face 1703 due to, for example, tanning, aging, injury, mood changes, weight changes, beard changes, cosmetic use, accessories, or other causes. ) can be adjusted over time to accommodate any change in enrollment biometric data. During each verification, one of the ECD 1404 and/or gateway 1402 may adjust a portion of the enrollment biometric data (eg, 20%, 30%, 40%, 50%). Other portions of the biometric data may remain unchanged.

It will be appreciated that various implementations of the above-disclosed and other features and functions, or alternatives or variations thereof, may be incorporated into many other different systems or applications, preferably by those skilled in the art. In addition, various alternatives, modifications, variations or improvements not presently anticipated or anticipated herein may subsequently be made by those skilled in the art, and are also intended to be encompassed by the appended claims.

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 are within the scope of the claims. As used in this description, the term “exemplary” means “serving as an example, illustration, or illustration,” rather than “be advantageous over other examples” or “preferred”. The detailed description includes specific details for the purpose of providing an understanding of the described technology. However, these techniques may be practiced without these specific details. For example, changes may be made in the function and arrangement of the elements described without departing from the scope of the present disclosure. In addition, various examples may omit, substitute, or add various procedures or components as appropriate. For example, methods described may be performed in an order different from that described, and various steps may be added, omitted, or combined. Also, features described for some examples may be combined in other examples. In some instances, well-known structures and mechanisms are shown in block diagram form in order to avoid obscuring the concepts of the described examples.

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 include voltage, current, electromagnetic wave, magnetic field or magnetic particle, light field or optical particle, computer-readable It may be represented as computer-executable code or instructions stored on a medium, or any combination thereof.

The various illustrative blocks and components described in connection with the disclosure herein may include a processor, digital signal processor (DSP), ASIC, FPGA or other programmable logic device, discrete gate or It may be implemented or implemented in a specially-programmed device such as, but not limited to, transistor logic, discrete hardware components, or any combination thereof. A specially programmed processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. Additionally, a specially-programmed processor may be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in combination with a DSP core, or any other such configuration.

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 this disclosure and the appended claims. For example, due to the nature of software, the functions described above may be implemented using software executed by a specially programmed processor, hardware, firmware, hardwiring, or any combination thereof. Features implementing a function may also be physically located in various positions, including being distributed such that portions of the function are implemented in different physical locations. Also, as used herein, including in the claims, "or" as used in a list of items beginning with "at least one of" means, for example, A list of “at least one of A, B, or C” refers to a list that is disjunctive to mean A or B or C or AB or AC or BC or ABC (ie, A and B and C).

Computer-readable media includes both computer storage media and communication media including any medium that enables 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 be used to store or carry desired program code means in the form of instructions or data structures and can be accessed by a general or special purpose computer, or a general or special purpose processor. RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage device, or any other medium. 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 coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio and microwave, coaxial Cables, fiber optic cables, twisted pair, DSL, or wireless technologies such as infrared, radio and microwave are included in the definition of a medium. As used herein, disk and disk are compact disk (CD), laser disk (disc), optical disk (disc), digital versatile disk (DVD). , floppy disks and Blu-ray disks, where disks usually reproduce data magnetically, while disks optically reproduce data using lasers. Combinations of the above are also included within the scope of computer-readable media.

Claims (20)

an edge capture device (ECD);
an illumination source configured to emit invisible light incident towards the requestor;
an optical sensor configured to detect the detected invisible light;
a processing circuit operatively coupled to an illumination source and an optical sensor;
The processing circuit is:
generating a sampled profile comprising a plurality of sampling points having a plurality of intensities of the detected invisible light;
identify one or more macroblocks each comprising coordinates and a subset of the plurality of sampling points;
calculate the number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks;
create a first array comprising a plurality of weights each calculated by dividing a corresponding number of occurrences of a local pattern value by a size of a corresponding macroblock of the one or more macroblocks;
assigning a unique index to each of the plurality of weights;
create a second array of unique indexes by ranking the unique indexes based on the associated weights;
generate a third array comprising a plurality of rank distances between a highest weight of the plurality of weights and each weight of the plurality of weights; and
configure the requestor's biometric template based on the third array and coordinates;
composed,
Edge Capture Device (ECD). According to claim 1,
sending an interrogation signal to the requestor's proximity card; and
further comprising a scanner configured to receive a response signal comprising an identification sequence associated with the proximity card;
Edge Capture Device (ECD). According to claim 1,
further comprising a keypad configured to receive a password entered by the requestor; the processing circuitry is configured to determine whether the password entered by the requestor matches any one of the plurality of authorized passwords;
Edge Capture Device (ECD). According to claim 1,
further comprising a microphone configured to receive voice input by the requestor, wherein the processing circuitry determines whether the input voice matches any one of a plurality of authorized voices;
Edge Capture Device (ECD). According to claim 1,
the illumination source is further configured to emit at least one of ultraviolet light or infrared light;
Edge Capture Device (ECD). According to claim 1,
further comprising a modem configured to transmit the biometric template to the remote gateway;
Edge Capture Device (ECD). According to claim 1,
The processing circuit is:
Bus;
a plurality of processing boards each having a processor and a memory; and
further comprising a controller configured to distribute copies of the biometric template to the plurality of processing boards;
the plurality of processing boards are configured to simultaneously compare the copy of the biometric template with the authorized biometric template stored in the memory;
Edge Capture Device (ECD). According to claim 1,
The optical sensor comprises a wide-angle lens,
Edge Capture Device (ECD). According to claim 1,
the optical sensor is further configured to detect at least one eye of the requestor;
Edge Capture Device (ECD). According to claim 1,
the optical sensor is further configured to detect at least one of reflected infrared light, reflected ultraviolet light, or emitted infrared light from the requestor;
Edge Capture Device (ECD). According to claim 1,
further comprising a display configured to display an image of the requestor's face in visible light, wherein the image is captured by the optical sensor;
Edge Capture Device (ECD). 12. The method of claim 11,
the display is further configured to show an alignment mark for aligning the requester's face to the optical sensor;
Edge Capture Device (ECD). The infrastructure is:
access-controlled assets; and
an edge capture device (ECD) associated with an access-controlled asset;
The edge capture device (ECD) comprises:
an illumination source configured to emit invisible light incident towards the requestor;
an optical sensor configured to detect the detected invisible light; and
processing circuitry operatively coupled to the illumination source and the optical sensor;
the processing circuit;
generating a sampled profile comprising a plurality of sampling points having a plurality of intensities of the detected invisible light;
identify one or more macroblocks each comprising coordinates and a subset of the plurality of sampling points;
calculate the number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks;
create a first array comprising a plurality of weights each calculated by dividing a corresponding number of occurrences of a local pattern value by a size of a corresponding macroblock of the one or more macroblocks;
assigning a unique index to each of the plurality of weights;
create a second array of unique indexes by ranking the unique indexes based on the associated weights;
generate a third array comprising a plurality of rank distances between a highest weight of the plurality of weights and each weight of the plurality of weights; and
configured to configure the requestor's biometric template based on the third array and coordinates;
infrastructure. 14. The method of claim 13,
Access-controlled assets include at least one of a safe, lock, security door, security gate, equipment, machinery, computing device, digital storage device, database, and file;
infrastructure. 14. The method of claim 13,
receive a positive match indication from an edge capture device (ECD); and
in response to the positive match indication, further comprising a gateway configured to send a positive match signal to the access controlled asset associated with the edge capture device (ECD) to allow the requestor to access the access controlled asset;
infrastructure. 14. The method of claim 13,
Edge capture devices (ECDs) are:
sending an interrogation signal to the requestor's proximity card; and
further comprising a scanner configured to receive a response signal comprising an identification sequence associated with the proximity card;
infrastructure. 14. The method of claim 13,
the illumination source is further configured to emit at least one of ultraviolet light or infrared light;
infrastructure. 14. The method of claim 13,
The processing circuit is:
Bus;
a plurality of processing boards each having a processor and a memory; and
further comprising a controller configured to distribute copies of the biometric template to the plurality of processing boards;
the plurality of processing boards are configured to simultaneously compare the copy of the biometric template with the authorized biometric template stored in the memory;
infrastructure. 14. The method of claim 13,
The optical sensor comprises a wide-angle lens,
infrastructure. The system is:
a plurality of edge capture devices (ECDs) comprising a first edge capture device (ECD) and the remaining edge capture devices (ECD);
a plurality of access controlled entry points each associated with an edge capture device (ECD) of the plurality of edge capture devices (ECD);
a gateway communicatively coupled to a plurality of access controlled entry points and a plurality of edge capture devices (ECDs);
The first edge capture device (ECD) comprises:
an illumination source configured to emit invisible light incident towards the requestor;
an optical sensor configured to detect the detected invisible light;
a memory for storing a plurality of authorized biometric templates; and
processing circuitry operatively coupled to the illumination source and the optical sensor;
The processing circuit is:
generating a sampled profile comprising a plurality of sampling points having a plurality of intensities of the detected invisible light;
identify one or more macroblocks each comprising coordinates and a subset of the plurality of sampling points;
calculate the number of occurrences of the local pattern value within each subset of the plurality of sampling points for each of the one or more macroblocks;
create a first array comprising a plurality of weights each calculated by dividing a corresponding number of occurrences of a local pattern value by a size of a corresponding macroblock of the one or more macroblocks;
assigning a unique index to each of the plurality of weights;
create a second array of unique indexes by ranking the unique indexes based on the associated weights;
generating a third array including a plurality of rank distances between a highest weight among the plurality of weights and each weight among the plurality of weights;
construct the requestor's biometric template based on the third array and coordinates; and
send the configured biometric template to the gateway via the modem in response to not finding a local match for the configured biometric template;
The gateway is:
receive a configured biometric template of the requestor from a first edge capture device (ECD) of the plurality of edge capture devices (ECD);
comparing the configured biometric template with a plurality of biometric templates;
identify a positive match between the configured biometric template and one of the plurality of biometric templates;
in response to the positive match, send a positive match signal to the first edge capture device (ECD) to allow the requestor to access an access controlled entry point associated with the first edge capture device (ECD);
system.

Images