Biometric authentication by vascular examination

Abstract

Presented here are approaches to authenticating unknown individuals based on temporal changes in the spatial characteristics and directionality of blood flow through blood vessels. At a high level, these approaches rely on monitoring vascular dynamics to recognize unknown individuals. For example, an authentication platform may examine a digital image of an anatomical region and establish how the characteristics of the vasculature within the anatomical region have changed as a result of deformation. Example characteristics include the location, size, volume, and pressure of the blood vessels contained within the vasculature, as well as the velocity and acceleration of blood flowing through the vasculature.

Description

[Technical Field]

[0001] (CROSS-REFERENCE TO RELATED APPLICATIONS) This application claims priority to U.S. Provisional Application No. 63 / 127,054, entitled "Vein Map Authentication with Image Sensor," filed December 17, 2020, and incorporated herein by reference in its entirety.

[0002] (Technical field) The present disclosure relates to biometric authentication in computer security, and more particularly to techniques in which physiological characteristics are examined to enable minimally disruptive authentication. [Background technology]

[0003] Biometric authentication procedures verify an individual's identity through a biometric. The term "biometric" refers to a physical or behavioral characteristic that can be used as a means of verifying identity. Biometrics are convenient because they are difficult to spoof and do not require the corresponding individual to remember passwords or manage tokens. Instead, the authentication mechanism is part of the person.

[0004] Fingerprints have historically been the most common biometric modality. However, as technology evolves, other biometric modalities have emerged. For example, vascular pattern recognition (also known as "vein pattern recognition") uses near-infrared light to create images of subcutaneous blood vessels (or simply "vessels"). Collectively, these subcutaneous vessels are called a "vascular pattern" or "vein map," which can be used for authentication. Vein pattern recognition is promising because vascular patterns are not only unique to a given individual, but also change little with age.

[0005] Vein pattern authentication typically involves identifying and then analyzing the vascular pattern along the back of the hand. For example, near-infrared light generated by a light-emitting diode (LED) is shone onto the back of the hand and penetrates the skin. Due to differences in absorbance between blood vessels and other tissues, the near-infrared light is reflected back into the skin at different depths. Based on an analysis of the reflected near-infrared light, the vascular pattern can be inferred, and features such as branching positions and angles can be identified from the vascular pattern (which can then be used for authentication).

[0006] Because vein patterns are difficult to reproduce, vein pattern authentication has attracted attention as a contactless biometric authentication method that is relatively immune to counterfeiting. Furthermore, vein pattern authentication offers significant advances over other biometric authentication approaches in terms of false acceptance rate (also known as "false positive rate") and false rejection rate (also known as "false negative rate"). However, vein pattern authentication has several drawbacks. For example, individuals tend to be uncomfortable exposing their bodies to the unfamiliar light source typically associated with the scanning equipment required for vein pattern authentication. Furthermore, this scanning equipment can be difficult (if not impossible) to install in some environments and can be prohibitively expensive for many merchants. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 contains a high-level diagram of a conventional authentication procedure in which an unknown person is prompted to present their hand to a vascular scanner.

[0008] [Figure 2] FIG. 2 shows how each cardiac cycle is represented as a peak in a photoplethysmogram.

[0009] [Figure 3] FIG. 3 contains a high-level representation of a system that can be used to authenticate the identity of an unknown person whose vasculature can be imaged.

[0010] [Figure 4] FIG. 4 shows an example of an electronic device that can implement an authentication platform designed to authenticate the identity of an unknown person based on image data generated by an image sensor.

[0011] [Figure 5] FIG. 5 illustrates how the underlying vasculature in an anatomical region (here, the face) can be altered by the execution of a gesture that causes physical deformation of the surrounding tissue.

[0012] [Figure 6A] 6A-C illustrate several different approaches to determining, calculating, or obtaining the pulse wave. [Figure 6B] 6A-C illustrate several different approaches to determining, calculating, or obtaining the pulse wave. [Figure 6C] 6A-C illustrate several different approaches to determining, calculating, or obtaining the pulse wave.

[0013] [Figure 7] FIG. 7 includes a flow diagram of a procedure for authenticating a user of the authentication platform based on analysis of visual evidence of vascular dynamics in an anatomical region.

[0014] [Figure 8] FIG. 8 contains a flow diagram of the processes performed by the authentication platform during the use phase (also called the "implementation phase").

[0015] [Figure 9] FIG. 9 contains a flow diagram of another process for determining whether to authenticate an unknown person as a predetermined individual through vascular studies.

[0016] [Figure 10]FIG. 10 contains a visual illustration of the process by which the authentication platform determines whether to authenticate an unknown person as a given individual.

[0017] [Figure 11] FIG. 11 contains a flow diagram of a process for creating a model that is trained to predict blood flow through the vasculature of an anatomical region as it is deformed.

[0018] [Figure 12] FIG. 12 is a block diagram illustrating an example of a processing system in which at least some of the operations described herein may be implemented.

[0019] Various features of the technology described herein will become more apparent to those skilled in the art from a consideration of the detailed description in conjunction with the drawings. Embodiments are shown in the drawings by way of example, and not limitation. While the drawings depict various embodiments for illustrative purposes, those skilled in the art will recognize that alternative embodiments may be employed without departing from the principles of the technology. Thus, while specific embodiments are shown in the drawings, the technology is susceptible to various modifications. DETAILED DESCRIPTION OF THE INVENTION

[0020] To enroll in an authentication program that relies on matching vein patterns, an individual (also referred to as a "user") may first be prompted to present their hand to a vascular scanner. The term "vascular scanner" may be used to refer to an imaging device that includes (i) an emitter operable to emit electromagnetic radiation (e.g., in the near-infrared range) into the body, and (ii) a sensor operable to sense the electromagnetic radiation reflected by physiological structures within the body. Typically, a digital image is created based on the reflected electromagnetic radiation, which serves as a reference template. At a high level, the reference template represents a "ground truth" vascular pattern that can be used for authentication.

[0021] FIG. 1 includes a high-level diagram of a conventional authentication procedure in which an unknown person is prompted to present their hand to a vascular scanner. As shown in FIG. 1, the vascular scanner emits electromagnetic radiation onto the hand and then creates a digital image (also called a "scan") based on the electromagnetic radiation reflected by the hand's blood vessels. This image represents the hand's vascular pattern and can therefore be matched to a reference template created for a given individual during the enrollment phase (also called the "registration phase"). If the digital image matches the reference template, the unknown person is authenticated as the given individual. However, if the digital image does not match the reference template, the unknown person is not authenticated as the given individual.

[0022] Vein pattern matching has become an attractive option for biometric authentication because vein scanners do not require direct contact with the body while the scan is being performed. However, vein pattern matching has been shown to be susceptible to spoofing. As an example, Jan Krissler and Julian Albrecht demonstrated at the 2018 Chaos Communication Congress that vein scanners can be bypassed using a fake hand made out of wax. While spoofing is unlikely to be successful in most real-world conditions, concerns related to vulnerability may hinder the reliable adoption of biometric authentication techniques.

[0023] Therefore, presented herein are approaches to authenticating unknown persons based on changes over time in the spatial properties and directionality of blood flow through blood vessels. At a high level, these approaches rely on monitoring vascular dynamics to recognize unknown persons. The term "vascular dynamics" refers to changes in the vasculature and its properties caused by, for example, deformation of the surrounding subcutaneous tissue due to the performance of a gesture. Examples of vascular properties include the location, size, volume, and pressure of blood vessels, as well as the velocity and acceleration of blood flowing through the vessels.

[0024] As explained further below, these approaches to authentication can be considered a form of ranged photoplethysmogram (PPG) monitoring. The term "photoplethysmogram" refers to an optically obtained plethysmogram that can be used to detect changes in blood volume within subcutaneous tissue. With each cardiac cycle, the heart pumps blood to the body's periphery. This pressure pulse attenuates somewhat by the time the blood reaches the skin, but is sufficient to detectably dilate blood vessels within the subcutaneous tissue. The volume change caused by the pressure pulse can be detected by shining light onto the skin and then measuring the amount of light transmitted or reflected by an image sensor. In PPG, each cardiac cycle appears as a peak, as shown in Figure 2.

[0025] Historically, pulse oximeters have been commonly used for PPG monitoring. Pulse oximeters typically contain at least one light-emitting diode (LED) that emits light toward a photodiode through a body part, such as a fingertip or earlobe. However, PPG can also be obtained through analysis of digital images of a target anatomical region. In such scenarios, pressure pulses can be indicated by subtle changes in the color of the skin and subcutaneous tissue. However, establishing subtle characteristics of pressure pulses can be challenging. For example, the timing and phase of pressure pulses can be difficult to detect through analysis of digital images of the face due to the complex structure of the underlying vasculature and the complex influence of body posture and facial expression. Deformation of the subcutaneous tissue caused by body posture and facial expression affects the resistance of blood flow through the facial venous network, which in turn affects the signal generated by an image sensor observing the subcutaneous tissue dominated by the venous network. Although the relationship between deformation of the subcutaneous tissue and the signal generated by the image sensor is difficult to quantify, the deformation has a predictable effect on the signal (and can therefore be used as an authentication measure).

[0026] To determine whether to authenticate an unknown person as a predetermined individual, the authentication platform (also referred to as an "authentication system") may determine the degree to which the unknown person's vascular dynamics compare to the vascular dynamics of the predetermined individual. For example, assume that an unknown person wishes to authenticate themselves as the predetermined individual. In such a scenario, the unknown person may be prompted to perform a gesture that causes deformation of the subcutaneous tissue (and therefore the vasculature) in an anatomical region. The gesture may be related to the anatomical region. For example, if the authentication platform examines the vascular dynamics of the face, the unknown person may be prompted to smile or frown, and if the authentication platform examines the vascular dynamics of the hands, the unknown person may be prompted to clasp their hands.

[0027] While the unknown person performs a gesture, a camera on the electronic device may generate digital images of the anatomical region. For example, the camera may generate digital images in rapid succession at a predetermined cadence. As another example, the camera may generate a video of the anatomical region, in which case the digital images may represent frames of the video. Based on analysis of the digital images, the authentication platform may generate a "biometric signature" or "vascular signature" of the unknown performer. For example, the authentication platform may generate a vein model that programmatically indicates how the vasculature deformed while the gesture was performed. At a high level, the vein model identifies how spatial properties of the vasculature changed as a result of the gesture. Alternatively, the authentication platform may estimate metrics of vascular properties based on analysis of the digital images. For example, the authentication platform may attempt to quantify how the directionality of blood flow through the vasculature changed as a result of the gesture.

[0028] The authentication platform can then compare the biometric signature with an enrolled biometric signature (also referred to as a "reference biometric signature") associated with the predetermined individual to determine whether the unknown person should be authenticated as the predetermined individual. For example, if the authentication platform generates a vein model programmatically indicating how the unknown person's vasculature deforms when a gesture is performed, the authentication platform may (i) obtain a vein map associated with the predetermined individual, and (ii) estimate, based on the vein map, the deformation that would be expected during the performance of the gesture by the predetermined individual. As another example, if the authentication platform estimates metrics that indicate how vascular characteristics change while a gesture is performed, the authentication platform may (i) obtain a vein map associated with the predetermined individual, and (ii) estimate, based on the vein map, the metric that would be expected during the performance of the gesture by the predetermined individual. As described further below, the vein map may be stored in a digital profile that includes information about the vasculature of the predetermined individual. For example, the digital profile may include vein maps for different anatomical regions, metrics for different vascular characteristics, etc.

[0029] In summary, the authentication platform may present a notification instructing a person to be authenticated to perform a gesture that causes a deformation of an anatomical region, acquire a digital image of the anatomical region generated by an electronic device as the person performs the gesture, estimate characteristics of blood flow through subcutaneous blood vessels in the anatomical region based on the digital image, and then determine whether to authenticate the person as a predetermined individual based on a comparison of the estimated characteristics with a digital profile associated with the predetermined individual. The estimated characteristics may be, for example, directionality, velocity, volume, phase, or pressure of blood flowing through the subcutaneous blood vessels.

[0030] Because the information being "read" resides within the body, authentication based on biometric signatures offers many of the same advantages as vein pattern matching: high accuracy, reliability, and consistency. However, this approach is easier to implement because specialized equipment (e.g., a vascular scanner) is not required. Instead, authentication can be performed based on analysis of digital images generated by an electronic device. While the electronic device may include specialized software, firmware, or hardware, general-purpose, standardized hardware (e.g., digital image sensors used in mobile phones, tablet computers, etc.) may be sufficient to capture high-quality digital images.

[0031] At a high level, the authentication platform is designed to facilitate an approach in which an individual can combine a vein map, which serves as an authentication factor, with measured blood flow, which serves as a PPG signal. In this manner, authentication can be achieved using electronic devices that cannot detect individual blood vessels but can detect spatially resolved PPG signals (e.g., through analysis of digital images). Specifically, the approach described herein (i) enables highly secure authentication without requiring specialized equipment; (ii) enables authentication based on knowledge factors (e.g., of variants) and biometric information for unknown and predetermined individuals; and (iii) enables authentication that is robust against spoofing and theft because new variants can be easily identified and claimed.

[0032] For illustrative purposes, embodiments may be described in the context of monitoring vasculature in a given anatomical region. For example, embodiments may be described in the context of examining digital images of a face, palm, or finger. However, the approaches described herein may be equally applicable to vasculature in other parts of the human body.

[0033] Although not required, implementations are described below in the context of instructions executable by an electronic device. The term "electronic device" is generally used interchangeably with the term "computing device" and may therefore be used to refer to computer servers, point-of-sale (POS) systems, tablet computers, wearable devices (e.g., fitness trackers and watches), mobile phones, etc.

[0034] Although aspects of the technology, such as particular modules, may be described as being performed exclusively or primarily by a single electronic device, some implementations are performed in a distributed environment in which modules are shared among multiple electronic devices linked through a network. For example, an unknown person may be asked to initiate an authentication procedure with a mobile phone that generates a digital image of an anatomical region, but the decision whether to authenticate the unknown person may be made by an authentication platform residing on a computer server to which the mobile phone sends the digital image. [term]

[0035] References herein to "one embodiment" or "one embodiment" mean that the described feature, function, structure, or characteristic is included in at least one embodiment of the technology. Appearances of such phrases do not necessarily refer to the same embodiment, nor do they necessarily refer to mutually exclusive alternative embodiments.

[0036] Unless the context clearly requires otherwise, the terms "comprises," "comprising," and "consisting of" are to be interpreted in an inclusive sense (i.e., in the sense of "including but not limited to"), rather than in an exclusive or exhaustive sense. The term "based on" is also to be interpreted in an inclusive sense, rather than in an exclusive or exhaustive sense. Thus, unless otherwise specified, the term "based on" is intended to mean "based at least in part on."

[0037] The terms "connected," "coupled," and variations thereof are intended to include any connection or coupling between two or more elements, either directly or indirectly. The connection / coupling may be physical, logical, or a combination thereof. For example, objects may be electrically or communicatively coupled to each other despite not sharing a physical connection.

[0038] The term "module" may refer to a software component, a firmware component, or a hardware component. A module is typically a functional component that generates one or more outputs based on one or more inputs. As an example, a computer program may include multiple modules responsible for completing different tasks, or a single module responsible for completing all tasks.

[0039] When used in connection with a list of multiple items, the term "or" is intended to cover all of the following interpretations: any of the items in the list, all of the items in the list, and any combination of the items in the list.

[0040] The order of steps performed in any of the processes described herein is exemplary. However, steps may be performed in various orders and combinations, provided that this does not violate physical feasibility. For example, steps may be added to or deleted from the processes described herein. Similarly, steps may be interchanged or reordered. As such, any process description is intended to be open-ended. [Authentication by analyzing blood vessel information]

[0041] Presented herein is an authentication platform that utilizes vascular dynamics as biometric proof that an unknown person is a predetermined individual. As described further below, the spatial properties and directionality of blood flow through blood vessels in an anatomical region can be estimated based on an analysis of one or more digital images of the anatomical region. When the surrounding subcutaneous tissue deforms (e.g., by performing a gesture), the spatial properties and directionality of blood flow change, and these changes can be used to determine whether to authenticate the unknown person as a predetermined individual.

[0042] The authentication platform can be used to secure biometric-driven transactions, such as payments approved through a hands-free interface. For example, suppose an unknown person wishes to authenticate themselves to complete a transaction. Rather than prompting the unknown person to place a body part (e.g., their hand) near a vascular scanner, biometric authentication can instead be performed using the electronic device the unknown person used to initiate the transaction. For example, if the unknown person initiates the transaction using a mobile phone they carry, the mobile phone may generate a digital image of an anatomical region (e.g., a face) that can be analyzed by the authentication platform. As described below, the authentication platform may reside on the mobile phone or another electronic device (e.g., a computer server) to which the mobile phone is communicatively connected. This authentication approach relies on analyzing blood vessels under the skin, but the mobile phone does not need to contact the skin. Instead, the unknown person need only be prompted to generate a digital image of the anatomical region for authentication purposes using the mobile phone. In this way, the authentication platform may enable a person to authenticate themselves in a minimally disruptive manner by relying on information about vascular dynamics.

[0043] In some embodiments, the authentication platform operates independently to authenticate the identity of an unknown person, while in other embodiments, the authentication platform operates in conjunction with other systems. For example, a payment system may interface with the authentication platform to ensure that a transaction is completed in a secure and hassle-free manner. As an example, the authentication platform may facilitate contactless payment procedures in which an unknown person is authorized to initiate or complete a transaction by making a part of their body imageable. As described above, an unknown person may make a part of their body imageable simply by placing the part within the field of view of the camera of the electronic device. The electronic device is typically, but not necessarily, the one used to initiate or complete the transaction.

[0044] While embodiments may discuss authentication in the context of initiating or completing a transaction, it should be noted that authentication is useful in a variety of contexts. For example, assume a group of individuals are invited to a network-accessible conference where confidential information will be shared. Each person attempting to enter the network-accessible conference may be required to be authenticated by an authentication platform before being granted access. [Authentication Platform Overview]

[0045] FIG. 3 includes a high-level representation of a system 300 that can be used to authenticate the identity of an unknown person whose vasculature is imageable. As shown in FIG. 3, the system 300 includes an authentication platform 302, which may have access to a user interface (UI) 304, an image sensor 306, a light source 308, a processor 310, or any combination thereof. As described further below, these elements of the system 300 can be incorporated into the same electronic device or distributed among multiple electronic devices. For example, the authentication platform 302 may reside, in part or in whole, on a network-accessible server system, while the UI 304, the image sensor 306, the light source 308, and the processor 310 may reside on a separate electronic device responsible for generating the digital image of the unknown person.

[0046] The UI 304 represents an interface through which the unknown person can interact with the system 300. The UI 304 may be a voice-driven graphical user interface (GUI) displayed on the display of an electronic device. Alternatively, the UI 304 may be a non-voice-driven GUI displayed on the display of an electronic device. In such an embodiment, the UI 304 may visually indicate the body part that should be presented for authentication purposes. For example, the UI 304 may visually prompt the unknown person to position their body so that anatomical regions can be observed by the image sensor 306. As one example, the UI 304 may include a “live view” of a digital image generated by the image sensor 306 so that the unknown person can easily align their face with the image sensor 306. As another example, the UI 304 may present an illustration to indicate where the unknown person should place their hand so that the palm or back of the hand can be imaged using the image sensor 306. Additionally, the UI 304 may present an authentication decision that is ultimately made by the authentication platform 302.

[0047] Image sensor 306 may be any electronic sensor capable of detecting and transmitting information to generate a digital image. Examples of image sensors include charge-coupled device (CCD) sensors and complementary metal-oxide semiconductor (CMOS) sensors. Image sensor 306 may be implemented in a camera module (or simply "camera"). In some embodiments, image sensor 306 is one of multiple image sensors implemented in an electronic device. For example, image sensor 306 may be included in a front- or rear-facing camera built into a mobile phone.

[0048] Typically, digital images are generated by the image sensor 306 in conjunction with ordinary visible light. However, the image data representing the digital image may be in a variety of formats, color spaces, etc. For example, the image sensor 306 may be implemented in a camera designed to output image data according to the Red-Green-Blue (RGB) color model, such that each pixel is assigned a separate chromaticity value for red, green, and blue. As another example, the image sensor 306 may be implemented in a camera designed to output image data according to one of the YCbCr color spaces, such that each pixel is assigned a single value for the luma component (Y) and a set of values ​​for the chroma components (Cb, Cr).

[0049] The light source 308 includes one or more illuminants capable of emitting light in the visible or non-visible range. For example, the light source 308 may include an illuminant capable of emitting white light while a digital image is generated by the image sensor 306. Additionally or alternatively, the light source 308 may include an illuminant capable of emitting ultraviolet or infrared light. Examples of illuminants include light-emitting diodes (LEDs), organic LEDs (OLEDs), resonant cavity LEDs (RCLEDs), quantum dots (QDs), lasers such as vertical cavity surface-emitting lasers (VCSELs), superluminescent diodes (SLEDs), and various phosphors.

[0050] Those skilled in the art will recognize that when image sensor 306 is instructed (e.g., by processor 310) to generate a digital image in conjunction with light (whether visible or non-visible) emitted by light source 308, image sensor 306 must be designed to detect an appropriate range of electromagnetic radiation. In addition to CCD and CMOS, other examples of image sensors include monolithically integrated germanium (Ge) photodiodes, indium gallium arsenide (InGaAs) photodiodes, mercury cadmium telluride (HgCdTe) photodiodes, and other photodetectors (e.g., photodiodes) designed for the infrared and ultraviolet regions of the electromagnetic spectrum.

[0051] Thus, image sensor 306 and light source 308 may operate together to generate digital images of an anatomical region under particular lighting conditions. For example, image sensor 306 may generate a series of digital images while light source 308 emits light in the visible range. As another example, image sensor 306 may generate at least one digital image while light source 308 emits light in the visible range and at least one digital image while light source 308 emits light in the non-visible range.

[0052] As described above, image sensor 306 and light source 308 may be incorporated into a single electronic device. In some embodiments, the electronic device is associated with the unknown person. For example, image sensor 306 and light source 308 may be incorporated into a mobile phone associated with the unknown person. In other embodiments, the electronic device is not associated with the unknown person. For example, image sensor 306 and light source 308 may be incorporated into a point-of-sale system through which the unknown person is attempting to complete a transaction.

[0053] This electronic device may be referred to as a "vascular monitoring device" because it is responsible for monitoring changes in the vasculature within the anatomical region of interest. During the imaging portion of the authentication session, the vascular monitoring device may collect image data related to the anatomical region. As described further below, the authentication platform 302 may be able to identify pulse waves by examining the image data. The term "pulse waves" may refer to color changes along the surface of the anatomical region caused by the movement of blood through the underlying subcutaneous tissue. While the color changes may be difficult, if not impossible, to detect with the human eye, the authentication platform 302 may be able to identify these changes through analysis of the image data. Because pulse waves are related to the cardiac cycle, information about the vasculature in the anatomical region (and the cardiovascular system as a whole) may be gleaned from the pulse waves.

[0054] 3, authentication platform 302 may include a flow prediction algorithm 312, a flow measurement algorithm 314, a pattern matching algorithm 316, an authentication algorithm 318, and a biometric database 320. Biometric database 320 may store biometric data representing collected information related to vascular characteristics that can be used to identify known persons. The biometric data in biometric database 320 may vary depending on the approach to authentication adopted by system 300. The biometric data in biometric database 320 may be encrypted, hashed, or obfuscated to prevent unauthorized access.

[0055] For example, the biometric database 320 may include digital profiles of various individuals, each of which may include a corresponding individual's vein map that can be used for authentication. Each vein map may be constructed from or composed of two-dimensional or three-dimensional image data of a corresponding anatomical region. For example, assume that the authentication platform 302 is programmed to determine whether to authenticate an unknown individual as a given individual based on the deformation of the vasculature within the anatomical region. In such a scenario, the authentication platform 302 may prompt the unknown individual to perform a gesture and then establish the deformation of the vasculature within the anatomical region through analysis of the image data generated by the image sensor 306. The authentication platform 302 may then compare the deformation to a vein model associated with the given individual. At a high level, the vein model may programmatically indicate how the given individual's vasculature deformed while the gesture was performed. In other words, the vein model may represent a series of discrete positions that indicate how the shape of a single blood vessel or a collection of blood vessels changes over time as a gesture is performed, thereby causing deformation of the surrounding subcutaneous tissue.

[0056] This vein model can be created in several different ways. In some embodiments, the individual is prompted to perform a gesture while being imaged during the enrollment phase, and the vein model is created based on an analysis of the resulting digital image. In other embodiments, the individual's anatomical region is imaged so that a vein map can be created by the authentication platform 302. In such embodiments, the authentication platform 302 can simulate deformation of the vasculature while performing the gesture based on the vein map.

[0057] A digital profile may include a single vein model associated with a single gesture, multiple vein models associated with a single gesture, or multiple models associated with different gestures. Similarly, a digital profile may include a single vein model associated with a single anatomical region, multiple vein models associated with a single anatomical region, or multiple models associated with different anatomical regions. During the enrollment phase, an individual may be permitted to specify which anatomical region(s) and gesture(s) can be used for authentication. While the authentication platform 302 may require that at least one vein model be created for each anatomical region and gesture pairing, an individual may be permitted to create multiple vein models (e.g., for improved robustness).

[0058] Additionally or alternatively, the digital profile may include reference values ​​for different vascular characteristics (e.g., the velocity of blood determined to be flowing through blood vessels in a given anatomical region) that can be used for authentication. Thus, the biometric database 320 may include data indicative of temporal variations in vascular characteristics for a single vessel or a collection of vessels when a gesture is performed. Authentication can be based on similarities between values ​​for vascular characteristics such as pressure and flow rate instead of, or in addition to, similarities between spatial deformations of the vasculature.

[0059] As described above, the biometric database 320 may include one or more biometric signatures. The nature of each biometric signature may depend on how authentication is performed. For example, each biometric signature may represent a vein model created for an individual during an enrollment phase. Alternatively, each biometric signature may represent one or more values ​​indicating temporal changes in vascular characteristics when deformation of subcutaneous tissue occurs in an anatomical region. As an example, a biometric signature may comprise a vector of length N, each element of which is a value specifying the rate at which blood flows through the vasculature of the anatomical region when a gesture is performed. N may represent the number of samples obtained while the gesture is performed. In other words, N may represent the number of digital images generated for the anatomical region when a gesture is performed, since the flow rate can be estimated independently for each digital image.

[0060] The biometric signatures in the biometric database 320 may be associated with a single individual, in which case the authentication platform 302 may be limited to authenticating an unknown person as that individual. Alternatively, these biometric signatures may be associated with multiple individuals, in which case the authentication platform 302 may be able to authenticate an unknown person as any of those individuals. Additionally, as described above, a single individual may have multiple biometric signatures in the biometric database 320. These biometric signatures may correspond to different types (e.g., vascular models versus vascular property values), different anatomical regions, or different gestures. For example, an individual may choose to create multiple biometric signatures for different anatomical regions during the enrollment phase, and there may be a different biometric signature for each anatomical region. As another example, an individual may choose to create multiple biometric signatures for different gestures during the enrollment phase, and there may be a different biometric signature for each gesture.

[0061] When executed by the processor 310, the algorithms implemented in the authentication platform 302 allow an individual to generate a biometric signature during an enrollment phase. The algorithms implemented in the authentication platform 302 then allow verification to occur during a use phase. The enrollment and use phases are further described below.

[0062] The flow prediction algorithm 312 may be responsible for determining the relative timing of pulse waves in an anatomical region through analysis of one or more digital images of the anatomical region. For example, the flow prediction algorithm 312 may determine the timing or phase of a pulse wave at a certain spatial coordinate (e.g., specifying an anatomical region) based on the digital image, with or without physical deformation. When determined without physical deformation, this measurement may be referred to as a "measured venous flow pattern" or "measured flow pattern," while when determined with physical deformation, this measurement may be referred to as a "measured deformed venous flow pattern" or "measured deformed flow pattern." As an example, the relative arrival timing of pulse waves may be estimated based on recognition of pulse wave features such as dicrotic notches. Based on this information, the flow prediction algorithm 312 can estimate the velocity at which blood is flowing through the vasculature in the anatomical region. Alternatively, the flow prediction algorithm 312 may estimate another vascular characteristic, such as the phase of the pressure pulse, the direction of blood flow, the volume of blood flow, or the pressure of the vasculature in the anatomical region.

[0063] The flow measurement algorithm 314 may be responsible for predicting the propagation pattern of a pulse wave that will occur in an anatomical region when deformed. This propagation pattern may be referred to as a “predicted deformed venous flow pattern” or a “predicted deformed flow pattern.” To accomplish this, the flow measurement algorithm 314 may create a PPG by modeling, estimating, or predicting how a pulse wave will propagate through the anatomical region in a deformed state. The flow measurement algorithm 314 may take as inputs a measured flow pattern, a vein map, and a deformed vein map. As described above, the deformed vein map may be determined based on at least one digital image of the anatomical region in a deformed state, or alternatively, the deformed vein map may be determined by modifying the vein map to simulate a deformation.

[0064] In some embodiments, the flow measurement algorithm 314 is a machine learning algorithm. For example, the flow measurement algorithm 314 may be based on a neural network with parameters that are predetermined based on best practice examples or tuned through experimentation.

[0065] The predicted deformed flow pattern may be expressed using two-dimensional or three-dimensional coordinates relative to the surface of the anatomical region. Furthermore, the predicted deformed flow pattern may be associated with (i) timing information and (ii) phase information. The timing information may relate to the relative time at which a pressure pulse may arrive at a coordinate after arriving at the anatomical region. The timing information may correspond to an identifiable feature of the pressure pulse, such as a dicrotic notch or another portion of the pulse wave representing the pressure pulse. The phase information may relate to the relative phase of the pressure pulse that may be present at all coordinates at a single point in time. Each coordinate in the anatomical region may have a different pulse waveform due to different influences on the pressure pulse across the anatomical region.

[0066] The pattern matching algorithm 316 may be responsible for calculating the strength of the match between the predicted deformed flow pattern and the measured deformed flow pattern. In other words, the pattern matching algorithm 316 may be responsible for establishing the degree of similarity between the predicted deformed flow pattern and the measured deformed flow pattern. The degree of similarity may be expressed using a metric called a "match score." The match score may be expressed using any suitable numerical scale. For example, the match score may indicate the degree of similarity using any integer value between 0 and 100 or any decimal value between 0 and 1.

[0067] The authentication algorithm 318 may be responsible for determining whether to authenticate the unknown person as the predetermined individual based on the match score. For example, the authentication algorithm 318 may be programmed to authenticate the unknown person as the predetermined individual if the match score exceeds a predetermined threshold. If the match score does not exceed a predetermined threshold, the authentication algorithm 318 may not authenticate the unknown person as the predetermined individual. Typically, the authentication algorithm 318 is designed to output a binary signal (e.g., pass or fail) indicating whether authentication is appropriate. However, the authentication algorithm 318 may also be designed to output a non-binary signal. As an example, the output generated by the authentication algorithm 318 may indicate (i) that the unknown person should be authenticated as the predetermined individual, (ii) that the unknown person should not be authenticated as the predetermined individual, or (iii) that further authentication attempts are required. If the authentication algorithm 318 cannot reliably establish whether authentication is appropriate, the authentication platform 302 may take further action (e.g., by prompting the unknown person to perform a different gesture or present a different anatomical region for imaging).

[0068] 4 illustrates an example of an electronic device 400 capable of implementing an authentication platform 414 designed to authenticate the identity of an unknown person based on image data generated by an image sensor 408. As described above, the image data may represent one or more digital images of an anatomical region of the body. In some embodiments, the digital image(s) are generated based on ambient light reflected by the anatomical region toward the image sensor 408. In other embodiments, the light source 410 emits light toward the anatomical region to illuminate the anatomical region while the digital image(s) are generated by the image sensor 408. It should be noted that the light source 410 may also be configured to emit a discrete series of "pulses" or "flashes" of light over a time interval.

[0069] In some embodiments, authentication platform 414 is embodied as a computer program executed by electronic device 400. For example, authentication platform 414 may reside on a mobile phone that can obtain image data on which a determination is made as to whether authentication is appropriate. As another example, authentication platform 414 may reside on a POS system that can obtain image data on which a determination is made. In other embodiments, authentication platform 414 is embodied as a computer program executed by another electronic device to which electronic device 400 is communicatively connected. In such an embodiment, electronic device 414 may transmit image data to the other electronic device for processing. For example, authentication of an unknown person may be sought by a POS system used to initiate a transaction, but the image data may be generated by a mobile phone located in proximity to the unknown person. The image data may be provided to the POS system or another electronic device (e.g., a computer server) for processing, or the image data may be processed by the mobile phone before being provided to the POS system or other electronic device. Those skilled in the art will also recognize that aspects of authentication platform 414 may be distributed among multiple electronic devices.

[0070] The electronic device 414 may include a processor 402, a memory 404, a UI output device 406, an image sensor 408, a light source 410, and a communication module 412. The communication module 412 may be, for example, a wireless communication circuit designed to establish a communication channel with another electronic device. Examples of wireless communication circuitry include integrated circuits (also referred to as "chips") configured for Bluetooth, Wi-Fi, NFC, etc. The processor 402 may have general-purpose characteristics similar to a general-purpose processor, or the processor 402 may be an application-specific integrated circuit (ASIC) that provides control functions for the electronic device 400. As shown in FIG. 4, the processor 402 may be directly or indirectly coupled to all components of the electronic device 400 for communication purposes.

[0071] The memory 404 may be comprised of any suitable type of storage medium, such as static random access memory (SRAM), dynamic random access memory (DRAM), electrically erasable programmable read-only memory (EEPROM), flash memory, or registers. In addition to storing instructions executable by the processor 402, the memory 404 may also store image data generated by the image sensor 408 and data generated by the processor 402 (e.g., when executing modules of the authentication platform 414). It should be noted that the memory 404 is merely an abstract representation of a storage environment. The memory 404 may be comprised of actual memory chips or modules.

[0072] As described above, the light source 410 may be configured to emit light (more specifically, electromagnetic radiation) in the visible or non-visible range toward an anatomical region on the body of the unknown person to be authenticated. Typically, the light source 410 emits light only when instructed to do so. For example, if the authentication platform 414 determines that authentication is necessary, the authentication platform 414 may generate an output that prompts the processor 402 to (i) instruct the light source 410 to emit light and (ii) instruct the image sensor 408 to generate image data.

[0073] The communications module 412 can manage communications between components of the electronic device 400. The communications module 412 can also manage communications with other electronic devices. Examples of electronic devices include mobile phones, tablet computers, personal computers, wearable devices, point-of-sale systems, and network-accessible server systems comprised of one or more computer servers. For example, in embodiments in which the electronic device 400 is a mobile phone, the communications module 412 can facilitate communications with a network-accessible server system responsible for examining image data generated by the image sensor 408.

[0074] For convenience, authentication platform 414 may be referred to as a computer program residing in memory 404. However, authentication platform 414 may be comprised of software, firmware, or hardware components implemented in or accessible to electronic device 400. According to embodiments described herein, authentication platform 414 may include various algorithms, such as those described above with reference to FIG. 3. Typically, these algorithms are executed by separate modules of authentication platform 414 that are separately addressable (and thus can execute independently without interfering with other modules). These modules may be integral parts of authentication platform 414. Alternatively, these modules may be logically separate from authentication platform 414 but can operate “in parallel” with it. Together, these algorithms may enable authentication platform 414 to authenticate the identity of an unknown person based on an analysis of vascular dynamics determined from image data generated by image sensor 408.

[0075] For example, assume that an unknown person wishes to authenticate themselves as a predetermined individual. In such a scenario, the unknown person may be prompted to perform a gesture that causes deformation of the vasculature within an anatomical region. While the unknown person performs the gesture, the image sensor 408 may generate image data representing a digital image of the anatomical region. Based on analysis of the image data, the authentication platform 414 may generate a "biometric signature" of the unknown performer. For example, the authentication platform 414 may generate a vein model that programmatically indicates how the vasculature deformed while the gesture was performed, or the authentication platform 414 may estimate metrics of vascular properties based on analysis of the image data.

[0076] The authentication platform 414 can then compare the biometric signature with enrolled biometric signatures associated with the predetermined individual to determine whether the unknown person should be authenticated as the predetermined individual. Typically, the enrolled biometric signature is stored in a biometric database 416. In FIG. 4, the biometric database 416 is located in the memory 404 of the electronic device 400. However, the biometric database 416 may alternatively or additionally be located in a remote memory accessible to the electronic device 400 via a network. If the biometric signature is sufficiently similar to the enrolled biometric signature, the authentication platform 414 may authenticate the unknown person as the predetermined individual.

[0077] Other elements may also be included as part of the authentication platform 414. For example, a UI module may be responsible for generating content to be output by the UI output device 406 for presentation to the unknown person. The form of the content may depend on the nature of the UI output device 406. For example, if the UI output device 406 is a speaker, the content may include audio instructions for positioning the electronic device 400 so that the anatomical region is observable by the image sensor 408. As another example, if the UI output device 406 is a display, the content may include visual instructions for positioning the electronic device 400 so that the anatomical region is observable by the image sensor 408. The UI output device 406 may also be responsible for outputting (e.g., issuing or displaying) the authentication decision made by the authentication platform 414. [Vascular deformation caused by performing gestures]

[0078] Within a given anatomical region, blood vessels beneath the skin define a vasculature. As an example, the face includes several anatomical regions (e.g., the forehead, cheeks, and chin) where the vasculature can be visually monitored. Figure 5 illustrates how the underlying vasculature in an anatomical region (here, the face) can change with the execution of a gesture that causes physical deformation of the surrounding tissue. Figure 5 also illustrates how blood flowing through the vasculature can be monitored in terms of pulse waves. That is, the movement of blood within the vasculature can be visually monitored as it flows toward the arterial ends of the capillaries within the anatomical region. In Figure 5, arbitrary units numbered 1-4 are used to illustrate an example of the arrival order of pulse waves.

[0079] As described above, measured flow patterns can change due to (i) physical movement of blood vessels, which changes their position, and (ii) deformation of surrounding tissue, which changes the hemodynamic flow characteristics of the target vasculature. For example, tissue compression can increase capillary pressure, thereby changing the relative pulse phase and pulse wave velocity. Typically, whenever a gesture is repeatedly performed, the vasculature will deform in a predictable manner. Knowing (i) the gesture, (ii) the undeformed vascular pattern (e.g., the upper left image of FIG. 5), and (iii) the measured flow pattern (e.g., the lower left image of FIG. 5), the measured deformed flow pattern (e.g., the lower right image of FIG. 5) and / or the deformed vascular pattern (e.g., the upper right image of FIG. 5) can be determined. This process is further described below with reference to step 904 of FIG. 9. [Establishing flow patterns through pulse wave analysis]

[0080] A key aspect of some of the approaches described herein is establishing the flow pattern of blood through the vasculature within a given anatomical region through analysis of image data. Figures 6A-C illustrate several different approaches for determining, calculating, or obtaining a pulse wave. Figure 6A illustrates an approach in which object recognition is used to define a region of interest (ROI) from which a digital image is generated. As shown in Figure 6B, red and green pixel values ​​can be extracted for the ROI over a period of at least one cycle. This can be done for multiple patches within the ROI, where pixel values ​​are averaged over each patch. The patches can have a fixed size and be distributed within the ROI according to a segmentation function. Alternatively, the patches can have a size that can be adjusted (e.g., based on the size of the ROI or the amount of available computational resources). The average pixel value from each patch can then be used to estimate a pulse wave value, which can be used to establish the phase of the pulse wave. As shown in Figure 6C, these pulse wave values ​​can indicate relative phase periods in arbitrary units.

[0081] Note that because red and green pixel values ​​relate to changes occurring at different depths (e.g., different vascular structures), this component may be affected differently by physical deformation. This difference may be a useful component that the algorithms described herein can be trained to detect and then use to create predictions. Therefore, it may be beneficial to ensure that red and green pixel values ​​are not only available to the authentication platform, but can also be used independently to calculate predictions and phase estimates. [Method for authentication]

[0082] 7 includes a flow diagram of a procedure 700 for authenticating a user of an authentication platform based on analysis of visual evidence of vascular dynamics in an anatomical region. As described further below, the authentication procedure has three phases: a training phase, an enrollment phase, and a usage phase. These phases may be designed to enable minimally disruptive authentication without requiring the user to interact with the electronic device in an unusual way. Instead, the user may simply perform gestures while the electronic device generates digital images of the anatomical region that are deformed by the performance of those gestures.

[0083] For illustrative purposes, the authentication procedure may be described in the context of monitoring vasculature while performing a gesture to deform the surrounding subcutaneous tissue. However, the surrounding subcutaneous tissue may be deformed in other ways. For example, if the vasculature to be monitored is located on a finger, the user may be prompted to position the finger adjacent to the electronic device such that haptic feedback generated by a haptic actuator (or simply "actuator") located within the electronic device can deform the surrounding subcutaneous tissue.

[0084] To begin the training phase, the flow prediction algorithm can undergo supervised, semi-supervised, or unsupervised learning, where training data is retrieved, created, or acquired and then provided to the flow prediction algorithm for training purposes. The training data may include measured flow patterns and / or measured deformed flow patterns associated with corresponding vein maps and / or deformed vein maps. Typically, the training data is associated with a single anatomical region because understanding the deformations in one anatomical region (e.g., the face) may not be simply applicable to another anatomical region (e.g., the hand). However, the training data may also be associated with multiple individuals. Thus, for various individuals, the training data may include measured flow patterns, measured deformed flow patterns, vein maps, deformed vein maps, or any combination thereof.

[0085] In some embodiments, gesture-based learning is achieved through transfer learning based on a model trained during normal times (e.g., when no gestures are present, also referred to as the "normal state"). For example, each layer of a neural network can be divided into (i) a first layer associated with corresponding individual characteristics that are independent of gestures, and (ii) a second layer associated with corresponding individual characteristics that change with gestures. By fixing the first layer and training only the second layer for each target gesture, it is expected that the image data for training (and the learning process) can be reduced.

[0086] As mentioned above, part of the training data (e.g., the vein map and deformation vein map) may be image data. In some embodiments, the image data is generated from different capture angles or positions, or with different image sensors (e.g., corresponding to different electronic devices), to provide greater robustness to these changes during use.

[0087] Further, the training data may be divided into a training set and a test set, e.g., 80 percent of the training data is assigned to the training set and 20 percent of the training data is assigned to the test set. Those skilled in the art will recognize that these values ​​are provided for illustrative purposes. More or less than 80 percent of the training data may be assigned to the training set. Similarly, more or less than 20 percent of the training data may be assigned to the test set. The percentage of the training data assigned to the training set is typically larger (e.g., 2x, 3x, 5x, etc.) than the percentage of the training data assigned to the test set. The training set may be used to train a flow prediction algorithm, as described below, while the test set may be used to verify that the flow prediction algorithm has properly learned how to predict flows.

[0088] At a high level, a flow prediction algorithm comprises a set of algorithms designed to generate outputs (also called "predictions") related to blood flow through the vasculature of an anatomical region given specific inputs. These inputs may include vein maps, deformed vein maps, or image data of the anatomical region. In some embodiments, this set of algorithms represents one or more neural networks. The neural networks learn by processing examples, each associated with known inputs and outputs, to form probability-weighted associations between the inputs and outputs. These probability-weighted associations may be referred to as "weights." During a training phase, randomly selected weights may be initially used by the neural network(s) of the flow prediction algorithm. These weights can be adjusted as the flow prediction algorithm learns from measured flow patterns, vein maps, and deformed vein maps. Thus, the flow prediction algorithm may adjust these weights as it learns how to output predicted deformed flow patterns.

[0089] Each predicted deformed flow pattern output by the flow prediction algorithm may be scored based on its deviation from the corresponding measured deformed flow pattern, which serves as ground truth. Typically, this is done for each example included in the training set. More specifically, the pattern matching algorithm may calculate a score using a threshold for timing or phase at each time coordinate (e.g., ±3, 5, or 10 milliseconds). If the timing difference between the measured and predicted deformed flow patterns exceeds the threshold, that time coordinate may be marked as a failure. The score may be calculated by the pattern matching algorithm based on the percentage of time coordinates classified as failures. Furthermore, the pattern matching algorithm may compare the score with a predetermined threshold and classify each example in the training set as either "pass" or "fail" depending on whether the corresponding score exceeds the predetermined threshold. The pattern matching algorithm may calculate the overall success rate of the flow prediction algorithm based on the percentage of passing examples in the training set.

[0090] It should be noted that the weights of the neural network(s) of the flow prediction algorithm may be adjusted for any scheme in which adjustments are made to optimize success. One example of a known scheme is the Monte Carlo approach. Thus, this portion of the training phase may be repeated a predetermined number of cycles, or this portion of the training phase may be repeated until an overall success rate reaches an acceptable value (e.g., 90, 95, or 98 percent).

[0091] A test set may be used to verify that the flow prediction algorithm is operating properly. Thus, the flow prediction algorithm may be applied to the vein maps and deformed vein maps included in the test set to generate predicted flow patterns or predicted deformed flow patterns. As described above, the pattern matching algorithm may calculate a score indicating the performance of the flow prediction algorithm based on a comparison of the predicted flow patterns or predicted deformed flow patterns with the measured flow patterns or measured deformed flow patterns, respectively.

[0092] During the enrollment phase (also referred to as the "setup phase"), the vein map and deformation vein map may be generated for the user during normal use of the electronic device. For example, upon receiving an input indicating a request to initiate the enrollment phase, the electronic device may generate a digital image of the anatomical region while the anatomical region is deformed (e.g., by the user performing a gesture). The deformation may be prompted via the UI by requesting (e.g., via text) that the user smile, frown, or purse their lips while a digital image of the face is generated by the electronic device. For example, the UI may display a graphical representation of the deformation with a general model of the anatomical region or human body to visually instruct the user. As another example, the UI may display a graphical representation of a gesture that may cause the deformation of the anatomical region. For example, the graphical representation may serve as a visual instruction to interact with the electronic device responsible for generating the digital image in a particular way (e.g., swiping a finger on the screen, holding the housing in a particular way, etc.).

[0093] From these digital images, vein maps and deformed vein maps for anatomical regions can be generated. As shown in Figure 7, the vein maps and deformed vein maps are typically stored in a biometric database and subsequently retrieved when an unknown person attempts to authenticate themselves as a user.

[0094] FIG. 8 includes a flow diagram of a process 800 performed by the authentication platform during the use phase (also referred to as the "implementation phase"). Initially, the authentication platform receives input from a source indicating a request to authenticate an unknown person as a predetermined individual (step 801). In some embodiments, the source is a computer program running on the same electronic device as the authentication platform. For example, if the authentication platform resides on a mobile phone, the authentication request may originate from a mobile application in which the unknown person is attempting to perform an activity requiring authentication. In other embodiments, the source originates from another electronic device. For example, assume that an unknown person attempts to complete a transaction using a POS system associated with a merchant. In such a scenario, the POS system may request that the authentication be performed. While the POS system may be responsible for generating the image data required for authentication, the authentication platform may reside on a computer server communicatively connected to the POS system via a network.

[0095] Next, the authentication platform may receive (i) a vein map and (ii) a deformed vein map associated with the given individual (step 802). As described above, the deformed vein map may be associated with a gesture performed by the given individual during the enrollment phase. Furthermore, the authentication platform may cause a notification prompting the unknown person to perform a gesture (step 803). This notification is intended to prompt the unknown person to perform the same gesture as was performed by the given individual during the enrollment phase.

[0096] As described above, performing a gesture may result in deformation of the target anatomical region. As the unknown person performs the gesture, the electronic device may monitor the deformation of the anatomical region. For example, an image sensor in the electronic device may generate image data through observation of the anatomical region, and the authentication platform may acquire this image data for analysis (step 804). In some embodiments, the image data comprises digital images generated before, during, or after deformation of the anatomical region. For example, the electronic device may generate a first series of digital images over a first time interval before the deformation occurs and a second series of digital images over a second time interval while the deformation is "held." Thus, the electronic device may generate digital images of the anatomical region while it is in a natural state (also referred to as a "relaxed" state) and a deformed state. Typically, the first and second time intervals are long enough to allow at least one entire cardiac pulse cycle to be observed. The duration of a cardiac cycle varies depending on various physiological factors, but it typically falls within the range of 0.5-2.0 seconds. Thus, the first and second time intervals may be at least 1, 2, or 3 seconds. Longer durations may optionally be used to capture more than one cardiac cycle.

[0097] The authentication platform may then analyze the image data to determine (i) a measured flow pattern and (ii) a measured deformed flow pattern (step 805). To accomplish this, the authentication platform may apply a flow measurement algorithm to the image data. When applied to the image data, the flow measurement algorithm may first perform a registration operation (also called a "mapping operation") to determine pixel locations in the image data that correspond to specific anatomical coordinates in the vein map and the deformed vein map. This mapping operation ensures that values ​​in these data sets relate to the same location in the anatomical region. Next, the flow measurement algorithm may average the red or green frequency components of the image data over various pixel regions (e.g., 3x3, 6x6, or 9x9 pixel regions). The frequency band selected may approximately correspond to the frequency of pressure pulses that carry blood to the anatomical region. However, other frequencies may be used instead, if appropriate. The average red or green frequency components may be calculated for image data corresponding to different time points, such that a time series identifying the intensity of the frequency bands of interest is created. For example, a flow measurement algorithm may average the red or green frequency components of different digital images (e.g., that represent frames of a video generated by an electronic device). Furthermore, the flow measurement algorithm may perform pattern recognition to determine the relative timing of a single identifiable phase of a pressure pulse, such as a dicrotic notch, based on an analysis of a time-varying sequence of average values ​​of the red or green frequency components. After this phase of the pressure pulse is identified, the flow measurement algorithm may assign a timing value to some or all of the pixel regions associated with the earliest detected occurrence of the identified phase of the pressure pulse. At a high level, these timing values ​​may represent a flow pattern indicative of how blood flows through the vasculature of an anatomical region, as determined from an analysis of the image data. When the image data is associated with an anatomical region in its natural state, this flow pattern may be referred to as a "measured flow pattern." When the image data is associated with an anatomical region in its deformed state, this flow pattern may be referred to as a "measured deformed flow pattern."

[0098] The authentication platform may then apply a flow prediction algorithm to (i) the measured flow pattern generated for the unknown person, (ii) the vein map for the given individual, and (iii) the deformed vein map for the given individual to generate a predicted deformed flow pattern (step 806). At a high level, the predicted deformed flow pattern may be a data structure including timing values ​​that represent a prediction of how blood may flow through the vascular system of the given individual when a gesture is performed.

[0099] The authentication platform can then apply a pattern matching algorithm to (i) the measured variant flow pattern and (ii) the predicted variant flow pattern to generate a metric indicating similarity (step 807). As mentioned above, this metric may be referred to as a "match score." At a high level, this metric may indicate, for each value, the degree to which the measured variant flow pattern is equivalent to the predicted variant flow pattern.

[0100] The authentication platform can then determine whether to authenticate the unknown person as the predetermined individual based on the metric (step 808). For example, the authentication platform may apply an authentication algorithm that compares the metric to a predetermined threshold. If the metric exceeds the predetermined threshold, the authentication algorithm may generate an output indicating that the unknown person should be authenticated as the predetermined individual. However, if the metric does not exceed the predetermined threshold, the authentication algorithm may generate an output indicating that the unknown person should not be authenticated as the predetermined individual.

[0101] FIG. 9 includes a flow diagram of another process 900 for determining whether to authenticate an unknown person as a predetermined individual through a vascular survey. Meanwhile, FIG. 10 includes a visual illustration of a process by which an authentication platform determines whether to authenticate an unknown person as a predetermined individual. Initially, the authentication platform may receive an input indicating a request to authenticate an unknown person as a predetermined individual (step 901). Step 901 of FIG. 9 may be substantially similar to step 801 of FIG. 8. For illustrative purposes, process 900 is described in the context of examining a digital image generated by an electronic device carried by an unknown person. However, one skilled in the art will recognize that process 900 may be equally applicable to scenarios in which an unknown person is in proximity to an electronic device but does not possess the electronic device (e.g., when the electronic device is a point-of-sale system).

[0102] The authentication platform can then retrieve a digital profile associated with the predetermined individual claimed by the unknown person (step 902). For example, the authentication platform may access a biometric database in which multiple digital profiles associated with different individuals are stored, and the authentication platform can then select a digital profile from among the multiple digital profiles based on the input. Typically, the input identifies the predetermined individual claimed by the unknown person (e.g., using a name or an identifier such as an email address or phone number), so that the authentication platform can easily identify the appropriate digital profile from among those stored in the biometric database.

[0103] A digital profile may include one or more vascular patterns (also called "vein maps") associated with a given individual. In addition to being associated with a given individual, each vascular pattern may be associated with a given anatomical region. For example, a digital profile may include separate vascular patterns for the face, palm, fingers, etc. Additionally, a digital profile may include vascular profiles for the same anatomical region in different states. For example, a digital profile may include (i) a first vascular pattern that provides spatial information about the blood vessels within the anatomical region while the anatomical region is in a natural state, and (ii) a second vascular pattern that provides spatial information about the blood vessels within the anatomical region while the anatomical region is in a deformed state (e.g., due to the performance of a gesture).

[0104] The authentication platform then instructs the unknown person to perform a gesture that causes deformation of the anatomical region while the camera of the electronic device is pointed at the anatomical region (step 903). As the unknown person performs the gesture, the camera may generate a series of digital images. These digital images may be generated discretely in quick succession at a predetermined cadence (e.g., every 0.1, 0.2, or 0.5 seconds). Alternatively, the camera may generate a video of the anatomical region, in which case the digital images may represent frames of the video. In this scenario, the digital images may be generated at a predetermined rate (e.g., 20, 30, or 60 frames per second).

[0105] The authentication platform can then estimate the unknown person's flow pattern based on the digital images generated by the camera. More specifically, the authentication platform can estimate (i) a first blood flow pattern while the unknown person's anatomical region is in a natural state and (ii) a second blood flow pattern while the person's anatomical region is in a deformed state based on an analysis of the digital images (step 904). As described above, these flow patterns may be estimated based on different digital images generated by the electronic device's camera. The first flow pattern may be generated based on an analysis of a digital image of the anatomical region in a natural state (e.g., generated before or after a gesture is performed), and the second flow pattern may be generated based on an analysis of a digital image of the anatomical region in a deformed state (e.g., generated while a gesture is being performed or held). As described above with respect to FIG. 8, the first and second flow patterns may be estimated based on a programmatic analysis of pixels of the corresponding digital images to identify color changes (e.g., in the red or green component) indicative of blood flow through blood vessels in the anatomical region.

[0106] Further, the authentication platform can predict a third flow pattern of blood through the anatomical region of the given individual that would be expected if the given individual were to perform the gesture based on the digital profile and the first flow pattern (step 905). As described above, the digital profile can include (i) a first vascular pattern that provides spatial information about blood vessels within the anatomical region of the given individual while the anatomical region is in a natural state, and (ii) a second vascular pattern that provides spatial information about blood vessels within the anatomical region of the given individual while the anatomical region is in a deformed state. By applying an algorithm to the first vascular pattern, the second vascular pattern, and the first flow pattern, the authentication platform can generate a third flow pattern as an output. At a high level, the algorithm can simulate blood flow through blood vessels during deformation of the anatomical region that would be caused by the gesture.

[0107] The authentication platform can then determine whether to authenticate the unknown person as the predetermined individual based on a comparison of the second flow pattern with the third flow pattern (step 906). For example, assume that the first flow pattern, the second flow pattern, and the third flow pattern are represented as matrices. The first flow pattern may be represented as a first vector or matrix, each element of which includes a value indicating an estimated blood flow through a corresponding portion of the unknown person's anatomical region while in a natural state. The second flow pattern may be represented as a second vector or matrix, each element of which includes a value indicating an estimated blood flow through a corresponding portion of the unknown person's anatomical region while in a deformed state. Meanwhile, the third flow pattern may be represented as a third vector or matrix, each element of which includes a value indicating an estimated blood flow through a corresponding portion of the predetermined individual's anatomical region while in a deformed state. In such a scenario, the authentication platform may apply an algorithm to the second and third vectors or matrices to generate a score indicating the similarity between the second and third flow patterns. The authentication platform can then establish the likelihood that the unknown person is a given individual based on the score. Note that as used herein, the term "matrix" can be used to refer to a set of row vectors or column vectors.

[0108] Other steps may also be included. As one example, the authentication platform may generate a signal (e.g., in the form of a message or notification) indicating whether the unknown person has been authenticated as a predetermined individual. The authentication platform may transmit this signal to the source from which the request to authenticate was received. For example, if the request to authenticate the unknown person is received from a computer program running on a mobile phone, the authentication platform may provide a signal to the mobile phone so that the computer program can establish whether to allow the unknown person to perform any tasks requiring authentication. Similarly, if the request to authenticate the unknown person is received from a point-of-sale system during a transaction, the authentication platform may provide a signal to the point-of-sale system so that the transaction can be completed.

[0109] As described above with reference to FIG. 8 , performing authentication may require the authentication platform to apply a flow prediction algorithm to (i) a measured flow pattern associated with an unknown person, (ii) a vein map associated with the given individual, and (iii) a deformed vein map associated with the given individual to generate a predicted deformed flow pattern. This predicted deformed flow pattern represents a prediction by the authentication platform of how blood will flow through the vasculature when the given individual performs a particular gesture. In some embodiments, the flow prediction algorithm is part of a collection of algorithms that collectively define a flow prediction model. Generally, a flow prediction model is a machine learning (ML) or artificial intelligence (AI) model that is “trained” with examples to make predictions, i.e., predict how blood will flow through the vasculature when deformed. Generally, a flow prediction model is a machine learning (ML) or artificial intelligence (AI) model that is “trained” with examples to make predictions, i.e., how blood will flow through the vasculature when deformed.

[0110] 11 includes a flow diagram of a process 1100 for creating a model trained to predict blood flow through the vasculature of an anatomical region when deformed. As described above, an anatomical region may be deformed by performing a gesture, or the anatomical region may be deformed by applying an external force (e.g., haptic feedback generated by a haptic actuator). The nature of the deformation may depend on the anatomical region. For example, the vasculature of a face may be susceptible to deformation by instructing a person to perform a gesture (e.g., smiling or frowning), and the vasculature of a finger may be susceptible to deformation by instructing a person to place a finger against an electronic device and then applying an external force (e.g., via haptic feedback).

[0111] First, the authentication platform can identify a model to be trained to predict blood flow through a venous network of an anatomical region as it deforms (step 1101). Note that the terms "venous network" and "vasculature" may be used interchangeably. Thus, the term "venous network" may refer to a portion of the vasculature located within an anatomical region. While an anatomical region can be any part of the body where vascular dynamics can be monitored by imaging, common anatomical regions include the fingers, the palmar and dorsal sides of the hand, and the face.

[0112] The authentication platform can then obtain (i) a first series of vascular patterns corresponding to the anatomical region in its natural state, (ii) a second series of vascular patterns corresponding to the anatomical region in its deformed state, (iii) for each vascular pattern in the first series, a series of flow patterns that convey how blood flows through that vascular pattern when the anatomical region is in its natural state, and (iv) for each vascular pattern in the second series, a series of deformed flow patterns that convey how blood flows through that vascular pattern when the anatomical region is in its deformed state (step 1102). Each vascular pattern in the first series can indicate the spatial relationship between subcutaneous blood vessels when the anatomical region is in its natural state, while each vascular pattern in the second series can indicate the spatial relationship between subcutaneous blood vessels when the anatomical region is in its deformed state.

[0113] Furthermore, each vascular pattern in the first series is associated with a corresponding vascular pattern in the second series, and corresponding vascular patterns in the first and second series may be associated with the same individual. Thus, a single individual may be associated with a vascular pattern in the first series, a vascular pattern in the second series, one of the flow patterns, and one of the modified flow patterns. Typically, each vascular pattern in the first series is associated with a different individual, but the same individual may be associated with multiple vascular patterns in the first series. For example, a single individual may be associated with vascular patterns corresponding to the same anatomical region but generated using image data generated by different electronic devices. Similarly, each vascular pattern in the second series is typically associated with a different individual. However, as noted above, each vascular pattern in the first series may be associated with the same individual as the corresponding vascular pattern in the second series.

[0114] The authentication platform may then provide (i) the first series of vascular patterns, (ii) the second series of vascular patterns, (iii) the series of flow patterns, and (iv) the series of deformed flow patterns as training data to the model (step 1103). Such an approach trains the model to predict blood flow through a venous network in an anatomical region of a person when applied to the vascular patterns associated with the person. In other words, the authentication platform may provide this information as training data to the model to generate a trained model capable of predicting blood flow. For example, if the authentication platform is tasked with predicting blood flow through an anatomical region of a given individual, the authentication platform may apply the trained model to a set of vascular patterns associated with the given individual. The set of vascular patterns may include one vascular pattern corresponding to the anatomical region in its natural state and another vascular pattern corresponding to the anatomical region in its deformed state. After training is complete, the authentication platform may store the trained model in a biometric database (step 1104). [Additional Considerations and Implementation] [A. Personalized Gestures]

[0115] As mentioned above, vein maps can play a key role in determining whether to authenticate an unknown person as a given individual. To tailor the authentication process, the authentication platform can design or select variations based on those vein maps.

[0116] For example, assume that an unknown person wishes to authenticate themselves as a given individual. As part of the authentication process, the authentication platform may obtain vein maps associated with the given individual (e.g., a first vein map for the anatomical region in its natural state and a second vein map for the anatomical region in a deformed state). In this situation, the authentication platform may design or select deformations that better surface or emphasize unique aspects of these vein maps. For example, the authentication platform may analyze some or all of the vein maps included in the biometric database to identify sufficiently unique features. These features may relate to spatial relationships between different blood vessels (e.g., unusual branching locations or unusual dimensions), or they may relate to vascular properties of the blood vessels (e.g., whether the velocity, volume, or pressure of blood flowing through the venous network changes more or less than average after deformation).

[0117] Additionally or alternatively, the authentication platform may utilize a system that delivers a request to the unknown person to perform a gesture (or prompt, induce, or cause a deformation) in a manner that surfaces a unique characteristic. For example, the authentication platform may ask the unknown person to apply pressure to a location along the palm of their hand to inhibit or occlude blood flow to a unique blood vessel location, thereby creating a unique effect on the venous resistance of other vessels that feed into that unique blood vessel. The effect of the deformation on the pressure pulse emanating through the anatomical region (and thus the image data visually capturing the pressure pulse) may be co-located with the deformation or may be located some distance from the location of the deformation.

[0118] As described above, the authentication platform may generate measured flow patterns and measured deformed flow patterns as part of the authentication process. In some embodiments, the authentication platform determines the differences between the measured flow patterns and the measured deformed flow patterns and then compares these differences against other examples to ensure that changes to the image data due to the deformations are also sufficiently unique. B. Matching Measured Flow Patterns

[0119] After the measured flow pattern and the measured deformed flow pattern are known for the unknown person, the authentication platform may authenticate the unknown person based solely on a match of the measured flow pattern since an association with the original vein map authentication element was previously established. This "lightweight" authentication process may be suitable only for some situations, such as those involving minimal sensitive information or behavior. However, this "lightweight" authentication process may be useful, for example, for quickly authenticating an unknown person when time or computational resources are limited. [C. Multiple Transformations]

[0120] As described above, a given individual may be prompted to perform multiple gestures during the enrollment phase. Such an approach provides significant security advantages because there are multiple options for authentication. If an unknown person seeks to be authenticated as the given individual, the authentication platform may require the unknown person to perform any combination of the gestures performed by the given individual during the enrollment phase. Thus, the authentication platform may require the unknown person to perform multiple different gestures during the authentication process, and the authentication platform may authenticate the unknown person as the given individual only if a predetermined percentage (e.g., greater than 50 percent, exactly 100 percent) of those gestures match the given individual.

[0121] The authentication platform may also require the unknown person to perform the same gesture multiple times. For example, the authentication platform may require the unknown person to perform a single gesture multiple times during the authentication process, and the authentication platform may authenticate the unknown person as the predetermined individual only if a predetermined percentage of those performances (e.g., greater than 50 percent, exactly 100 percent) match the predetermined individual.

[0122] In embodiments in which the authentication platform allows an individual to perform multiple gestures during the enrollment phase, the authentication platform may manage separate biometric databases for those different gestures. For example, the authentication platform may manage a first biometric database containing information for a first gesture (e.g., a vein map and a deformed vein map), a second biometric database containing information for a second gesture, etc. Alternatively, the authentication platform may store information associated with different gestures in different portions of a single biometric database.

[0123] Furthermore, entries in the biometric database may be associated not only with a name or identifier (e.g., an email address or phone number) that identifies the corresponding individual, but also with a label that identifies the corresponding gesture. Thus, different gestures (e.g., a smile and a frown) may be associated with different labels that can be added to entries in the biometric database.

[0124] In some embodiments, the appropriate label is identified based on an analysis of the image data used for authentication. For example, if the image data includes digital images of faces, the authentication platform may examine those digital images to determine which gestures were performed. Automatic analysis of the image data may be useful in several ways. First, the authentication platform may be able to infer which gestures were performed by an unknown person without explicitly instructing the unknown person to perform the gestures. Second, the authentication platform may be able to establish an appropriate vein map to retrieve from a biometric database. For example, if the authentication platform determines that an unknown person in a digital image has performed a predetermined gesture, the authentication platform may retrieve a vein map associated with the predetermined gesture from a biometric database. [Advantages of authenticating unknown individuals through ranged vascular studies] [A. Changes due to aging and environment]

[0125] It is understood that blood vessel shape generally does not change, but blood vessel properties (e.g., flow rate) can be affected by factors such as age, disease, etc. Environmental factors such as temperature and humidity can also affect blood vessel properties. For example, cold temperatures can cause blood vessels to compress or constrict. Blood flow can also be affected by physiological factors (e.g., tension and stress) and physiological activity (e.g., exercise).

[0126] The authentication platform can be designed to be robust to such changes in vascular properties. One important aspect of the authentication platform is that it focuses on the local spatial characteristics and directionality of blood flow as a result of deformation. Therefore, the influence of the aforementioned factors that tend to affect the entire body is typically negligible or manageable (e.g., through modeling). For example, while the directional pattern of blood flow through the venous network in an anatomical region due to gesture execution is observable in stressed and relaxed states, the absolute signal intensity (e.g., determined through analysis of image data) may differ.

[0127] Changes in overall blood flow can introduce noise into individual measurements, but this effect has been adequately addressed in recent studies that have shown that blood flow remains readily observable even after physical activity (e.g., exercise). [B. Robustness in difficult scenarios]

[0128] Ongoing research is improving the accuracy and robustness of establishing or monitoring vascular dynamics through the analysis of digital images. There is also growing interest in remote monitoring, especially approaches that utilize readily available electronic devices such as mobile phones and tablet computers.

[0129] In scenarios where changes in an unknown user's health affect local-scale spatial characteristics (e.g., due to traumatic injury, stroke, etc.), the authentication platform may employ modeling techniques to account for these changes. For example, if the authentication platform detects, based on analysis of image data associated with an unknown individual, that a deformation has occurred, the authentication platform can apply an ML-based model designed to adjust the vein map accordingly. Thus, the authentication platform may be able to intelligently manipulate the vein map to account for changes in an individual's health after those individuals complete the enrollment phase. As another example, if the authentication platform detects a high heart rate based on analysis of image data associated with an unknown individual, the authentication platform can apply an ML-based model to determine appropriate adjustments to vascular characteristics, such as flow rate or pressure. However, because spatial information and vascular characteristics tend to be fairly uniform over time, it is unlikely that these types of adjustments would be widely required. [Processing System]

[0130] 12 is a block diagram illustrating an example of a processing system 1200 in which at least some operations described herein may be performed. For example, components of processing system 1200 may be hosted on an electronic device that includes an image sensor. As another example, components of processing system 1200 may be hosted on an electronic device that includes an authentication platform responsible for examining image data generated by the image sensor.

[0131] Processing system 1200 may include a processor 1203, main memory 1206, non-volatile memory 1210, network adapter 1212 (e.g., network interface), video display 1218, input / output device 1220, control device 1222 (e.g., mechanical input such as a keyboard, pointing device, or buttons), drive unit 1224 including recording medium 1226, or signal generating device 1230, all communicatively coupled to bus 1216. Bus 1216 is illustrated as an abstraction representing one or more physical buses and / or point-to-point connections connected by appropriate bridges, adapters, or controllers. Thus, bus 1216 may be any of a variety of buses, including a system bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express bus, a HyperTransport bus, an Industry Standard Architecture (ISA) bus, a Small Computer System Interface (SCSI) bus, a Universal Serial Bus (USB), an I / O bus, a Serial Bus (SPI), a Serial Adapter (SAC ... 2 This may include an Inter Integrated Circuit (C) bus, or a bus conforming to the Institute of Electrical and Electronics Engineers (IEEE) standard 1394.

[0132] Processing system 1200 may share a computer processor architecture similar to that of a computer server, a router, a desktop computer, a tablet computer, a mobile phone, a video game console, a wearable electronic device (e.g., a watch or fitness tracker), a network-connected (“smart”) device (e.g., a television or home assistant device), an augmented or virtual reality system (e.g., a head-mounted display), or another electronic device capable of executing (sequentially or otherwise) a set of instructions that specify action(s) to be taken by processing system 1200.

[0133] Although main memory 1206, non-volatile memory 1210, and storage medium 1224 are depicted as a single medium, the terms "storage medium" and "machine-readable medium" should be interpreted to include a single medium or multiple media that store one or more sets of instructions 1226. Also, the terms "storage medium" and "machine-readable medium" should be interpreted to include any medium that can store, encode, or carry a set of instructions for execution by processing system 1200.

[0134] Generally, the routines executed to implement embodiments of the present disclosure may be implemented as part of an operating system or a specific application, component, program, object, module, or sequence of instructions (collectively referred to as a "computer program"). A computer program typically comprises one or more instructions (e.g., instructions 1204, 1208, 1228) stored at various times in various memories and storage devices within a computing device. When read and executed by processor 1202, the instructions cause processing system 1200 to perform operations to implement various aspects of the present disclosure.

[0135] While embodiments have been described in the context of a fully functional computing device, those skilled in the art will understand that various embodiments can be distributed as program products in various forms. The present disclosure applies regardless of the particular type of machine- or computer-readable medium used to actually achieve the distribution. Further examples of machine- and computer-readable media include volatile memory devices, non-volatile memory devices 1210, removable disks, hard disk drives, recordable-type media such as optical disks (e.g., compact disk read-only memories (CD-ROMs) and digital versatile disks (DVDs)), cloud-based storage, and transmission-type media such as digital and analog communication links.

[0136] Network adapter 1212 enables processing system 1200 to broker data within network 1214 with entities external to processing system 1200 through any communication protocol supported by processing system 1200 and the external entity. Network adapter 1212 may include a network adapter card, a wireless network interface card, a switch, a protocol converter, a gateway, a bridge, a hub, a receiver, a repeater, or a transceiver including integrated circuits (e.g., enabling communication over Bluetooth or Wi-Fi). [remarks]

[0137] The foregoing description of various embodiments of the claimed subject matter has been provided for purposes of illustration and description. It is not intended to be exhaustive or to limit the claimed subject matter to the precise form disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments have been chosen and described in order to best explain the principles of the invention and its practical application, so that others skilled in the relevant art will be able to understand the claimed subject matter, various embodiments, and various modifications that are suitable for the particular use contemplated.

[0138] Although the detailed description describes specific embodiments and the best mode contemplated, no matter how detailed the detailed description appears, the technology can be implemented in many ways. While embodiments are encompassed within this specification, they may vary considerably in their implementation details. Certain terms used when describing particular features or aspects of various embodiments should not be construed as meaning that the terms are redefined herein to be limited to any particular characteristic, feature, or aspect of the technology with which they are associated. In general, the terms used in the following claims should not be construed to limit the technology to the specific embodiments disclosed herein unless those terms are expressly defined herein. Therefore, the actual scope of the technology encompasses not only the disclosed embodiments, but also all equivalent ways of practicing or implementing the embodiments.

[0139] The terminology used herein has been selected primarily for ease of reading and description; it has not been selected to define or enclose subject matter. Accordingly, it is intended that the scope of the technology be limited not by this detailed description, but rather by any claims that issue in an application based hereon. Accordingly, the disclosure of various embodiments is intended to be illustrative, but not limiting, of the scope of the technology, which is defined in the following claims. [Industrial Applicability]

[0140] The present disclosure is applicable to biometric authentication in computer security.

Claims

1. receiving input indicating a request to authenticate a person possessing an electronic device including a camera; obtaining a digital profile associated with a given identity claimed by said person; causing the person to perform a gesture that causes a deformation of the anatomical region while the camera of the electronic device is pointed at the anatomical region; estimating (i) a first flow pattern of blood while the anatomical region of the person is in a natural state, and (ii) a second flow pattern of blood while the anatomical region of the person is in a deformed state based on an analysis of the digital images produced by the camera; predicting a third flow pattern of blood through the anatomical region of the individual that would be expected if the individual were to perform the gesture based on the digital profile and the first flow pattern; and determining whether to authenticate the person as the predetermined individual based on a comparison of the second flow pattern and the third flow pattern; A method for providing the above.

2. 2. The method of claim 1, wherein the digital profile includes (i) a first vascular pattern that provides spatial information about blood vessels in the anatomical region while the anatomical region is in a natural state, and (ii) a second vascular pattern that provides spatial information about blood vessels in the anatomical region while the anatomical region is in a deformed state.

3. The predicting step comprises:

3. The method of claim 2, comprising applying an algorithm to the first vascular pattern, the second vascular pattern, and the first flow pattern that generates the third flow pattern as an output by simulating blood flow through the vessels during deformation of the anatomical region that would be caused by the gesture.

4. The estimating step comprises: The method of claim 1 , comprising examining the digital image to identify color variations indicative of blood flow through blood vessels in the anatomical region.

5. the first flow pattern is represented as a first vector, each element of which includes a value indicative of an estimated blood flow through a corresponding portion of the anatomical region of the person while in the natural state; and the second flow pattern is represented as a second vector, each element of which includes a value indicative of an estimated blood flow through the corresponding portion of the anatomical region of the person while in the deformed state; and 2. The method of claim 1, wherein the third flow pattern is represented as a third vector, each element of which includes a value indicative of an estimated blood flow through the corresponding portion of the anatomical region of the given individual while in the deformed state.

6. The determining step comprises: applying an algorithm to the second and third matrices to generate a score indicative of the similarity between the second and third flow patterns; and establishing a likelihood that the person is the predetermined individual based on the score; The method of claim 5 , comprising:

7. The obtaining includes: accessing a biometric database in which multiple digital profiles associated with different individuals are stored; and selecting the digital profile from among the plurality of digital profiles based on the input; The method of claim 1 , comprising:

8. The method of claim 7 , wherein the input identifies the predetermined individual to whom the person claims to be.

9. When executed by a processor of an electronic device, the electronic device Identifying a model that is trained to predict blood flow through a venous network in an anatomical region as it deforms; (i) a first series of vascular patterns corresponding to said anatomical region in its natural state; (ii) a second series of vascular patterns corresponding to said anatomical region in a deformed state; and (iii) for each vascular pattern in the first series, a series of flow patterns that describe how blood would flow through that vascular pattern when the anatomical region is in the native state; (iv) for each vascular pattern in the second series, obtaining a series of deformed flow patterns that describe how blood flows through that vascular pattern when the anatomical region is in the deformed state; and providing (i) the first series of vascular patterns, (ii) the second series of vascular patterns, (iii) the series of flow patterns, and (iv) the series of modified flow patterns to the model as training data, so as to generate a trained model that, when applied to vascular patterns associated with a person, is capable of predicting blood flow through the venous network in an anatomical region of the person; A computer program that causes a computer to perform the operations comprising:

10. 10. The computer program of claim 9, wherein each vascular pattern in the first series is associated with a corresponding vascular pattern in the second series, and corresponding vascular patterns in the first and second series are associated with the same individual.

11. each vascular pattern in the first series represents a spatial relationship between subcutaneous blood vessels when the anatomical region is in the natural state, and each vascular pattern in the second series represents a spatial relationship between the subcutaneous blood vessels when the anatomical region is in the deformed state; 10. A computer program according to claim 9.

12. 10. The computer program of claim 9, wherein each vascular pattern in the first series is associated with a different individual.

13. 10. The computer program of claim 9, wherein the anatomical region is a finger, a palmar or dorsal side of a hand, or a face.

14. 10. The computer program of claim 9, wherein the model comprises a neural network having parameters that are adjusted based on the results of the providing.

15. presenting a notification instructing the person to be authenticated to perform a gesture that causes a deformation of the anatomical region; acquiring a digital image of the anatomical region generated by an electronic device as the person performs the gesture; estimating characteristics of blood flow through subcutaneous blood vessels of the anatomical region based on the digital images; and determining whether to authenticate the person as the predetermined individual based on a comparison of the estimated characteristics with a profile associated with the predetermined individual; Equipped with the characteristic is a deformation of a venous network formed by the subcutaneous blood vessels; A method characterized by:

16. The method of claim 15 , wherein the digital image is generated in conjunction with visible, infrared, or ultraviolet light emitted by the electronic device.

17. The method of claim 15 , wherein the property relates to directionality, velocity, volume, phase, or pressure of blood flowing through the subcutaneous vessel.

Images