Systems and methods for biometric identification
By updating and adjusting the biometric data template in the biometric identification system, using auxiliary matching judgment and additional certification standards, the identification accuracy problems caused by biometric changes are solved, and the system's robustness and safety are improved.
Patent Information
- Application Number
- CN201811249260.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2018-10-24
- Filing Date
- 2018-10-25
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2038-10-25
AI Technical Summary
Biometric identification systems are highly sensitive to biometric changes, resulting in reduced identification accuracy and reliability. Especially in fingerprint recognition, when environmental changes such as skin humidity changes, it is difficult for the system to accurately match registration and verification.
By updating templates in the biometric data repository, using auxiliary matching judgment and additional certification standards, and combining changes in biometric inputs, biometric data will be automatically updated and adjusted to improve matching accuracy and system robustness.
It improves the identification accuracy and reliability of the biometric identification system under different environmental conditions, reduces the error rejection rate and imposter risks, and enhances the security and user experience of the system.
Smart Images

Figure CN109858316B_ABST
Abstract
Description
[0001] Cross-references
[0002] This application claims the benefit of U.S. Provisional Patent Application No. 62 / 577,115, filed October 25, 2017, the contents of which are incorporated by reference. Technical Field
[0003] Embodiments disclosed herein relate to systems and methods for electronic biometric identification. Background Art
[0004] Biometric identification is a pervasive tool used in a wide range of environments and is often used to enhance the security and usability of electronic systems. Among other possibilities, biometric identification systems can be used to automatically identify an individual's unique personal characteristics (or features) for purposes such as secure access control and personalization of identity-based device settings. Typically, biometric identification involves the detection of a biometric for comparison with reference biometric data (or enrollment data). Unfortunately, biometric identification systems can be highly sensitive to changes that make one or more detected biometrics unrecognizable. This in turn reduces the accuracy and reliability of the biometric identification system. Summary of the invention
[0005] One embodiment provides a device for updating biometric data. The device includes a biometric sensor and a processor. The processor is configured to: reject an authentication attempt based on a biometric input from the biometric sensor that fails a match determination, accept an additional authentication attempt based on an additional biometric input from the biometric sensor that passes an auxiliary match determination, and update a biometric data repository based on the biometric input.
[0006] Another embodiment includes a method of updating biometric data. The method includes rejecting an authentication attempt based on a biometric input from a biometric sensor that fails a match determination, accepting an additional authentication attempt based on an additional biometric input from the biometric sensor that passes an auxiliary match determination, and updating a biometric data repository based on the biometric input.
[0007] Yet another embodiment includes a non-transitory computer-readable storage medium storing instructions that, when executed by a processor, configure the processor to: reject an authentication attempt based on biometric input from a biometric sensor that fails a match determination, accept an additional authentication attempt based on additional biometric input from the biometric sensor that passes an auxiliary match determination, and update a biometric data repository based on the biometric input. BRIEF DESCRIPTION OF THE DRAWINGS
[0008] Figure 1 is a block diagram depicting an embodiment of an electronic device.
[0009] Figure 2 is a pair of images depicting an embodiment of a varying biometric.
[0010] Figure 3 is a flow chart depicting an embodiment of a method of updating biometric data.
[0011] Figure 4A-4B is a flow chart depicting an embodiment of a method of updating biometric data based on multiple update criteria.
[0012] Figure 5 is a flow chart depicting an embodiment of a method of updating biometric data based on an auxiliary match determination.
[0013] Figure 6 is a flow chart depicting an embodiment of a method of updating biometric data based on an auxiliary match determination.
[0014] Figure 7A-7B is a collection of images and diagrams depicting an embodiment of a method of aligning and comparing biometric images. DETAILED DESCRIPTION
[0015] The following specific embodiments are merely exemplary in nature and are not intended to limit the invention or the application and use of the invention. In addition, it is not intended to be bound by any express or implied theory presented in the previous technical field, background technology, invention content, description of the drawings, the following specific embodiments or the accompanying abstract.
[0016] Go to the attached picture, Figure 1 1 is a block diagram of an electronic device 100 (or electronic system) according to some embodiments. Some non-limiting examples of electronic devices 100 include personal computers of all sizes and shapes, such as desktop computers, laptop computers, netbook computers, tablet computers, web browsers, e-book readers, and personal digital assistants (PDAs). Additional example electronic devices 100 include composite input devices, such as physical keyboards and separated joysticks or key switches. Other example electronic devices 100 include peripheral devices, such as data input devices (including remote controls and mice) and data output devices (including display screens and printers). Other examples include remote terminals, information stations, video game consoles (e.g., video game consoles, portable game devices, etc.), communication devices (including cellular phones, such as smart phones) and media devices (including recorders, editors, and players, such as televisions, set-top boxes, music players, digital photo frames, and digital cameras).
[0017] The electronic device 100 includes a sensor 102 and a processing system 104. As an example, basic functional components of the electronic device 100 that may be used during a biometric matching attempt or authentication attempt are illustrated. The processing system 104 includes (one or more) processors 106, a memory 108, a biometric data repository 110, an operating system (OS) 112, and (one or more) power supplies 114. Each of the (one or more) processors 106, the memory 108, the biometric data repository 110, and the operating system 112 is physically, communicatively, and / or operatively interconnected for inter-component communication. The power supply 114 is interconnected to various system components to provide power when necessary.
[0018] As shown, (one or more) processors 106 are configured to implement functions and / or process instructions for execution within the electronic device 100 and the processing system 104. For example, the processor 106 executes instructions stored in the memory 108 to recognize a biometric or determine whether an authentication attempt is accepted or rejected. The memory 108, which can be a non-transitory computer-readable storage medium, is configured to store information within the electronic device 100 during operation. In some embodiments, the memory 108 includes temporary storage, that is, an area that does not retain information when the electronic device 100 is turned off. Examples of such temporary storage include volatile memory, such as random access memory (RAM), dynamic random access memory (DRAM), and static random access memory (SRAM). The memory 108 also retains program instructions for execution by the processor 106.
[0019] The biometric data repository 110 may be stored on one or more non-transitory computer-readable storage media and / or memory 108. The biometric data repository 110 may include enrollment data (or reference data) for one or more enrolled biometrics (or reference biometrics). The enrollment data may be obtained from a biometric image (e.g., the image itself, a compressed image, an extracted feature set, a digital representation of the image, etc.), one or more biometric objects (e.g., a finger, an iris, a palm print), and all or part of (one or more) biometric objects. In some embodiments, the biometric data repository 110 includes multiple templates that may be associated with appropriate identifiers. For example, the multiple templates may correspond to multiple users, multiple biometric objects, multiple enrollment views of the same biometric object, and / or other enrollment information. More generally, the biometric data repository 110 may be configured to store information about one or more biometrics. The biometric data repository 110 may also be configured for long-term storage of information. In some embodiments, the biometric data repository 110 is stored on a non-volatile storage element. Non-limiting examples of non-volatile storage elements include, among other elements, a magnetic hard disk, a solid state drive (SSD), an optical disk, a floppy disk, flash memory, or in the form of an electrically programmable memory (EPROM) or an electrically erasable and programmable (EEPROM) memory. Although the biometric data repository 110 is shown as being included in the processing system 104, in some embodiments, the biometric data repository 110 or portions thereof may be stored remotely from the processing system 104 and / or the sensor 104.
[0020] The processing system 104 may also host an operating system (OS) 112. The operating system 112 may control the operation of the components of the processing system 104. For example, the operating system 112 may facilitate the interaction of (one or more) processors 106, memory 108, and a biometric data repository 110. Although shown as including a processor 106 and a memory 108, the processing system 104 may also include a microprocessor, a microcontroller, and / or a dedicated circuit. In some embodiments, the processing system 104 sends a message to the sensor 102 to instruct the sensor 102 to capture an image of an object (e.g., such as for an authentication attempt). In addition, in some embodiments, when the device 100 is in sleep mode, the sensor 102 may send an interrupt signal to the processing system 104 when the sensor detects a press on the sensor 102, which causes the electronic device 100 to wake up from the sleep mode. In addition, the processing system 104 may be configured to output prompts or feedback to the user through a user interface (e.g., an electronic display, a speaker, a tactile device, etc.).
[0021] The processor(s) 106 may implement hardware and / or software to obtain an image describing an input object or data describing a biometric. The processor(s) 106 may also align two images and compare the aligned images to each other to determine if a match exists. The processor(s) 106 may also be operable to reconstruct a larger image from a series of smaller partial images or sub-images (such as fingerprint images) when multiple partial fingerprint images are collected during a biometric process, such as an enrollment or matching process for verification or identification. The processor(s) 106 may also perform image processing operations, such as contrast enhancement, scaling, background subtraction, and the like. In some embodiments, the processor(s) 106 include one or more dedicated processors associated with the sensor 102. Additionally or alternatively, the processor(s) 106 may include one or more main processors associated with the electronic device 100.
[0022] The processing system 104 includes one or more power sources 114 to provide power to the electronic device 100. Non-limiting examples of power sources 114 include disposable power sources, rechargeable power sources, and / or power sources developed from nickel-cadmium, lithium-ion, or other suitable materials and power cords and / or adapters that in turn connect to the power sources.
[0023] The sensor 102 may be configured to detect biometric and / or other inputs by interacting with one or more users. The interaction may be active or passive, and may be direct or indirect. As an example, in some embodiments, the sensor 102 and / or one or more input devices such as a keyboard, a touch screen, etc. may be configured to capture input for authentication, such as biometric input and passwords (e.g., passwords, PIN codes, authentication gestures, etc.), etc.
[0024] The sensor 102 may be implemented as a physical part of the processing system 104, or may be physically separated from the processing system 104. Where appropriate, the sensor 102 may communicate with portions of the electronic device 100 using any one or more of the following: a bus, a network, and other wired or wireless interconnections. In some embodiments, the sensor 102 is implemented as a fingerprint sensor to capture a fingerprint image of a user. In other embodiments, the sensor 102 is implemented as a camera that can be used to capture facial or eye biometrics. In yet another embodiment, the sensor is implemented as a microphone for capturing voice biometrics. The sensor device 102 may be incorporated as part of a display, for example, or may be a discrete sensor. In some embodiments, the sensor 102 may be configured for image sensing (e.g., using optical imaging, capacitive imaging, acoustic imaging, etc.).
[0025] The sensor 102 can utilize any suitable combination of sensor components and sensing technologies to detect user input in a sensing area. Some implementations utilize an array of multiple sensing elements or other regular or irregular patterns to detect input. Example sensing technologies that the sensor 102 can use include capacitive sensing technology, optical sensing technology, resistive sensing technology, thermal sensing technology, inductive sensing technology, magnetic sensing technology, and / or radar sensing technology.
[0026] In one example, the sensor 102 may use an inductive sensing technique in which one or more sensing elements pick up a loop current induced by a resonant coil or coil pair. Some combination of the current's amplitude, phase, and frequency may then be used to determine spatial information corresponding to an input object.
[0027] In another example, the sensor 102 may use optical sensing technology in which one or more sensing elements detect light from a sensing area. The detected light may be reflected from an input object, transmitted through an input object, emitted by an input object, or some combination thereof. The detected light may be in the visible light or invisible light spectrum (e.g., infrared or ultraviolet light). Example optical sensing elements include photodiodes, CMOS image sensor arrays, CCD arrays, thin film detectors, and other suitable photosensors that are sensitive to light of the wavelength(s) of interest. Active illumination may be used to provide light to the sensing area, and reflections of the illumination wavelength(s) from the sensing area may be detected to determine input information corresponding to the input object.
[0028] One example optical technique utilizes direct illumination of an input object, which may or may not be in contact with the input surface of the sensing region, depending on the configuration. One or more light sources and / or light guide structures are used to direct light to the sensing region. When an input object is present, the light is reflected directly from the surface of the input object, which reflection can be detected by the optical sensing element and used to determine input information about the input object.
[0029] Another example optical technique utilizes indirect illumination based on internal reflection to detect input objects in contact with the input surface of the sensing area. Due to the different refractive indices at opposite sides of the interface defined by the input surface, one or more light sources are used to direct light in the transmissive medium at an angle at which the light is internally reflected at the input surface of the sensing area. The contact of the input surface by the input object causes the refractive index to change across this boundary, which changes the internal reflection characteristics at the input surface. If the frustrated total internal reflection (FTIR) principle is used to detect the input object, where the light is directed to the input surface at an incident angle at which it is totally internally reflected, a higher contrast signal can sometimes be obtained except where the input object is in contact and causes the light to be scattered in the area where the input object is in contact and partially transmitted through this interface. An example of this situation is the presence of a finger introduced to the input surface defined by a glass to air interface. The higher refractive index of human skin compared to air causes light incident at the input surface at the critical angle of the interface to air to be partially transmitted through the input interface and scattered by the finger, where it would otherwise be completely internally reflected at the glass to air interface. This optical response can be detected by the system and used to determine spatial information. In some embodiments, this can be used to image small-scale surface variations of an input object, such as a fingerprint pattern, where the internal reflectivity of the incident light is different depending on whether a ridge or valley of a finger is in contact with that portion of the input surface.
[0030] In another example, the sensor 102 may use acoustic sensing technology, where one or more acoustic sensing elements detect sound waves from nearby input objects. The sound waves may be audible frequencies or ultrasonic frequencies. The detected sound waves may include echoes of ambient sound waves and / or echoes of sound waves emitted by the input device that are reflected from the surface of the input object. Some combination of the amplitude, phase, frequency, and / or delay of the electrical signal may be used to determine spatial information corresponding to the input object.
[0031] An exemplary acoustic sensing technology utilizes active ultrasonic sensing to transmit high-frequency source waves that propagate to a sensing area. One or more ultrasonic transmitter elements (also referred to as "ultrasonic transmitters") may be used to transmit high-frequency sound waves to the sensing area, and one or more ultrasonic receiving elements (also referred to as "ultrasonic receivers") may detect echoes of the transmitted sound waves. Separate elements may be used for transmission and reception, or an element that both transmits and receives (e.g., an ultrasonic transceiver) may be used. In one implementation, an acoustic pulse is transmitted and reflected at an interface corresponding to an input surface. The acoustic impedance at the interface is different depending on whether a ridge or valley of a finger is in contact with that portion of the input surface, thereby affecting the intensity of the detection wave reflected back from the input surface. In some embodiments, the transmitted ultrasonic wave is capable of penetrating a sub-surface of an input object, such as the dermis of a human finger.
[0032] In another example, the sensor 102 can use capacitive technology, where a voltage or current is applied to generate an electric field. A nearby input object causes a change in the electric field and generates a detectable change in capacitive coupling that can be detected as a change in voltage, current, etc. The sensor electrodes can be used as capacitive sensing elements. An array or other regular or irregular pattern of capacitive sensing elements can be used to generate the electric field. Individual sensor electrodes can be ohmically shorted together to form a larger sensing element.
[0033] One example technique utilizes a "self-capacitance" (or "absolute capacitance") sensing method based on changes in capacitive coupling between sensor electrodes and input objects. An input object near the sensor electrodes changes the electric field near the sensor electrodes, thereby changing the measured capacitive coupling. The absolute capacitance sensing method can operate by modulating the sensor electrodes relative to a reference voltage (e.g., system ground), and by detecting the capacitive coupling between the sensor electrodes and the input object. For example, an array of sensing elements can be modulated, or a drive ring or other conductive element that is ohmically or capacitively coupled to the input object can be modulated. The reference voltage can be a substantially constant voltage or a varying voltage, or the reference voltage can be system ground.
[0034] Another example technology utilizes a "mutual capacitance" (or "transcapacitive") sensing method based on changes in capacitive coupling between sensor electrodes. An input object near the sensor electrodes can change the electric field between the sensor electrodes, thereby changing the measured capacitive coupling. The transcapacitive sensing method can operate by detecting the capacitive coupling between one or more transmitter sensor electrodes (also referred to as "transmitter electrodes") and one or more receiver sensor electrodes (also referred to as "receiver electrodes"). The transmitter sensor electrode can be modulated relative to a reference voltage to transmit a transmitter signal. The receiver sensor electrode can be kept substantially constant relative to the reference voltage to facilitate receiving the result signal. The reference voltage can be a substantially constant voltage or system ground. The transmitter electrode is modulated relative to the receiver electrode to transmit the transmitter signal and to facilitate receiving the result signal. The result signal can include (one or more) effects corresponding to one or more transmitter signals, and / or (one or more) effects corresponding to one or more environmental interference sources (e.g., other electromagnetic signals). The sensor electrode can be a dedicated transmitter or receiver, or can be configured to both transmit and receive. In addition, the sensor electrode can be a dedicated transcapacitive sensing element or an absolute capacitance sensing element, or can be operated as both a transcapacitive and an absolute capacitance sensing element.
[0035] Such as Figure 1Some embodiments of electronic devices such as the electronic device 100 shown in the figure can be configured to detect and recognize biometrics. In general, biometrics include personal physiological or behavioral characteristics (or features) that are unique and provide information about the identity of the user. Examples of biometrics include fingerprints, faces, hand or palm prints, eye characteristics (e.g., irises, eye veins, retina, etc.), vascular characteristics (e.g., finger veins, eye veins, wrist veins, etc.), voice, and behavioral biometrics such as keystroke patterns.
[0036] Recognition of physiological biometrics such as fingerprints and faces is typically based on capturing an image of the biometric using a sensor (e.g., using optical, capacitive, acoustic, or other appropriate sensing methods). Figure 2 , depicts two images of biometrics. More specifically, Figure 2 Two fingerprint images captured using an optical sensor are shown. Figure 2 The two images in exemplify how changing conditions can affect the detection of biometrics, and more specifically, how changing finger skin conditions can affect fingerprint images from an optical fingerprint sensor.
[0037] The first fingerprint image IMG1 is an image obtained from a finger having a typical amount of moisture in the skin. The second fingerprint image IMG2 is an image obtained from the finger shortly after washing hands. Washing hands changes the skin condition on the finger and affects the resulting biometric image. More specifically, washing hands reduces the moisture content of the skin relative to unwashed or typical conditions. When the finger touches the input surface (or platen) of the optical sensor, this changed condition affects the contact area between the input surface and the finger, which in turn affects the interaction of light at the input surface detected by the optical sensor. As shown in Figure 2 As can be seen in the , this change significantly affects the appearance of the fingerprint in the resulting image. This in turn reduces the performance of the biometric recognition system and significantly increases the false rejection rate (FRR) of the authentication system employing an optical fingerprint sensor due to the system's inability to recognize images captured from the altered fingerprint. For example, a user may enroll their fingerprint under a typical skin condition similar to that shown in the first image IMG1. When the same user later provides the same fingerprint in connection with an authentication or matching attempt under a washed finger condition similar to that in the second image IMG2, a false match failure determination is likely due to the altered appearance relative to the enrolled image. The same problem may occur if a user enrolls their fingerprint under a washed finger condition and then later provides their same finger under a typical (unwashed) condition.
[0038] Similar issues of changing biometrics may also occur with other types of fingerprint sensors. For example, fingers that are too dry or too wet can also affect images captured with capacitive and acoustic fingerprint sensors (having a changing effect). In addition, other changes associated with biometric detection can degrade the performance of fingerprint and other biometric recognition systems. For example, posture and lighting changes during the capture phase can affect the appearance of biometrics in facial recognition and fingerprint recognition systems. Voice recognition performance may be degraded due to changes in the detected voice signal and template aging. Any time a biometric recognition system is not robust to biometric-related changes at different instances of detection and recognition, the performance of the system may be affected.
[0039] Reference again Figure 2 , the second fingerprint image IMG2 appears to have low contrast areas with broken and unclear fingerprint ridges and valleys. To the untrained eye, it appears to have performance that is insufficient for use in a biometric recognition system. However, despite this, the second fingerprint image IMG2 is not necessarily low quality or unusable. On the contrary, it has rich biometric information that can be utilized by a biometric recognition system. In some embodiments, recognition performance can be improved by storing a registration data set that is more representative of changes associated with biometrics and / or by adopting a matching process that is more robust to changes associated with biometrics.
[0040] One way to store a more complete biometric data set is to model the effects of changes in the appearance of a biometric, or more generally, the effects of changes in the data or resulting signals obtained from a detected biometric. For example, a mapping from a typical fingerprint illustrated in a first image IMG1 to a washed fingerprint illustrated in a second image IMG2 (or vice versa) can be modeled and artificially applied to generate an artificially altered (e.g., artificially washed) representation of the biometric. The artificially altered representation can then be matched with greater accuracy to a truly altered representation of the biometric detected by the biometric sensor. The transformation that maps the enrolled biometric from one condition to another can be modeled or estimated using any suitable technique, such as by using machine learning, self-organizing methods, and / or empirically based on observed data. In some embodiments, such a transformation can be applied at enrollment or to the enrolled biometric, and the artificially transformed representation can be stored in the biometric data repository. Directly storing an artificial representation of the enrolled biometric can have benefits in terms of reduced matching time or reduced waiting time compared to methods that attempt to apply artificial transformations to new authentication inputs at matching time.
[0041] Another approach to storing a more complete biometric data set is to detect the biometric under various changing conditions and store the data obtained from the biometric detected under the different conditions. The resulting stored data derived from the changed biometric can then provide an enrollment data set that is more representative of the biometric under the changed conditions for later comparison with new inputs. Some approaches may attempt to store a more representative data set through the use of explicit user feedback or prompts that request the user to create the changed conditions during an active enrollment period. However, it may be cumbersome, impractical, or unreliable for the user to create or recreate the same conditions or environment. For example, in a case such as Figure 2 In the case of the washed finger illustrated in , it may be troublesome for the user to register using a typical finger, wash the finger, and then register again using the washed finger.
[0042] To this end, template update schemes that automatically update stored biometric data sets based on new biometric inputs can work well and reduce the burden on users. The template update scheme can obtain new biometric inputs associated with new matching attempts or authentication attempts, and update the enrollment data set based on the new biometric inputs when certain criteria are met. More generally, the template update scheme can detect biometrics outside of an effective enrollment process or in association with other uses of the device. This template update scheme can be called passive enrollment. There may be challenges and risks associated with this approach. For example, an impostor or fake user can compromise the security or performance of the system by entering incorrect biometric data into the enrollment data set. Although new biometric inputs that form the basis of a successful (qualified) match or an accepted authentication attempt can provide a lower risk of an impostor entering the template, they also provide less new information to the enrollment data set. Since they have already fully matched the existing template, the information contained in these successful inputs may be slightly redundant to the existing enrollment data set.
[0043] Figure 3-6Embodiments for updating biometric data in conjunction with biometric match failures and / or biometric-based authentication rejections are disclosed. Data obtained from a match failure can be added to an enrollment data set and / or a match failure can be used as an indicator that a new biometric input will be captured. A successful (accepted) authentication with additional input can be used to increase the confidence that the match failure belongs to an enrolled (one or more) user (real user) and can be used for template updates. Auxiliary matches for unqualified biometric inputs can be introduced in conjunction with additional accepted authentications, which further increases the confidence that the match failure is derived from a real enrolled user and can form the basis for implicit template updates. Additionally or alternatively, auxiliary match determinations for unqualified biometric inputs can be performed in the case of multiple templates to identify appropriate templates for updating and / or use the identification determination for other purposes.
[0044] In some embodiments, the auxiliary match determination can adopt one or more different matching parameters, such as one or more different thresholds, different features, different alignment processes and / or different scoring functions associated with the verification match determination. For example, in some embodiments, the auxiliary match determination uses the same score as the score used for verification, but has a different, relaxed threshold. In some embodiments, the verification match determination and the auxiliary match determination can be based on different scores corresponding to different features. For example, the verification match determination can be based on local features, while the auxiliary match determination can be based on global features. An example of a match determination based on global features would be a match determination based on the calculation of the Euclidean distance between the pixels in the new input image (e.g., the verification image) and the stored (e.g., the registration image) in the best possible alignment determined. As another example, the intensity of the correlation peak can be used. The auxiliary match determination can also be based on a lower security level and / or a more computationally expensive process than the unqualified verification match determination. Since the auxiliary match determination can be used in conjunction with additional strong authentication (e.g., password, alternative biometrics, etc.), a weaker auxiliary match can be tolerated based on the overall high security level provided by the combination of weak auxiliary matches plus strong alternative authentications. Additionally or alternatively, because auxiliary match determinations may be employed in special cases where an initial match fails or in the context of validating implicit template updates, a computationally more expensive or slower process may be accepted without unduly adding frustration to the user experience.
[0045] Reference again Figure 3-6, depicting a flowchart in which one or more steps may be implemented or configured to be implemented by an electronic device 100, (one or more) processors 106, and / or any appropriate device or processing circuitry. In some embodiments, (one or more) processors 106 or other processing circuitry may be configured to implement one or more steps using specially programmed general-purpose hardware (e.g., a CPU configured to perform the steps of the method via one or more programs stored in a memory) and / or dedicated hardware (e.g., a dedicated digital logic circuit block). Although these methods are illustrated in the flowcharts as flowing in a particular order, it should be understood that, unless otherwise stated or clearly contradictory to the context or logic of the method, certain steps may be performed in any appropriate order or sequence. In addition, it should be understood that, unless otherwise stated or clearly contradictory to the context or logic of the method, various steps may be performed serially or in parallel. Similarly, for simplicity, several intermediate steps that are optionally included in various flowcharts are omitted.
[0046] Go to Figure 3 , depicts a method 350 for updating biometric data. In particular, Figure 3 The method 350 depicted in relates to updating a biometric data repository in conjunction with a biometric authentication rejection (or a failed authentication attempt).
[0047] At step 352, the method includes obtaining a biometric input associated with the authentication attempt. At step 354, an authentication decision is made based on the biometric input from step 352 and one or more authentication criteria. When it is determined that the authentication criteria are met, the authentication attempt is accepted at step 356. When it is determined that the authentication criteria are not met (or when the authentication attempt is rejected), an update decision is made at step 358 based on whether one or more update criteria are met. When it is determined that the update criteria are met, the biometric data repository is updated at step 360. When it is determined that the update criteria are not met, the authentication attempt is rejected at step 362 without updating the biometric data repository.
[0048] The biometric input at step 352 may be obtained directly or indirectly from a biometric sensor (or input device) and corresponds to a biometric (e.g., fingerprint, face, voice, etc.) detected by the biometric sensor associated with the authentication attempt. The biometric input may include a signal or data (e.g., an image, an audio signal, etc.) that corresponds to the detected biometric and may be analyzed and further processed to make one or more decisions about the biometric input. In some embodiments disclosed herein, each authentication attempt may be implicit or explicit, as appropriate. For example, the biometric input at step 352 may be obtained explicitly in response to a user's request for authentication and / or in conjunction with the output of a prompt for the user to provide a biometric. Alternatively, the biometric input at step 352 may be obtained implicitly without providing an indication to the user, such as in conjunction with a persistent or continuous authentication attempt. In addition, the acceptance of the authentication attempt at step 356 and / or the rejection of the authentication attempt at step 362 may be performed passively without providing an indication to the user.
[0049] The authentication criteria at step 354 may include one or more biometric matching criteria, and the authentication decision may be based, in whole or in part, on the biometric input from step 352 matching or failing to match the one or more biometric matching criteria. An example of an authentication criterion includes determining a match score based on a comparison between the biometric input received at step 352 and data in a biometric data repository, where the match score may be associated with a multi-modal or multi-factor authentication attempt.
[0050] The match determination may involve comparing the biometric input to stored biometric data (or reference data) from a biometric data repository. The stored biometric data may include enrollment data corresponding to one or more enrolled biometrics, which may have been previously detected by the same biometric sensor used to capture the biometric input from step 352 and / or may have been obtained in other ways (e.g., from a different sensor or database). When it is determined that the biometric input matches the enrollment data with a sufficiently high confidence, the biometric input passes the match determination and the authentication attempt may be accepted at step 356. When it is determined that the biometric input does not match the enrollment data with a sufficiently high confidence, the biometric input fails the match determination and the authentication attempt may be rejected at step 362.
[0051] At step 360, the biometric data repository may be updated with data from the biometric input (from step 352) and / or with data from one or more newly obtained biometric inputs. In some embodiments, the biometric data repository is updated by storing data from biometric inputs that failed a match determination. Additionally or alternatively, when the update criteria at step 358 are met, the biometric sensor may be scanned or operated to obtain one or more additional biometric inputs, and the one or more additional inputs may be used to update the template. The update criteria at step 358 may include any appropriate criteria, and in some embodiments, the update criteria include Figure 4A-4B One or more of the update criteria depicted in the illustrated embodiment, as further described below, may include additional matching determinations and / or additional authentication based on additional input.
[0052] Additionally, it should be noted that biometric data updates may optionally be used separately (e.g., based on other update criteria or conditions) in conjunction with the verification acceptance at step 356. For example, in some embodiments, the biometric input data collected at step 352 may be used to update the data in the biometric data repository when the data meets the authentication criteria at step 354.
[0053] Reference now Figure 4A-4B , a method 450 for updating biometric data based on an auxiliary matching determination is disclosed. Some embodiments of the method 450 may be configured according to other suitable embodiments of the methods disclosed herein. In addition, similar elements are labeled with similar reference numerals where appropriate, and, where appropriate, a more detailed description of elements that may be commonly shared in other disclosed embodiments of these methods is omitted for simplicity.
[0054] Method 450 involves updating biometric data with data from a biometric input that failed a verification match determination (e.g., feature data extracted from the biometric input and / or the biometric input itself). Figure 4A-4B In method 450, multiple update criteria 458a-d may be used to verify a failed authentication to determine that the biometric input that failed to achieve authentication is still used to update the enrollment data set. Figure 4A-4B A plurality of update criteria 458a-d are shown, but may be used Figure 4A-4B Any combination of the update criteria shown in . In addition, it should be understood that other appropriate update criteria not shown in the figure may also be optionally used.
[0055] Additionally, each update criteria determination in method 450, taken individually or collectively, may also be used as a precondition to updating the biometric data at step 460. In particular, in the illustrated embodiment shown in method 450, only in response to positive satisfaction of each of the update conditions 458a-d is the biometric data updated at step 460. Furthermore, as shown, in response to failure of any of the update conditions 458a-d, the biometric data is not updated at 460 or data derived from the failed biometric input is discarded at step 470. However, it should be understood that in alternative embodiments, the determinations at steps 458a-d may more generally form the basis for updating the biometric data at step 460 using more complex logic or decision making. For example, in some embodiments, the update at step 460 may be performed when a weighted or unweighted combination of the various update criteria 458a-d is satisfied. Yet another example would be to perform the update at step 460 when only one or more of the update criteria 458a-d is satisfied.
[0056] Reference now Figure 4A As shown, at step 452, a biometric input is obtained. At step 454, a verification match determination is made on the biometric input based on comparing the biometric input with the enrollment data stored in the biometric data repository. The verification match determination is used as a verification criterion to accept or reject the authentication attempt based on the biometric input passing or failing the verification match determination. When the biometric input passes the verification match determination, the authentication attempt is accepted at step 456. In conjunction with accepting the authentication attempt, access to the operation of the electronic device is authorized at step 464. For example, authorizing access to the operation at step 464 may include unlocking the mobile device or performing a financial transaction in response to accepting the authentication attempt at step 456. Additionally or alternatively, an acceptance message (or indicator) may be output with an output device (e.g., a display, a speaker, a tactile actuator, etc.) in conjunction with or in response to accepting the authentication attempt at step 456. When the biometric input fails to pass the verification match determination at step 454, the authentication attempt is rejected at step 462. In connection with rejecting the authentication attempt at step 462, access to the operation of the electronic device is rejected at step 468. As an example, denying access to the operation at step 468 may include maintaining a locked state of the mobile device or denying the financial transaction in response to denying the authentication attempt at step 462. Additionally or alternatively, a denial message (or indicator) may be output using an output device (e.g., a display, a speaker, a tactile actuator, etc.) in conjunction with or in response to denying the authentication attempt at step 462.
[0057] Also in conjunction with the failed authentication match determination at step 454, data from the failed biometric input is temporarily stored in memory at step 466. Figure 4B , when it is determined that one or more of the plurality of update criteria 458a-d are not satisfied, the data derived from the ineligible biometric input is discarded at step 470. When it is determined that the plurality of update criteria 458a-d are satisfied, the biometric data repository is updated with the data from the ineligible biometric input at step 460. The data from the ineligible biometric input may be stored in the biometric data repository for future matching attempts (e.g., for comparison with future new biometric inputs). As an example, in an embodiment where the biometric input from step 452 is a biometric image, the data stored at step 460 may be a biometric template determined from the image, including a set of features extracted from the image, a digital representation of the image, and / or the full image itself.
[0058] The updated criteria at steps 458a-d include authentication criteria at step 458a (based on additional authentication decisions for additional inputs), timing criteria at step 458b (based on time elapsed since authentication was denied), match-independent criteria at step 458c (based on match-independent classification of failed biometric inputs), and auxiliary match criteria at step 458d (based on auxiliary match decisions for failed biometric inputs). As further described below, updating criteria 458a-d can increase the confidence that data from failed biometric inputs can be added to the template and reduce the likelihood of increasing the false acceptance rate (FAR) by inappropriately adding the wrong user or user data to the enrollment data set.
[0059] refer to Figure 4B, additional input is obtained at step 474 to facilitate the additional authentication determination at step 458a in conjunction with the additional authentication attempt. The additional input at step 474 may be the same biometric modality (or pattern) or a different authentication modality (or pattern) as the biometric input obtained at step 452. The additional input may be a non-biometric authentication modality. For example, in one embodiment, the biometric input may correspond to a fingerprint biometric modality, and the additional input may correspond to a facial biometric modality, an eye biometric modality, or a voice biometric modality. In another embodiment, the biometric input may correspond to a biometric modality (e.g., fingerprint, face, eye, voice, handprint, etc.), and the additional input may correspond to a password modality (e.g., a password, PIN code, or user-defined gesture). In some embodiments, each individual input (e.g., the biometric input at step 452 and the additional input at step 474) may be associated with or obtained from: a separate scan of a biometric sensor, a separate input action with a target biometric by a user, or a combination thereof.
[0060] In one embodiment, the additional input corresponds to the same biometric modality and is obtained from the same biometric sensor as the biometric input from step 452, and the sensing parameters of the same biometric sensor are modified before obtaining the additional input. For example, the integration time or gain setting of the optical sensor may be adjusted to improve detection relative to the biometric input from step 452. In one embodiment, the biometric sensor is configured to repeatedly scan or capture biometrics and match each of the scans (or captures) until one of them successfully matches. In this embodiment, the additional input from step 474 may correspond to any one or more of the repeated scans. In another embodiment, the biometric sensor is configured to capture multiple scans and combine two or more of those scans to obtain the additional input at step 474. For example, an image in the biometric input from step 452 may be averaged together with one or more images from one or more subsequent scans obtained with the biometric sensor, or two or more subsequent scans obtained with the biometric sensor may be averaged together to provide the additional input at step 474. In one embodiment, this may improve the signal-to-noise ratio (SNR) for the additional input from step 474 relative to the biometric input from step 452.
[0061] In the illustrated method 450, the additional input at 474 is obtained after or in conjunction with outputting a prompt to the user at step 476, but in alternative embodiments, the prompt at step 476 may be omitted. The prompt may include a request for another input mode (e.g., another biometric or a password). Additionally or alternatively, the prompt may include user feedback to perform an action with the detected biometric from step 452. For example, the user feedback may include a prompt to increase pressure with a biometric object or to increase pressure with a fingerprint against an input surface. Additionally or alternatively, it may include a prompt to contact the biometric input location with the input surface or to contact the fingerprint with the input surface. In the case of an optical fingerprint sensor, increasing pressure or maintaining contact for a longer period of time may improve the quality of contact between the input surface and the fingerprint, thereby improving the quality of subsequent fingerprint detection. In another embodiment, the prompt may include a request for the user to touch their nose with their fingerprint, which in the case of a fingerprint sensor may improve the wet condition of the fingerprint. More generally, the prompt may include user feedback or requests.
[0062] At step 458a, method 450 determines whether one or more additional authentication criteria are satisfied based on the additional input from step 474. The authentication criteria at step 458a may be satisfied based on biometric and / or non-biometric authentication as described above. For example, the authentication criteria at step 458a may be satisfied when it is determined that the additional input matches a registered biometric or when the additional input matches a stored password. When the authentication criteria at step 458a are satisfied, the additional authentication attempt is accepted at step 478. When the authentication criteria cannot be satisfied, the additional authentication attempt is rejected at step 480. In addition, data derived from the failed biometric input may be discarded at step 470 in response to the rejection of the additional authentication attempt at step 480.
[0063] At steps 458b-d, method 450 determines whether a plurality of additional update criteria are satisfied. Figure 4B4, update criteria 458b-d are shown as being performed serially, and further, the authentication determination at step 458a is shown as being performed in parallel with update criteria 458b-d. However, in various embodiments of method 450, any two or more of the update determinations at steps 458a-d may be performed in parallel and / or serially with each other. At step 458b, method 450 determines whether a time period (e.g., time period T=2 seconds, 10 seconds, or another suitable time period) has passed since the authentication attempt (or more specifically the verification attempt from step 454) based on the biometric input from step 452. In some embodiments, the time period is determined by starting a counter in response to the biometric input at step 452, the verification determination at step 454, or another suitable proxy for the authentication attempt associated with the biometric input from step 452. When it is determined at step 458b that the time period has passed before an additional authentication attempt, data from the biometric input is discarded at step 470. When it is determined at step 458b that an additional authentication attempt occurred within the time period, the data from the biometric data repository is updated at step 460 to include data from the unqualified biometric input. Figure 4B In some embodiments, method 450 may additionally or alternatively determine whether the additional authentication attempt from step 458a immediately follows the verification authentication attempt from step 454. The additional authentication attempt may be determined to immediately follow the verification authentication attempt when there are no intervening authentication attempts between the additional authentication attempt and the verification authentication attempt.
[0064] At step 458c, method 450 determines a match-independent classification of the biometric input. The match-independent classification may include one or more determinations regarding inherent characteristics of the biometric input from step 474, regardless of comparison to the enrolled biometric data. In some embodiments, the match-independent classification at step 485c includes a quality determination, a sensor coverage determination, a liveness determination, and / or a changed condition determination.
[0065] The quality determination may be based on any suitable quality metric or image quality analysis (e.g., based on a contrast metric, extracted features, etc.) The sensor coverage determination may be based on a determination of whether a fingerprint sensor or other biometric sensor is adequately covered by a fingerprint or another biometric (e.g., based on analysis of an image from a biometric input and / or by segmenting the image into fingerprint pattern and non-fingerprint pattern regions).
[0066] The biometric input may pass the sensor coverage determination when it is determined that it adequately covers the sensor and fail the sensor coverage determination when it is determined that it has insufficient coverage of the sensor. As an example, the sensor coverage determination may be used to abandon the template update in the case of a partial touch (containing insufficient biometric information).
[0067] A liveness determination may involve a determination of whether a biometric input is obtained from a genuine living biometric (not a spoof or a copy). When the biometric input is determined to be from a living biometric, the biometric input may pass the liveness determination, and when the biometric input is determined to be from a spoof (or not from a living biometric), the liveness determination may not pass. For example, the liveness determination may be based on any suitable liveness metric or anti-spoofing analysis (e.g., based on image intensity, statistics such as average grayscale value, etc.).
[0068] When the biometric input is determined to correspond to the desired changed condition for inclusion in the enrollment data set, the changed condition determination may pass, and when the biometric input is determined not to correspond to the desired changed condition, the changed condition determination may fail. Figure 2 In the case of the fingerprint illustrated in , when it is determined that the biometric input corresponds to a washed finger rather than a wet finger or a typical (unwashed) finger, the changed condition determination can pass. As an example, any of the match-independent classifications can be based on the following: thresholding the score calculation, a neural network based on features extracted from the image, a deep learning network, a non-linear scoring function, etc.
[0069] It should be noted that while the classifications at step 458c are shown as match-independent decisions, in some alternative embodiments, one or more of these classifications are made in a match-dependent manner based on comparisons to stored or enrolled biometric data. For example, a quality or liveness decision may be made based in whole or in part on the comparison, or features or statistics derived from the comparison, which may be separate from or combined into the auxiliary match decision below. For example, a neural network trained on all different classes of desired images and receiving all appropriate features, statistics, or metrics as input may be used to make multiple decisions simultaneously.
[0070] At step 458d, a secondary match determination is made based on the biometric input that failed the verification match determination from step 454. For example, the secondary match determination can be based on different thresholds, parameters, alignments, scores, or features as further described herein. When the biometric input passes the secondary match determination, the biometric data repository is updated at step 460, and when the biometric input fails the secondary match determination, the data derived from the biometric input is discarded at step 470.
[0071] Template updates may involve modifications to the enrollment data set, which could potentially allow the imposter unrestricted future access if the template is inadvertently updated with the imposter's data. In addition, unlike making a single false rejection, updating a template with incorrect data may result in an increase in FRR or otherwise negatively impact future performance in many instances of authentication. Accordingly, in some embodiments, template updates may be associated with an overall higher security requirement (compared to ordinary access control functions such as device unlocking). This higher security level may be enforced using an auxiliary matching criterion (which itself has lower security than the primary matching criterion used to make the authentication acceptance / rejection decision), but coupled with an additional strong authentication (e.g., password authentication or a different biometric modality) to jointly provide an overall higher security level. Alternatively, an auxiliary matching criterion may be used that itself provides a higher security level (without requiring another authentication), but is computationally more expensive than the matching criterion used for the authentication acceptance / rejection decision.
[0072] Reference now Figure 5 , a method 550 for updating biometric data based on an auxiliary matching determination is disclosed. Certain embodiments of the method 550 may be configured according to other suitable embodiments of the method disclosed herein. In addition, similar elements are labeled with similar reference numerals where appropriate, and, where appropriate, a more detailed description of elements that may be commonly shared in other disclosed embodiments of the method is omitted for simplicity.
[0073] exist Figure 5 In the embodiment of , the auxiliary match determination forms the basis for updating the biometric data repository 510 in conjunction with the failed authentication match attempt. Figure 2 The auxiliary match determination shown in FIG. 1 is based on a relaxed threshold, and further, the auxiliary match determination can be used to identify templates associated with authentication failures. The relaxed threshold can correspond to a lower degree of consistency between the biometric input and the stored biometric data. When used alone, such a relaxed threshold can generally indicate a lower degree of confidence in the match of the failed biometric input. However, when the relaxed threshold is used in conjunction with additional authentication (e.g., based on a Figure 4B 474 in step 476), the relaxed threshold along with the additional authentication can provide an overall high confidence level in allowing the template to be updated without compromising the security or performance of the biometric recognition system.
[0074] Reference again Figure 5 , at step 582, a plurality of enrollment biometric inputs are detected and data from the plurality of enrollment biometric inputs is stored in a plurality of templates 584 contained in biometric data repository 510. As an example, in an embodiment where the enrollment input from step 582 is a biometric image, the data in the plurality of templates 584 may include a set of features extracted from the image, a digital representation of the image, and / or the full image itself. In one embodiment, the plurality of enrollment inputs are detected during an active enrollment process prior to detection of a new biometric input at step 552. Figure 5 , but in alternative embodiments, any suitable number of templates may be employed. Each of the plurality of templates 584 may be associated with a corresponding identifier (e.g., a template ID) to facilitate an identification decision based on which of the plurality of templates 584 is determined to match the new biometric input. In general, the plurality of templates may correspond to any suitable plurality of registered biometric data sets for which identification or separate template storage may be desired. For example, the plurality of templates may be derived from: the same or different users, the same or different biometric modalities, the same or different portions of the biometrics, etc. The identification decision may involve: identification of a particular registered user among a plurality of registered users, identification of a particular finger among a plurality of registered fingers (from the same or different users), or more generally, identification of a particular template among a plurality of registered templates 584. In an example embodiment employing fingerprint identification, the plurality of templates 584 may correspond to a plurality of different registered fingers, respectively, wherein each of the templates includes data derived from one or more images of the corresponding registered finger.
[0075] Refer to the figure again Figure 5 , a biometric input is obtained at step 552. At step 586, a match score is determined based on a comparison of the biometric input with the enrollment data stored in the biometric data repository 510. At step 554, a verification match determination is made based on comparing the match score to a verification threshold. When the match score meets the verification threshold, the authentication attempt is accepted at step 556. When the match score fails to meet the verification threshold, the authentication attempt is rejected at step 562. When the authentication attempt is rejected, a secondary match determination is made based on whether the match score from step 586 meets a secondary threshold at step 558d. When the match score meets the secondary threshold (and such as Figure 4BWhen the other criteria for the additional authentication from step 458a shown in FIG. 5 are met), the biometric data repository 510 is updated at step 560. More specifically, the biometric data repository 510 can be updated by storing data from the biometric input that failed the verification match determination. When the match score fails to meet the auxiliary match threshold, the biometric data repository 510 is not updated with the biometric input from step 552, and the data from the failed biometric input is discarded at step 570. Figure 5 In the method 550 shown in FIG. 4 , the auxiliary match threshold is a relaxed threshold relative to the verification threshold from step 554. More specifically, the auxiliary match threshold at step 558d can correspond to a lower consistency between the biometric input and the enrollment data set than the verification threshold. Depending on the (one or more) scoring metrics used, the relaxed threshold can be numerically lower or higher. For example, when a difference metric is used, a numerically lower score can indicate a higher degree of consistency, while when a similarity metric is used, a numerically higher score can indicate a higher degree of consistency. When used in conjunction with additional authentication (such as step 458a in FIG. 4 ), the auxiliary match determination based on the relaxed threshold can provide an additional confidence measure that the previous biometric input was obtained from the real registered user, rather than an imposter or another user who interacted with the biometric sensor before the subsequent authentication accepted at step 458a. For example, in some instances a real registered user may wish to provide temporary access to another, unregistered user. If an unenrolled user intentionally or unintentionally interacts with the biometric sensor prior to an accepted additional authentication attempt, data from the unenrolled user may be inadvertently added to the template in embodiments where auxiliary match determination is not employed.
[0076] In some embodiments, a template in the template set 584 can be identified based on a template that generates the best match score that fails to pass a verification threshold but satisfies an auxiliary threshold. The auxiliary recognition determination can be used to identify an appropriate template to be updated in the plurality of templates 584 and / or the recognition determination can be used for any other appropriate purpose. Accordingly, in some auxiliary recognition embodiments, determining the match score at step 586 includes determining one or more match scores for the plurality of templates 584 stored in a biometric data repository, and the verification determination at step 554 can be further based on the one or more match scores. When one of the match scores satisfies the verification threshold, the match authentication attempt can be accepted at step 556. When the match score fails to satisfy the verification threshold, the verification attempt is rejected. At step 558d, the match scores that fail to satisfy the verification threshold are checked against the auxiliary threshold, and when at least one match score satisfies the auxiliary threshold (and such as Figure 4B458a are met), updating the biometric data repository 510. When more than one match score is determined and at least one of the match scores meets the secondary threshold, the method 550 may also determine the best match score among the multiple match scores that failed to meet the verification threshold and identify the template that generated the best match score. The biometric data repository 510 may be updated at step 560 by storing data from the failed biometric input into the identified template. The match score at step 586 may be determined using any suitable match scoring function (e.g., a neural network based on extracted features, a deep learning network, a linear or non-linear scoring function, etc.).
[0077] In some embodiments, the biometric is susceptible to changes that are so significant that different matching parameters, alignment processes, and / or feature sets are appropriate to assist in the match determination. For example, local features (e.g., points, local patches, etc.) and / or feature-based image alignment can provide high match confidence at the verification stage in most cases, while global features (e.g., intensity images, gradient images, binary images, global statistics, etc.) and / or intensity-based image alignment (e.g., correlation alignment, coarse-to-fine pyramid-based alignment, minimization / maximization of distance functions, etc.) can provide higher accuracy verification at the auxiliary matching stage in the case of a changing biometric and / or in cases where a lower security level may be acceptable for assisting the match determination (due to other additional authentication).
[0078] Reference again Figure 2 In the fingerprint images shown in FIG. 1 , the appearance difference between image IMG1 and image IMG2 of the same finger but in different conditions is different enough that a different alignment process and / or feature set may be more suitable for each condition. Figure 2 It is shown that the image IMG1 of a finger in a typical skin condition displays well-defined ridges and valleys and, in contrast to those shown in the image IMG2 of a washed finger, highly oriented textures. Accordingly, when comparing typical fingerprint images such as shown in image IMG1, feature points such as minutiae (e.g., bifurcations and ridge ends) and points of interest (e.g., corresponding to spots, SIFT points, etc.) can provide strong performance in the alignment and / or scoring stages due to their relatively high reliability, repeatability, and discrimination. However, such feature points may not be repeatable, or there may be errors in the point feature detection or feature-based alignment (or registration) process in the washed finger image IMG2. Accordingly, direct comparison of strengths for image alignment and / or match scoring can provide better or more robust matching performance for certain altered images or when attempting to match between altered images.
[0079] refer to Figure 6 , a method 650 is disclosed, which involves verification matching decisions based entirely or partially on feature points and auxiliary matching decisions based entirely or partially on intensity images (or grayscale images). Figure 7A-7B Depicts the use of intensity-based correlation alignment ( Fig. 7A ) compares the unaltered biometric images (enrolment view and verification view) and uses intensity-based correlation alignment ( Figure 7B ) compares an example of changing biometric images (enrollment view and verification view). In this example, the alignment includes determination of rotation, but in various embodiments, the alignment may involve determination of any suitable transformation (e.g., rigid or affine transformation involving rotation, translation, scaling, etc.). In addition, in Figure 7A-7B In the example shown, the strength of the correlation peak may be used as a score or scoring metric for aiding in making a match decision, but other metrics such as the Euclidean distance between ridge pixels may additionally or alternatively be used.
[0080] Some embodiments 650 of the method can be configured according to other suitable embodiments of the method disclosed herein. In addition, similar elements are indicated by similar reference numerals where appropriate, and, where appropriate, for simplicity, a more detailed description of elements that may be shared in other disclosed embodiments of the method is omitted.
[0081] Reference again Figure 6 At step 682, a plurality of enrollment biometric inputs are obtained, and data from the plurality of enrollment biometric inputs is stored in a plurality of templates 684 contained in a biometric data repository 610. The enrollment inputs may be detected based on image sensing performed with a biometric sensor. Each of the plurality of templates 684 may be associated with a corresponding identifier (e.g., a template ID) to facilitate a recognition decision based on which of the plurality of templates is determined to match the new biometric input. Each of the enrollment inputs is stored in one or more representations that allow a plurality of desired features and / or image representations to be derived from the same template. For example, this may involve storing the desired feature sets and image representations themselves, or storing a sufficient enrollment data set that allows these feature sets and image representations to be determined when matching. For example, in one embodiment, both the feature points themselves and the intensity image may be stored in each of the plurality of templates 684. In another embodiment, the intensity image itself is stored in each of the plurality of templates 684, and the feature points are determined (or extracted) from the enrollment intensity image when matching.
[0082] At step 652, a biometric input is obtained. At step 686a, a match score is determined based on a comparison of the biometric input with the enrollment data stored in the biometric data repository 610. More specifically, Figure 6 In the method 650 shown in FIG. 6 , a verification match score is determined based in whole or in part on comparing feature points from the biometric data repository 610 with feature points from the biometric input. At step 654, a verification match determination is made based on comparing the verification match score to a threshold. When the verification match score meets the threshold, the authentication attempt is accepted at step 656. When the verification match score fails to meet the threshold, the authentication attempt is rejected at step 662.
[0083] When the authentication attempt is denied, a secondary match score is determined at step 686b based on a comparison of the biometric input with the enrollment data stored in the biometric data repository 610. More specifically, the secondary match score may be determined in whole or in part based on a direct comparison of the intensity from the intensity image corresponding to the biometric input with the intensity of the intensity image from the biometric data repository 610.
[0084] See again Figure 6 , at step 658d, an auxiliary match determination is performed to determine whether the auxiliary match score from step 686b satisfies a threshold. When the auxiliary match score satisfies the auxiliary threshold, the biometric data repository 610 is updated at step 660. More specifically, the biometric data repository 610 is updated by storing data derived from the biometric input from step 652 that failed the verification match determination. For example, the stored data may be image intensity and / or feature points. When the auxiliary match score fails to satisfy the auxiliary match threshold, the biometric data repository 610 is not updated with the biometric input from step 652, and the data derived from the failed biometric input is discarded at step 670.
[0085] In some embodiments, templates in template set 684 can be identified based on templates that generate the best auxiliary match scores separate from failed verification match scores. Auxiliary recognition decisions can be used to identify the correct template to update in the template set and / or recognition decisions can be used for any other suitable purpose.
[0086] According to one assisted recognition embodiment, determining the assisted match score at step 658b includes determining a plurality of assisted match scores for a plurality of corresponding templates 684 stored in a biometric data repository. The assisted match determination at step 658d is based on each of the plurality of assisted match scores. When one of the plurality of assisted match scores satisfies an assisted threshold, the template that produces the best assisted match score may be updated at step 660. When each of the plurality of assisted match scores fails to satisfy the assisted threshold, the data derived from the biometric input 652 is discarded at step 670.
[0087] Figure 6 and Figure 7A-7B Specific examples are provided with reference to feature points and intensity images for verification matching and auxiliary matching. However, in various embodiments of method 650, verification matching decisions can generally rely on any suitable local features (e.g., points, local patches, etc.) and / or feature-based image alignment. In addition, auxiliary matching decisions can generally rely on any suitable global representation of the input (e.g., intensity image, gradient image, binary image, global statistics, etc.) and / or intensity-based image alignment (e.g., correlation alignment, coarse-to-fine pyramid-based alignment, minimization / maximization of distance functions, etc.).
[0088] The use of the terms "one", "an", "said", and "at least one" and similar indicators in the context of describing the present invention (particularly in the context of the following claims) is to be interpreted as covering both the singular and the plural, unless otherwise specified herein or clearly contradictory according to the context. The use of the term "at least one" (e.g., "at least one of A and B") followed by a series of one or more items is to be interpreted as meaning an item (A or B) selected from the listed items or any combination of two or more items (A and B) in the listed items, unless otherwise specified herein or clearly contradictory according to the context. Unless otherwise specified, the terms "comprising", "having", "including" and "containing" are to be interpreted as open terms (i.e., meaning "including, but not limited to"). The recording of the range of values herein is intended to serve as a shorthand method of referring to each separated value falling within the range individually, unless otherwise specified herein, and each separated value is incorporated into the specification as if it is recorded individually herein. All methods described herein can be performed in any suitable order, unless otherwise specified herein or clearly contradictory according to the context. Unless otherwise stated, the use of any and all examples or exemplary language (e.g., "such as") provided herein is intended merely to better illustrate the invention and does not cause limitations on the scope of the invention. No language in the specification should be construed as indicating any non-claimed element as essential to the practice of the invention.
[0089] Preferred embodiments are described herein, but variations of those preferred embodiments may become apparent to those of ordinary skill in the art after reading the foregoing description. The inventors expect that skilled artisans will employ such variations as appropriate, and the inventors intend for the invention to be practiced in a manner different from that specifically described herein. Accordingly, the invention includes all modifications and equivalents of the subject matter recited in the appended claims as permitted by applicable law. Furthermore, any combination of the above elements in all possible variations thereof is encompassed by the invention, unless otherwise specified herein or otherwise clearly contradicted by context.
Claims
1. A device for updating biometric data, the device include: Biometric sensors; as well as A processor, the processor being configured to: denying an authentication attempt based on a biometric input from the biometric sensor failing a match determination; accepting an additional authentication attempt based on additional biometric input from the biometric sensor determined by a secondary match; as well as updating a biometric data repository based on the biometric input, The processor is further configured to: The biometric data repository is updated by storing in the biometric data repository data from the biometric input that failed the match determination.
2. The apparatus of claim 1, wherein the processor is further configured to: further rejecting the authentication attempt based on the biometric input failing to pass a plurality of first match determinations against a plurality of templates stored in the biometric data repository; making the plurality of first match determinations based on a plurality of match scores that fail to meet a first match threshold; as well as A first template is identified by determining a best match score among the plurality of match scores that fail to meet the first match threshold.
3. The apparatus of claim 2, wherein the processor is further configured to: A second template among the plurality of templates is identified based on the auxiliary match decision by determining a best auxiliary match score among a plurality of auxiliary match scores. The apparatus of claim 3 , wherein the first template is identical to the second template.
5. The apparatus of claim 3, wherein when the processor updates the biometric data repository based on the biometric input, the processor updates one of the first template or the second template.
6. The device of claim 1, wherein the processor is further configured to: determining a match score based on comparing the biometric input to enrollment data stored in the biometric data repository; Making the match determination based on the match score failing to meet a match threshold; and The auxiliary match determination is performed based on the match score satisfying an auxiliary match threshold, wherein the auxiliary match threshold is relaxed relative to the match threshold.
7. The apparatus of claim 1, wherein the processor is further configured to: determining a match score based on comparing the biometric input to enrollment data stored in the biometric data repository; determining a secondary match score based on comparing the secondary biometric input to the enrollment data stored in the biometric data repository; Performing the matching determination based on the matching score failing to meet a threshold; as well as The auxiliary matching determination is performed based on the auxiliary matching score satisfying the threshold.
8. The apparatus of claim 1, wherein the processor is further to update the biometric data repository based on a determination that the additional authentication attempt occurs within a specified time period of a rejected authentication attempt.
9. The device of claim 1, wherein after rejecting the authentication attempt, the processor is configured to repeatedly scan the biometric sensor until a qualified match determination is made.
10. The device of claim 1, wherein the biometric input and the additional biometric input correspond to different modalities.
11. The apparatus of claim 10, wherein the additional biometric input corresponds to a different biometric modality than the biometric input.
12. The apparatus of claim 10, wherein the additional biometric input corresponds to a non-biometric modality.
13. A method for updating biometric data, the method include: rejecting the authentication attempt based on biometric input from the biometric sensor failing a match determination; accepting an additional authentication attempt based on additional biometric input from the biometric sensor determined by a secondary match; as well as updating a biometric data repository based on the biometric input, Wherein updating the biometric data repository based on the biometric input includes updating the biometric data repository by storing in the biometric data repository data from the biometric input that failed the match determination.
14. The method according to claim 13, further comprising: include: further rejecting the authentication attempt based on the biometric input failing to pass a plurality of first match determinations against a plurality of templates stored in the biometric data repository; making the plurality of first match determinations based on a plurality of match scores that fail to meet a first match threshold; as well as A first template is identified by determining a best match score among the plurality of match scores that fail to meet the first match threshold.
15. The method according to claim 14, further comprising: include: A second template among the plurality of templates is identified based on the auxiliary match decision by determining a best auxiliary match score among a plurality of auxiliary match scores. The method of claim 15 , wherein the first template is identical to the second template.
17. A non-transitory computer-readable storage medium storing instructions, which, when executed by a processor, configure the processor to: rejecting the authentication attempt based on biometric input from the biometric sensor failing a match determination; accepting additional authentication attempts based on additional biometric input from the biometric sensor determined by a secondary match; and updating a biometric data repository based on the biometric input, Wherein when the processor updates the biometric data repository based on the biometric input, the processor updates the biometric data repository by storing in the biometric data repository data from the biometric input that failed the match determination.
18. A computer program product, the computer program product comprising a computer program, It is characterized in that The computer program, when executed by a processor, implements the steps of the method of updating biometric data according to any one of claims 13-16.
