A living body authentication method and device

Abstract

The application discloses a living body authentication method and device, relates to the field of data processing, and aims to solve the problem of misjudging a measured object as a living body object. The method comprises the following steps: in response to a living body authentication instruction, emitting an optical signal through a display screen; determining eye movement information of the measured object based on an image sequence collected by an image collection unit and the optical signal; collecting device movement information of a terminal device through an inertial measurement unit; and performing cross-correlation analysis on human biological characteristics based on the eye movement information and the device movement information, so as to determine whether the measured object meets the living body authentication. Through the above scheme, the living body object can be accurately identified.

Description

Technical Field

[0001] This application relates to the field of data processing, and more particularly to a method and apparatus for liveness authentication. Background Technology

[0002] Liveness detection refers to the process of collecting and analyzing images, actions, and other information of the person being verified to determine whether the person being verified is a real, living person, in order to prevent photos, videos, or other forged media from impersonating legitimate users.

[0003] Therefore, how to more accurately identify living objects is a problem that needs to be solved. Summary of the Invention

[0004] This application provides a liveness detection method and apparatus for accurately identifying living objects.

[0005] In a first aspect, this application provides a liveness authentication method, which is applied to a terminal device including a display screen, an inertial measurement unit, and an image acquisition unit. The method includes: in response to a liveness authentication command, transmitting an optical signal through the display screen; determining the eye movement information of the tested object by acquiring an image sequence and the optical signal through the image acquisition unit; acquiring the device motion information of the terminal device through the inertial measurement unit; and performing cross-correlation analysis of human biological characteristics based on the eye movement information and the device motion information to determine whether the tested object meets the requirements for liveness authentication.

[0006] This application emits optical signals through a display screen and collects eye movement information and device movement information of the terminal device, so that liveness detection no longer relies solely on the surface image or preset actions of the tested object, but comprehensively utilizes eye movement information (i.e. the test object's response to the optical signal) and physical device movement information of the terminal device, thereby improving the recognition capability of liveness detection.

[0007] In one possible implementation, a cross-correlation analysis of human biological characteristics is performed based on eye movement information and device motion information to determine whether the subject meets the requirements for liveness certification. This includes: performing cross-correlation analysis on eye movement information and device motion information to obtain phase delay parameters and motion gain parameters; the phase delay parameter is used to characterize the time difference between hand movement and eye movement of the subject; the motion gain parameter is used to characterize the amplitude relationship between hand movement and eye movement of the subject; when the phase delay parameter satisfies the central nervous system conduction mechanism of the human body and the motion gain parameter satisfies the vestibulo-ocular reflex mechanism, the subject is determined to meet the requirements for liveness certification.

[0008] When a terminal device is held in hand, slight hand tremors cause spatial movement, and the eyes of a real living person will produce a compensating rotation in the opposite direction and with a matching amplitude. By performing cross-correlation analysis on the motion information of the terminal device and the angular velocity vector of the eyeball, and verifying the inverse compensation relationship between the two using motion gain parameters, it is possible to determine whether the eye image motion is bound to the actual physical motion of the terminal device, thus avoiding misidentifying two-dimensional forged videos as living individuals. Simultaneously, phase delay parameters are used to verify whether the neural conduction time from motion perception to the eyeball executing the compensating rotation is within a preset physiological range, distinguishing between mechanically synchronized motion lacking neural conduction delay and forged videos with significant delays. Therefore, by simultaneously verifying phase delay and motion gain parameters, the ability to identify two-dimensional video playback, robotic arms holding forged media, and real-time generative video attacks can be improved.

[0009] In one possible implementation, the eye motion information and device motion information are obtained from image sequences acquired within a preset authentication time window. Cross-correlation analysis is performed on the eye motion information and device motion information to obtain phase delay parameters and motion gain parameters. This includes: performing cross-correlation analysis on the eye motion information and device motion information within a sliding time window to determine the frame offset when the cross-correlation value is maximized; determining the phase delay parameter based on the frame offset and the sampling frame interval; and calculating the motion vector for the eye motion information and the corresponding device motion information at the time point indicated by the maximum cross-correlation value to obtain the motion gain parameters.

[0010] One possible implementation also includes: if the device motion information does not meet the certification criteria, then the head motion information of the tested object is determined through the image sequence; based on the eye motion information and head motion information, a cross-correlation analysis of the vestibular-ocular reflex mechanism is performed to determine whether the tested object meets the liveness certification criteria.

[0011] In one possible implementation, before performing cross-correlation analysis of human biological characteristics based on eye movement information and device movement information, the method further includes: determining the retrograde light rays after the optical signal passes through the eyeball based on the optical signal corresponding to any image frame and a preset multilayer media optical model of the eyeball; determining the theoretical imaging point in the image frame after the eye region of the test subject reacts to the optical signal based on the retrograde light rays; and determining that the test subject has a living eyeball based on the actual imaging point and the theoretical imaging point in the image frame.

[0012] Therefore, by utilizing the multi-layered optical path constraints of a real eyeball, two-dimensional display images or three-dimensional forged media that do not conform to a preset multi-layered optical model of the eyeball can be identified.

[0013] In one possible implementation, before emitting optical signals through the display screen, the process further includes: the terminal device emitting a high-frequency micro-vibration sequence; determining the vibration response information of the terminal device through an inertial measurement unit, and determining the state of the terminal device based on the vibration response information, including a handheld state and a fixed state; and performing a cross-correlation analysis of human biological characteristics based on eye movement information and device motion information to determine whether the tested object meets the requirements for liveness authentication, including: when the terminal device is in a handheld state, performing a cross-correlation analysis of human biological characteristics based on eye movement information and device motion information to determine whether the tested object meets the requirements for liveness authentication.

[0014] In one possible implementation, the method further includes: when the terminal device is in a fixed state: based on the image sequence, determining the optical flow displacement information of the image captured by the image acquisition unit due to the vibration of the terminal device; and based on the device motion information and the optical flow displacement information, determining whether the object under test meets the requirements for liveness authentication.

[0015] By actively emitting a high-frequency micro-vibration sequence from the terminal device, and determining whether the terminal device is in a handheld or fixed state based on the motion information and vibration response information collected by the inertial measurement unit, the appropriate liveness detection method can be selected according to the load status of the terminal device, thus taking into account both handheld and fixed scenarios and reducing misjudgments caused by scenario differences.

[0016] In one possible implementation, the optical signal is emitted through the display screen, including: generating a pseudo-random spatial encryption matrix based on a one-time random number; the one-time random number becomes invalid after the liveness authentication is completed; and emitting an optical signal conforming to the pseudo-random spatial encryption matrix through a sub-pixel array set in the display screen.

[0017] By generating invisible optical signals through a pseudo-random spatial encryption matrix, the optical challenges used in each authentication are made random and imperceptible. This makes it difficult for attackers to pre-record or accurately predict the corresponding eye optical responses, thereby improving resistance to historical video replay and real-time forgery attacks, while reducing the impact of explicit action commands on the user authentication experience.

[0018] Secondly, this application provides a liveness detection device, the device comprising: The transmitting module is used to transmit optical signals through the display screen in response to a liveness authentication command; The acquisition module is used to determine the eye movement information of the object under test by acquiring image sequences and optical signals through the image acquisition unit; and to acquire the device motion information of the terminal device through the inertial measurement unit. The determination module is used to perform cross-correlation analysis of human biological characteristics based on eye movement information and device movement information to determine whether the tested object meets the requirements for liveness certification.

[0019] In one possible implementation, the determination module is specifically used to: perform cross-correlation analysis on eye movement information and device movement information to obtain phase delay parameters and motion gain parameters; the phase delay parameters are used to characterize the time difference between hand movement and eye movement of the test subject; the motion gain parameters are used to characterize the amplitude relationship between hand movement and eye movement of the test subject; when the phase delay parameters satisfy the central nervous system conduction mechanism of the human body and the motion gain parameters satisfy the vestibular-ocular reflex mechanism, the test subject is determined to meet the requirements for liveness authentication.

[0020] In one possible implementation, the eye motion information and device motion information are obtained from image sequences acquired within a preset authentication time window; the determination module is specifically used to: perform cross-correlation analysis on the eye motion information and device motion information within a sliding time window to determine the frame offset when the cross-correlation value is maximized; determine the phase delay parameter based on the frame offset and the sampling frame interval; and calculate the motion vector for the eye motion information and the corresponding device motion information at the time point indicated by the maximum cross-correlation value to obtain the motion gain parameter.

[0021] In one possible implementation, the determination module is further configured to: determine the head motion information of the tested object through an image sequence if the device motion information does not meet the certification criteria; and perform a cross-correlation analysis of the vestibular-ocular reflex mechanism based on the eye motion information and head motion information to determine whether the tested object meets the requirements for liveness certification.

[0022] In one possible implementation, the determining module is further configured to: determine the retrograde light rays after the optical signal passes through the eyeball based on the optical signal corresponding to any image frame and a preset multi-layer media optical model of the eyeball; determine the theoretical imaging point in the image frame after the eye region of the test object reacts to the optical signal based on the retrograde light rays; and determine that the test object has a living eyeball based on the actual imaging point and the theoretical imaging point in the image frame.

[0023] In one possible implementation, the determining module is further configured to: emit a high-frequency micro-vibration sequence from the terminal device; determine the vibration response information of the terminal device through an inertial measurement unit; determine the state of the terminal device based on the vibration response information, including a handheld state and a fixed state; and, when the terminal device is in a handheld state, perform a cross-correlation analysis of human biological characteristics based on eye movement information and device movement information to determine whether the tested object meets the requirements for liveness authentication.

[0024] In one possible implementation, the determining module is further configured to: when the terminal device is in a fixed state: determine the optical flow displacement information of the image acquired by the image acquisition unit due to the vibration of the terminal device based on the image sequence; and determine whether the object under test conforms to the liveness authentication based on the device motion information and the optical flow displacement information.

[0025] In one possible implementation, the transmitting module is specifically used to: transmit optical signals through the display screen, including: generating a pseudo-random spatial encryption matrix based on a one-time random number; the one-time random number becoming invalid after the liveness authentication is completed; and transmitting optical signals conforming to the pseudo-random spatial encryption matrix through a sub-pixel array set in the display screen.

[0026] Thirdly, this application also provides a liveness authentication device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method described in various possible designs of the first aspect.

[0027] Fourthly, this application also provides a computer-readable storage medium storing a computer program or instructions that, when executed by a processor, implement the method described in various possible designs of the first aspect.

[0028] Fifthly, this application also provides a computer program product that, when run on a computer, causes the computer to perform any of the methods described in the first aspect above.

[0029] These or other implementations of this application will become clearer and easier to understand in the following description of the embodiments. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 A flowchart illustrating a liveness authentication method provided in this application embodiment; Figure 2 A schematic diagram illustrating the propagation of an optical signal in a preset multilayer optical model of the eyeball, provided as an embodiment of this application; Figure 3 This is a schematic diagram illustrating how a terminal device determines whether a tested object meets a second liveness characteristic when it is being held in a handheld state, as provided in an embodiment of this application. Figure 4 This is a schematic diagram illustrating how a terminal device determines whether a tested object meets the second liveness characteristic when it is in a fixed state, according to an embodiment of this application. Figure 5 A schematic diagram of a liveness detection device provided in an embodiment of this application; Figure 6This is a schematic diagram of another liveness authentication device provided in an embodiment of this application. Detailed Implementation

[0032] To make the objectives, technical solutions, and beneficial effects of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0033] Existing liveness detection technologies capture facial images through image acquisition units and identify fake individuals using commands such as blinking, head turning, and mouth opening. However, existing technologies determine whether an individual is alive by judging whether their appearance or preset actions match the characteristics of a live person. When black market industries use high refresh rate displays or real-time generated fake videos, they can simulate blinking, head turning, and other actions in the video. In this case, existing technologies may misidentify the individual as a live person.

[0034] Based on this, this application provides a liveness detection method that accurately detects whether the subject is a living person by simultaneously referencing eye movement information and device movement information of the terminal device.

[0035] The following is a detailed description of the proposed solution in conjunction with the accompanying drawings.

[0036] The liveness authentication method described in this application is applied to a terminal device. The terminal device may be, but is not limited to, smartphones, tablets, laptops, smart wearable devices (such as smart glasses and smartwatches), and other electronic devices with display and image acquisition capabilities.

[0037] Please see Figure 1 The diagram illustrates a flowchart of a liveness authentication method applied to a terminal device, which includes the following steps: Step 110: In response to the liveness authentication command, an optical signal is emitted through the display screen.

[0038] In some embodiments, when the business side or risk control intelligent agent triggers the liveness authentication process, the terminal device enters the authentication session and opens a preset authentication time window. Within the authentication time window, the terminal device controls the display screen to emit optical signals for liveness determination to the eye area of ​​the tested object.

[0039] In some embodiments, the display screen includes a plurality of sub-pixel arrays; the terminal device can generate an invisible optical image based on a pseudo-random spatial encryption matrix, and drive the sub-pixel arrays set in the display screen to emit optical signals according to the invisible optical image. For example, the terminal device can generate a pseudo-random spatial encryption matrix based on a one-time random number, which expires after liveness authentication, and then emit optical signals conforming to the pseudo-random spatial encryption matrix through the sub-pixel arrays set in the display screen.

[0040] In some embodiments, the terminal device can drive a short-wavelength blue sub-pixel array to send optical signals at a refresh / modulation frequency of not less than 120Hz according to an invisible optical image, so that the frequency of change of the optical signal is higher than the critical flicker fusion frequency of the human eye, thereby completing the optical challenge without the user's visual perception.

[0041] By generating invisible optical signals through a pseudo-random spatial encryption matrix, the optical challenges used in each authentication are made random and imperceptible. This makes it difficult for attackers to pre-record or accurately predict the corresponding eye optical responses, thereby improving resistance to historical video replay and real-time forgery attacks, while reducing the impact of explicit action commands on the user authentication experience.

[0042] Step 120: Determine the eye movement information of the tested object by acquiring the image sequence and optical signals through the image acquisition unit.

[0043] In some embodiments, after the display screen emits optical signals, the image acquisition unit continuously acquires multiple frames of image data of the eye region at a preset frame rate. Each frame contains optical response information (i.e., image sequence) generated by the multiple layers of the eyeball (e.g., cornea, aqueous humor) in response to the optical signals, such as reflection, refraction, scattering, and occlusion. Furthermore, the eye movement information of the subject is determined based on the image sequence and the optical signals. The image acquisition unit can be a front-facing camera, an infrared camera, or other imaging devices capable of acquiring eye image information.

[0044] In some embodiments, the terminal device generates an optical challenge coding sequence based on a pseudo-random spatial encryption matrix and drives the sub-pixel array of the display screen to emit corresponding optical signals. After the image acquisition unit acquires an image sequence of the eye region, the terminal device can perform inter-frame differencing on adjacent frames to suppress background components, and perform mask matching / demodulation on the differencing results based on the pseudo-random spatial encryption matrix, thereby extracting the invisible response image (i.e., the image sequence) corresponding to the optical challenge. For example, the invisible response image can be extracted in the following manner:

[0045] in, and Let represent the pixel intensity at coordinates (x, y) of the image frame at time t and time t-1, respectively. This represents a pseudo-random spatial encryption matrix, for example. It is a pseudo-random spatial encryption matrix based on a one-time random number, and the one-time random number becomes invalid after the liveness authentication is completed.

[0046] In one possible implementation, the retrograde rays after the optical signal passes through the eyeball are determined based on the optical signal and a preset multi-layered optical model of the eyeball; the theoretical imaging point of the subject's eye region after responding to the optical signal is determined based on the retrograde rays; and the eye movement information of the subject is determined based on the image sequence and the theoretical imaging point.

[0047] In a specific example, the terminal device traces the propagation path of the optical signal in the eye in reverse, based on the emission position and direction of the optical signal emitted by the display screen, as well as the geometric parameters and refractive index parameters of each optical interface in the preset multi-layer media optical model of the eye, to obtain the reverse light rays after the optical signal passes through the cornea, aqueous humor, and other media of the eye. Based on the intersection points and propagation directions of the reverse light rays with each optical interface, the reverse light rays are projected onto the image coordinate system of the image acquisition unit to obtain the theoretical imaging point when the eye region responds to the optical signal.

[0048] Further, actual imaging points corresponding to the optical signals are extracted from each image frame of the image sequence, and the actual imaging points in each image frame are matched with the corresponding theoretical imaging points. The eye pose parameters are adjusted according to the positional deviation between the two until the reprojection residual between the theoretical imaging point and the actual imaging point meets the preset convergence condition, thereby obtaining the eye pose corresponding to each image frame. Based on the eye poses corresponding to adjacent image frames and the time interval between adjacent image frames, the rotation direction, rotation angle, and / or angular velocity of the eye are determined as the eye movement information of the tested object.

[0049] Step 130: Collect device motion information of the terminal device through the inertial measurement unit.

[0050] An inertial measurement unit (IMU) is used to acquire motion information of a terminal device. An IMU can be a sensor assembly used to measure the acceleration and / or angular velocity of the terminal in three-dimensional space, typically including, but not limited to, a three-axis accelerometer and a three-axis gyroscope.

[0051] Step 140: Perform cross-correlation analysis of human biological characteristics based on eye movement information and device movement information to determine whether the tested object meets the requirements for liveness certification.

[0052] This application emits optical signals through a display screen and collects eye movement information and device movement information of the terminal device, so that liveness detection no longer relies solely on the surface image or preset actions of the tested object, but comprehensively utilizes eye movement information (i.e. the test object's response to the optical signal) and physical device movement information of the terminal device, thereby improving the recognition capability of liveness detection.

[0053] The liveness authentication process of this application can be completed on the terminal device side, without requiring the tested object to perform preset actions such as blinking or turning its head, and can reduce the transmission of eye image sequences to the server, thereby improving the efficiency of liveness authentication.

[0054] In some embodiments, step 140 includes: performing cross-correlation analysis on eye movement information and device movement information to obtain phase delay parameters and motion gain parameters; and determining that the tested object meets the requirements for liveness authentication when the phase delay parameters satisfy the central nervous system conduction mechanism of the human body and the motion gain parameters satisfy the vestibular-ocular reflex mechanism.

[0055] Among them, the phase delay parameter is used to characterize the time difference between the hand movement and the eye movement of the test object, and the motion gain parameter is used to characterize the amplitude relationship between the hand movement and the eye movement of the test object.

[0056] For example, the formula for cross-correlation analysis of eye movement information and device motion information is as follows:

[0057] in, Indicates equipment movement information. Representing eye movement information, based on the cross-correlation analysis of the above eye movement information and device motion information, phase delay parameters and motion gain parameters can be obtained.

[0058] In some embodiments, eye motion information and device motion information are obtained by acquiring image sequences within a preset authentication time window; cross-correlation analysis is performed on eye motion information and device motion information within a sliding time window to determine the frame offset that satisfies the largest cross-correlation value, and the phase delay parameter is determined based on the frame offset and the sampling frame interval; For example, the formula for calculating the phase delay parameter is as follows:

[0059] in, This represents the cross-correlation value at a frame offset of m, where N represents the number of sampling points within the sliding time window. The optimal frame offset is determined by iterating through the preset offset range to maximize the cross-correlation value. Determine the phase delay parameter based on the optimal frame offset. ,in, This indicates the sampling interval after resampling.

[0060] In some embodiments, motion vectors are calculated for the eye motion information and the corresponding device motion information at the time point indicated by the maximum cross-correlation value to obtain motion gain parameters.

[0061] For example, the formula for calculating the motion gain parameter is as follows:

[0062] In some embodiments, if the device motion information does not meet the certification criteria, the head information of the tested object is determined by the image sequence, and the cross-correlation analysis of the vestibular-ocular reflex mechanism is performed based on the eye motion information and head motion information to determine whether the tested object meets the liveness certification criteria.

[0063] For example, the failure of equipment motion information to meet the criteria may include: the amplitude of the equipment motion information is less than a preset motion amplitude threshold, the signal-to-noise ratio of the equipment motion information is less than a preset signal-to-noise ratio threshold, or the equipment motion information is affected by environmental vibration interference and cannot extract motion components that meet the analysis requirements.

[0064] Specifically, the terminal device can detect facial key points of the subject from the image sequence, and estimate the head pose corresponding to each image frame based on the positional changes of facial key points in adjacent image frames and the imaging parameters of the image acquisition unit. For example, the head pose may include rotation and translation parameters of the head relative to the image acquisition unit. The head angular velocity vector is determined based on the head rotation parameters corresponding to adjacent image frames and the time interval between adjacent image frames. This serves as the head motion information. Eye motion information may include an eye angular velocity vector determined based on eye rotation parameters corresponding to adjacent image frames. Before performing cross-correlation analysis, the head angular velocity vector and the eye angular velocity vector can be transformed to the same coordinate system, and resampling and time alignment can be performed according to a unified sampling time point.

[0065] When the head angular motion within the preset authentication time window meets the preset motion conditions, and the sampling frame rate, exposure time, time synchronization error, and estimation accuracy of head pose and eye pose of the image acquisition unit can resolve the response delay to be detected, a normalized cross-correlation analysis is performed on the eye angular velocity vector and the head angular velocity vector:

[0066] in, This represents the normalized cross-correlation value between the eye angular velocity vector and the head angular velocity vector when the frame offset is m; N represents the number of sampling points involved in the calculation within the preset authentication time window; ε represents a preset positive number used to avoid a denominator of zero. The optimal frame offset is determined by iterating through the preset frame offset range to maximize the normalized cross-correlation value.

[0067] in, This represents the optimal frame offset. This indicates the sampling frame interval after resampling. This represents the response delay between head movement and eye movement. Within the alignment window corresponding to the optimal frame offset, the eye-head angular velocity gain is calculated:

[0068] in, This indicates the amplitude relationship between eye-compensation movements and head movements. The terminal device can determine this based on normalized cross-correlation values ​​and response latency. and eye-head angular velocity gain At least two of the criteria are used to determine whether eye movements and head movements satisfy the vestibulo-ocular reflex mechanism. For example, if the maximum normalized cross-correlation value is greater than a preset correlation threshold, the response delay is within a preset candidate delay range, and the eye-head angular velocity gain is within a preset gain range, then the eye movements and head movements are determined to satisfy the vestibulo-ocular reflex mechanism, and the tested subject is determined to be a liveness detector; otherwise, the eye-head movement analysis is determined to have failed or not met the valid criteria.

[0069] The preset candidate delay range, preset gain range, and preset relevant thresholds can be calibrated based on the sampling capability of the terminal device, the exposure characteristics of the image acquisition unit, the time synchronization error, the pose estimation error, and the sample data of the target population, and are not limited to a single fixed value. In embodiments with high sampling rates and satisfactory time synchronization accuracy, the response delay can be searched within a range of several milliseconds to tens of milliseconds. When the sampling period, exposure time, or time synchronization error is insufficient to support the corresponding time resolution, a rejection conclusion that the tested object does not meet the liveness authentication requirement is not made based on the response delay. Instead, the image sequence can be re-acquired, or the process can be switched to another liveness authentication branch.

[0070] In some embodiments, when the head angular movement within a preset authentication time window does not meet preset movement conditions, or when eye movements are primarily induced by optical signals emitted from the display screen, a visually evoked eye-tracking model can be used for liveness authentication. Specifically, the spatial position change, movement direction, phase change, and / or brightness change of the optical signals emitted from the display screen are used as inputs, and the eyeball angular velocity vector and / or gaze point trajectory are used as outputs. A time-series correlation analysis is performed on the inputs and outputs to obtain visual response delay, response gain, and / or trajectory matching degree. Based on the visual response delay range, response gain range, and / or trajectory matching threshold calibrated by the target population sample, it is determined whether the tested subject produces a corresponding eye movement response to the optical signal. This visually evoked eye-tracking analysis is used to evaluate the correspondence between optical challenge and eye movement response; it is not used as a vestibular reflex analysis, and the response delay range and gain range corresponding to the vestibular reflex are not adopted.

[0071] In some embodiments, this application employs a two-layer liveness determination mechanism: based on optical signals and image sequences, it determines whether the tested object meets the first liveness characteristic; based on device motion information and image sequences, it determines whether the tested object meets the second liveness characteristic; if the tested object meets both the first and second liveness characteristics, the tested object is determined to be a live object.

[0072] The first liveness feature is used to characterize whether the eye region of the tested object possesses the multi-layered media optical properties of a real eye. Specifically, the terminal device can determine the retrograde ray of the optical signal after passing through the eye and the corresponding theoretical imaging point based on the optical signal emitted by the display screen, the preset multi-layered media optical model of the eye, and the actual imaging points detected in the image sequence, and calculate the reprojection residual between the theoretical imaging point and the actual imaging point. When the reprojection residual can converge to within a preset residual threshold, it is determined that the eye region of the tested object conforms to the multi-layered media optical path constraints of a real eye, that is, the tested object satisfies the first liveness feature; when the reprojection residual cannot converge to within the preset residual threshold, it is determined that the tested object does not satisfy the first liveness feature.

[0073] In some embodiments, the eye rotation direction, rotation angle, and / or eye angular velocity can be determined based on the eye poses corresponding to multiple image frames in the image sequence, and the determination result of the first liveness feature can be further verified based on the continuity of eye movement, the range of movement, and the response relationship between the eye movement and the optical signal. It should be noted that eye angular velocity can be used as an auxiliary determination parameter for the first liveness feature, but the first liveness feature is not determined solely based on eye angular velocity.

[0074] The second liveness feature is used to characterize whether the motion response reflected in the image sequence matches the actual physical motion of the terminal device. The determination of the second liveness feature can be based on the state of the terminal device, which includes a handheld state and a fixed state.

[0075] When the terminal device is in a handheld state, it can perform cross-correlation analysis of human biological characteristics based on eye movement information and device movement information, as described in step 140 above, to obtain phase delay parameters and motion gain parameters. When the phase delay parameter meets preset timing conditions, the motion gain parameter meets preset inverse compensation relationships, and the cross-correlation value between the two meets preset correlation conditions, it is determined that the eye movement and the actual movement of the terminal device conform to preset human biological characteristics, thereby determining that the tested object meets the second liveness characteristic.

[0076] If the device motion information does not meet the criteria, such as insufficient amplitude, low signal-to-noise ratio, or interference from environmental vibrations, the head motion information of the tested object can be determined through image sequences, and a cross-correlation analysis of the vestibulo-ocular reflex mechanism can be performed on the eye motion information and head motion information. Based on the motion direction, response delay, motion gain, and cross-correlation degree between eye motion and head motion, it can be determined whether the tested object meets the second liveness characteristic.

[0077] When the terminal device is in a fixed state, it can emit a high-frequency micro-vibration sequence. An inertial measurement unit (IMU) collects the device motion information corresponding to the high-frequency micro-vibration sequence, and the optical flow displacement information caused by the high-frequency micro-vibration sequence in the image acquisition frame is determined through an image sequence. The terminal device transforms the device motion information to the image coordinate system and compares the transformed device motion information with the optical flow displacement information. When at least one of the two satisfies a preset matching condition in terms of motion direction, motion amplitude, and motion timing, it is determined that the image acquisition frame has produced a corresponding response to the actual physical motion of the terminal device, thereby determining that the measured object meets the second liveness characteristic.

[0078] By verifying the optical properties of the multi-layer media in the eye region and the correspondence between the image response and the actual physical motion of the terminal device, the liveness authentication of the tested object can be performed from two aspects: the authenticity of the optical media and the consistency of cross-modal motion. This reduces the misjudgment caused by authentication based solely on surface images or preset actions, and improves the ability to identify attacks such as two-dimensional image replay, three-dimensional forged media, and video stream injection.

[0079] This application does not limit the combination of the first liveness feature and the second liveness feature. In some embodiments, both the first liveness feature and the second liveness feature can be used as necessary determination conditions, and the tested object is determined to meet the liveness authentication when both are satisfied. In other embodiments, the first liveness feature or only the second liveness feature can be determined based on the security level corresponding to the authentication scenario, the hardware capabilities of the terminal device, and the quality of the collected data. In still other embodiments, corresponding confidence levels or weights can be set for the first liveness feature and the second liveness feature respectively, and a comprehensive liveness confidence level can be determined based on the weighted result of the two; when the comprehensive liveness confidence level reaches a preset confidence threshold, the tested object is determined to meet the liveness authentication.

[0080] When the first or second liveness feature fails to meet the valid judgment criteria due to reasons such as sampling frame rate, exposure time, time synchronization error, motion amplitude, or environmental interference, the authentication data can be re-acquired, the parameters of the optical signal or high-frequency micro-vibration sequence can be adjusted, or the process can be switched to other liveness authentication branches, instead of directly concluding that the tested object does not meet the liveness authentication criteria based on invalid data.

[0081] The following describes the first and second liveness detection methods.

[0082] In some embodiments, the determination of whether the subject meets the first living characteristic is based on the following scheme: determining the retrograde light rays after the optical signal passes through the eyeball based on the optical signal corresponding to any image frame and a preset multilayer media optical model of the eyeball; determining the theoretical imaging point in the image frame after the eye region of the subject responds to the optical signal based on the retrograde light rays; and determining whether the subject meets the first living characteristic (i.e., determining whether the subject has a living eyeball) based on the actual imaging point and the theoretical imaging point in the image frame.

[0083] In some embodiments, if the first layer determines that the tested object does not meet the liveness characteristics, then the liveness authentication is terminated and an indication message indicating that the tested object does not meet the liveness characteristics is output. That is, after determining that the tested object does not have a live eyeball, the subsequent liveness authentication steps are terminated.

[0084] In a specific example, the preset multi-layer media optical model of the eyeball includes the geometric parameters and refractive index parameters of the cornea and aqueous humor, wherein the refractive index of the cornea can be n1≈1.376 and the refractive index of the aqueous humor can be n2≈1.336.

[0085] For example, the direction of the retrograde rays after the optical signal passes through the eyeball is determined according to the three-dimensional vector Snell's law:

[0086] in, Indicates the direction of incidence. This represents the normal vector of the medium interface. Represents the refractive index of the incident medium. Indicates the refractive index of the exiting medium. Indicates the direction of refraction.

[0087] Please see Figure 2 This illustrates a schematic diagram of an optical signal propagating in a pre-defined multilayered optical model of the eye. For example... Figure 2 As shown, the light emitted from the display screen (i.e., the screen light source) is incident on the corneal surface through the air medium and passes through the cornea and aqueous humor in sequence; the image acquisition unit collects the optical response formed after refraction and reflection through the multiple media of the eyeball in the corresponding direction (i.e., camera observation). The refraction direction between different media can be determined according to the refractive index of each medium and the normal vector of the medium interface.

[0088] Furthermore, based on the direction of the reverse light rays, the theoretical imaging point in the image acquisition unit after the subject's eye region responds to the optical signal is determined:

[0089] Where j represents the optical interface number in the multilayer media optical model of the eyeball; This represents the intersection point of the backward rays at the j-th optical interface; This represents the intersection point of the retro-light rays on the (j+1)th optical interface; Indicates that the backlight comes from spread to Distance parameters; This represents the propagation direction vector of the retrograde ray after it leaves the j-th optical interface; Let represent the implicit surface equation of the (j+1)th optical interface. Indicates the intersection point Located on the (j+1)th optical interface; This indicates the final exit point of the retrograde light rays after passing through multiple layers of the eyeball; and These represent the rotation matrix and translation vector from the eye coordinate system to the image acquisition unit coordinate system, respectively; K represents the intrinsic parameter matrix of the image acquisition unit. This represents the theoretical imaging point calculated based on a preset multi-layered optical model of the eye.

[0090] Based on the theoretical imaging points and the actual imaging points detected in the image sequence, establish the reprojection residual function:

[0091] in, This represents the i-th theoretical imaging point. Let R represent the i-th actual imaging point, R represent the eye rotation parameter, T represent the eye translation parameter, and M represent the number of imaging points involved in the calculation.

[0092] In some embodiments, the Levenberg-Marquardt (LM) algorithm is used to adjust the eye rotation and translation parameters to make the reprojection residual E converge. If the reprojection residual E fails to converge to within a preset residual threshold, the tested object is determined not to meet the first liveness characteristic (i.e., it does not have a live eye); if the reprojection residual E converges to within the preset residual threshold, the tested object is determined to meet the first liveness characteristic. Thus, the multi-layered optical path constraint of a real eye can be used to identify two-dimensional display images or three-dimensional forged media that do not conform to a preset multi-layered optical model of the eye.

[0093] In some embodiments, the terminal device exists in at least two usage states: a handheld state and a fixed state. To distinguish between the two states, the terminal device can make a determination based on the spectral characteristics of the device's motion information; for example, the terminal device can actively emit a high-frequency micro-vibration sequence, such as calling the terminal device's X-axis and / or Z-axis linear motors to emit a pseudo-randomly encoded asymmetric high-frequency micro-vibration sequence (e.g., a frequency sweeping shock wave with a duration of 50ms), and the inertial measurement unit can collect the vibration response information of the terminal device to determine the state of the terminal device accordingly.

[0094] For example, the physical ring-down curve of the shock wave can be recorded by the inertial measurement unit: if the terminal is rigidly fixed by a robotic arm or metal support, it may exhibit high-frequency rigid resonance characteristics; if the terminal is held by a real human hand (even if it is deliberately kept still), the viscoelastic biodamping of the human soft tissue will absorb the shock wave energy, making the attenuation process smoother, thereby achieving the distinction of the physical bearing environment.

[0095] In some embodiments, the above-described attenuation process can be characterized by the attenuation envelope equation of a damped harmonic oscillator, for example:

[0096] Where x(t) represents the vibration displacement, A represents the initial amplitude, and ζ represents the damping ratio. Indicates the natural angular frequency. Indicates the angular frequency of damped vibration. This indicates the phase. Terminal devices can obtain parameters such as the damping ratio by fitting the acquired attenuation curves. It also determines whether the terminal device is in a handheld or fixed state based on threshold conditions.

[0097] For example, when a terminal device is fixed by a rigid bracket, the physical attenuation curve can exhibit rigid resonance characteristics; when the terminal device is held in the hand, the viscoelastic damping effect of the human soft tissue attenuates the vibration energy, which can be used to distinguish between the held state and the fixed state.

[0098] By actively emitting a high-frequency micro-vibration sequence from the terminal device, and determining whether the terminal device is in a handheld or fixed state based on the device motion information and vibration response information collected by the inertial measurement unit, an appropriate liveness detection method can be selected according to the load status of the terminal device, thus taking into account both handheld and fixed scenarios and reducing misjudgments caused by scenario differences.

[0099] In some embodiments, when the terminal device is in a handheld state, a cross-correlation analysis of human biological characteristics is performed based on eye movement information and device movement information to determine whether the tested object meets the requirements for liveness authentication.

[0100] In some embodiments, when the terminal device is in a handheld state, determining whether the tested object meets the second liveness characteristic based on motion information includes: determining the eye posture characteristics of the tested object based on optical signals and image sequences; performing cross-correlation analysis on motion information and eye posture characteristics to obtain phase delay parameters and motion gain parameters; and determining that the tested object meets the second liveness characteristic when the phase delay parameter meets a first preset condition and the motion gain parameter meets a second preset condition.

[0101] In some embodiments, after the reprojection residual E converges to a preset residual threshold and it is determined that the tested object meets the first liveness characteristic, the eye rotation parameter R (one of the eye posture characteristics) is determined based on the converged reprojection residual E, and the eye angular velocity vector is determined based on the eye rotation parameter R. Then, cross-correlation analysis is performed on the device motion information and the eye angular velocity vector to determine whether the tested object meets the second liveness characteristic.

[0102] In some embodiments, the formula for determining the eye angular velocity vector based on the eye rotation parameter R is as follows:

[0103] Where ΔR_k represents the relative rotation matrix between adjacent image frames, Δt_k represents the time interval between adjacent image frames, and vee represents the operation of converting the antisymmetric matrix into a vector. The translation parameter T is used for eye center localization and translation compensation.

[0104] In some embodiments, when the terminal device is in a handheld state, the inertial measurement unit (IMU) collects the device motion information of the terminal device as follows: The IMU can collect the raw device motion information of the terminal device at a sampling rate of not less than 500Hz, and extract the physiological micro-tremor components in the 8Hz to 12Hz frequency band through Fast Fourier Transform. In some cases, when there is environmental turbulence, the low-frequency components in the 1Hz to 3Hz frequency band can be filtered out. The filtered device motion information can be expressed as:

[0105] Where F represents the Fast Fourier Transform, This represents the inverse fast Fourier transform. Indicates the original equipment motion information. This represents the transfer function of the preset bandpass filter. This represents the filtered device motion information.

[0106] To reduce the clock skew between the image acquisition unit and the inertial measurement unit, the motion information of the equipment... and eyeball angular velocity vector Resampling is performed, and cross-correlation analysis is conducted within the sliding time window:

[0107] in, This represents the cross-correlation value at a frame offset of m, where N represents the number of sampling points within the sliding time window. The optimal frame offset is determined by iterating through the preset offset range to maximize the cross-correlation value.

[0108] Determine the phase delay parameter based on the optimal frame offset. .

[0109] Determine motion gain parameters based on device motion information and eye angular velocity vector:

[0110] For example, the first preset condition can be that the phase delay parameter Δt is within a preset physiological range of 10ms to 15ms. The phase delay parameter Δt is used to characterize the time interval between the movement of the terminal device and the compensating movement of the eye. When the phase delay parameter Δt meets the first preset condition, it indicates that the temporal relationship between the movement of the terminal device and the compensating movement of the eye conforms to the preset liveness timing characteristics; when the phase delay parameter Δt does not meet the first preset condition, it indicates that the temporal relationship between the two is abnormal, and it can be determined that the tested object does not meet the second liveness characteristics. For example, when the phase delay parameter is close to 0ms, the movement of the terminal device and the eye movement in the image may be mechanically synchronized movements lacking the corresponding physiological time delay; when the phase delay parameter is greater than the preset upper limit, the eye movement in the image may have additional delays introduced by video generation, processing, or transmission. By simultaneously verifying the motion gain parameter and the phase delay parameter, it is possible to determine whether the eye movement corresponds to the actual movement of the terminal device based on the motion amplitude relationship and the motion timing relationship, thereby improving the ability to identify video playback, mechanical devices driving fake media movement, and generative video attacks.

[0111] For example, the second preset condition can be that the motion gain parameter k satisfies 0.85 ≤ k ≤ 1.15. When the terminal device is held in hand, even slight movements of the terminal device can cause the subject's eyeballs to produce compensating movements in the opposite direction. The motion gain parameter k is used to characterize the amplitude correspondence between the eyeball's compensating movement and the terminal device's movement. When the motion gain parameter k satisfies the second preset condition, it indicates that the eyeball's posture characteristics and the terminal device's motion information conform to the preset reverse compensation relationship in terms of movement direction and amplitude; when the motion gain parameter k does not satisfy the second preset condition, it indicates that there is no corresponding motion correlation between the two, and it can be determined that the subject does not meet the second liveness characteristic. Thus, it is possible to identify video playback situations where the movement in the eye image does not match the actual movement of the terminal device.

[0112] Please see Figure 3 This illustrates a schematic diagram of determining whether a tested object meets the second liveness characteristic when the terminal device is being held in a handheld state. Figure 3 As shown, the terminal device emits optical signals (e.g., 120Hz high-frequency implicit pseudo-random micro-stroboscopic) to the eye region of the test subject (the multi-layered medium of the user's eyeball) through a display screen (the terminal screen and its display driver). The image acquisition unit (high frame rate front camera) acquires the image sequence (composite image stream) of the eye region's response to the optical signals, and the reverse ray tracing engine determines the eyeball angular velocity vector based on the image sequence. (t). Simultaneously, the inertial measurement unit (terminal IMU sensor) acquires device motion information generated when the terminal device is held in hand, and extracts the motion vector corresponding to the physiological micro-tremors of the hand from the device motion information. (Spatial transient vector). The terminal device analyzes the eyeball angular velocity vector through the cross-modal spatiotemporal collaborative analysis module (cross-modal spatiotemporal collaboration and VOR neurobiological computation engine). (t) and motion vector The analysis yields the phase delay parameter Δt and motion gain parameter k. Based on whether the phase delay parameter Δt and motion gain parameter k meet the corresponding preset conditions, it is determined whether the tested object meets the second liveness characteristic (outputting the physical liveness determination result).

[0113] When a terminal device is held in hand, slight hand tremors cause spatial movement, and the eyes of a real living person will produce a compensating rotation in the opposite direction and with a matching amplitude. By performing cross-correlation analysis on the device's motion information and the eye's angular velocity vector, and verifying the inverse compensation relationship between the two using motion gain parameters, it is possible to determine whether the eye image motion is bound to the actual physical motion of the terminal device, thus avoiding misidentifying two-dimensional forged videos as live objects. Simultaneously, phase delay parameters are used to verify whether the neural conduction time from motion perception to the eye's compensating rotation is within a preset physiological range, distinguishing between mechanically synchronized movements lacking neural conduction delay and forged videos with significant delays. Therefore, by simultaneously verifying phase delay and motion gain parameters, the ability to identify two-dimensional video playback, robotic arms holding forged media, and real-time generative video attacks can be improved.

[0114] In some embodiments, when the terminal device is in a fixed state, determining whether the object under test meets the second liveness characteristic based on the device motion information includes: determining the optical flow displacement information of the image acquired by the image acquisition unit due to the vibration of the terminal device based on the image sequence; and determining whether the object under test meets the second liveness characteristic based on the device motion information and the optical flow displacement information.

[0115] Specifically, after the terminal device emits a high-frequency micro-vibration sequence, the image acquisition unit continuously acquires images within the time domain corresponding to the high-frequency micro-vibration sequence. The terminal device can determine the optical flow displacement information (also known as the global optical flow displacement vector) of the image based on the changes in pixel positions in adjacent images. Based on a preset coordinate projection transformation relationship, the terminal device converts the device motion information acquired by the inertial measurement unit to the image coordinate system under the image acquisition unit, and compares the converted device motion information with the optical flow displacement information to determine the motion deviation between the two.

[0116] For example, the motion deviation can be expressed as the following error. express:

[0117] in, Represents the global optical flow displacement vector. This represents the displacement vector of the terminal equipment determined based on the equipment motion information collected by the inertial measurement unit. This represents the projection transformation matrix from the coordinate system of the inertial measurement unit to the coordinate system of the image acquisition unit. This indicates the preset error threshold.

[0118] exist ≤ In this case, the device motion information is determined to match the optical flow displacement information, and the measured object is determined to meet the second living characteristic; > In such cases, it is determined that the device motion information and optical flow displacement information do not match, and that the object under test does not meet the second liveness characteristic. By comparing the actual motion of the terminal device with the predicted motion caused by that motion in the image acquisition frame, it is possible to determine whether the frame acquired by the image acquisition unit produces a corresponding motion response to the current high-frequency micro-vibration sequence, thereby identifying video stream injection attacks that inject pre-recorded video or replace image acquisition data.

[0119] In some embodiments, eye pose features can be replaced with deformation information of facial skin or soft tissue under inertial action, and a liveness determination can be made based on the damping and time delay relationship between the deformation information and the device motion information of the terminal device. For terminal devices with limited computing resources, optical flow information of the eye reflection area can also be extracted, and time-locked analysis can be performed on the optical flow information, the optical signal encoding sequence of the display screen, and the device motion information collected by the inertial measurement unit to determine whether the object under test is a live body.

[0120] Please see Figure 4 This illustrates a schematic diagram of determining whether a tested object meets the second liveness characteristic when the terminal device is in a fixed state. For example... Figure 4As shown, in response to the liveness authentication command (the command issued by the risk control agent), the terminal device controls the vibration unit (terminal linear motor) to emit a high-frequency micro-vibration sequence (asymmetric sweep frequency shock wave, i.e., chirp pulse). The high-frequency micro-vibration sequence is physically transmitted through a physical bearing medium (e.g., a fixed desktop or stand). The inertial measurement unit (terminal IMU sensor) collects the device motion information (vibration attenuation response curve) corresponding to the high-frequency micro-vibration sequence, and can determine the bio-damping coefficient ζ based on the vibration attenuation response curve. At the same time, the image acquisition unit (high frame rate front-facing camera) acquires the image of the tested object and its background (front-facing camera image), and uses an optical flow calculation method (optical flow method) to determine the predicted device motion information (image relative displacement vector) caused by the terminal device vibration in the image. The terminal device compares the device motion information with the predicted device motion information (isomorphism test), and determines whether the tested object meets the second liveness characteristic based on whether the two match; when the two do not match, it can be determined that the tested object does not meet the second liveness characteristic (interception of rigid resonance or bypass video injection).

[0121] Based on the same concept, this application also provides a liveness authentication device that can perform the liveness authentication method described above.

[0122] Please see Figure 5 This application provides a schematic diagram of the structure of a liveness authentication device according to an embodiment of the present application, as shown below. Figure 5 As shown, the liveness detection device includes: Transmitter module 501 is used to transmit optical signals through the display screen in response to a liveness authentication command; The acquisition module 502 is used to determine the eye movement information of the object under test by acquiring image sequences and optical signals through the image acquisition unit; and to acquire the device motion information of the terminal device through the inertial measurement unit. The determination module 503 is used to perform cross-correlation analysis of human biological characteristics based on eye movement information and device movement information to determine whether the tested object meets the requirements for liveness certification.

[0123] In one possible implementation, the determination module 503 is specifically used to: perform cross-correlation analysis on eye movement information and device movement information to obtain phase delay parameters and motion gain parameters; the phase delay parameters are used to characterize the time difference between hand movement and eye movement of the test subject; the motion gain parameters are used to characterize the amplitude relationship between hand movement and eye movement of the test subject; when the phase delay parameters satisfy the central nervous system conduction mechanism of the human body and the motion gain parameters satisfy the vestibular-ocular reflex mechanism, the test subject is determined to meet the requirements for liveness authentication.

[0124] In one possible implementation, the eye motion information and device motion information are obtained from image sequences acquired within a preset authentication time window; the determination module 503 is used to: perform cross-correlation analysis on the eye motion information and device motion information within a sliding time window to determine the frame offset when the cross-correlation value is maximized; determine the phase delay parameter based on the frame offset and the sampling frame interval; and calculate the motion vector for the eye motion information and the corresponding device motion information at the time point indicated by the maximum cross-correlation value to obtain the motion gain parameter.

[0125] In one possible implementation, the determining module 503 is further configured to: determine the head motion information of the tested object through the image sequence if the device motion information does not meet the certification conditions; and perform cross-correlation analysis of the vestibular-ocular reflex mechanism based on the eye motion information and head motion information to determine whether the tested object meets the liveness certification criteria.

[0126] In one possible implementation, the determining module 503 is further configured to: determine the retrograde light rays after the optical signal passes through the eyeball based on the optical signal corresponding to any image frame and a preset multi-layer media optical model of the eyeball; determine the theoretical imaging point in the image frame after the eye region of the test object reacts to the optical signal based on the retrograde light rays; and determine that the test object has a living eyeball based on the actual imaging point and the theoretical imaging point in the image frame.

[0127] In one possible implementation, the determining module 503 is further configured to: emit a high-frequency micro-vibration sequence from the terminal device; determine the vibration response information of the terminal device through an inertial measurement unit; determine the state of the terminal device based on the vibration response information, including a handheld state and a fixed state; and, when the terminal device is in a handheld state, perform a cross-correlation analysis of human biological characteristics based on eye movement information and device movement information to determine whether the tested object meets the requirements for liveness authentication.

[0128] In one possible implementation, the determining module 503 is further configured to: when the terminal device is in a fixed state: determine the optical flow displacement information of the image acquired by the image acquisition unit due to the vibration of the terminal device based on the image sequence; and determine whether the object under test conforms to the liveness authentication based on the device motion information and the optical flow displacement information.

[0129] In one possible implementation, the transmitting module 501 is specifically used to: transmit optical signals through the display screen, including: generating a pseudo-random spatial encryption matrix based on a one-time random number; the one-time random number becomes invalid after the liveness authentication is completed; and transmitting optical signals conforming to the pseudo-random spatial encryption matrix through a sub-pixel array set in the display screen.

[0130] Please see Figure 6 This shows a schematic diagram of another liveness authentication device provided in an embodiment of this application, such as... Figure 6 As shown, the liveness authentication device includes a memory 601 and a processor 602, with the processor 602 coupled to the memory 601. The memory 601 stores program instructions, and the processor 602 calls the program instructions stored in the memory 601 to execute the aforementioned liveness authentication method according to the obtained program.

[0131] Optionally, the liveness authentication device may further include an interface circuit 603, which may be a transceiver or an input / output interface. The input / output interface is used for inputting and / or outputting information; output can be understood as sending, and input as receiving. The processor 602 can communicate with other devices in the liveness authentication device or other devices besides the liveness authentication device through the interface circuit 603 to obtain the information required to perform the above-described liveness authentication method.

[0132] When the liveness authentication device 600 is used to achieve Figure 1 In the method shown, the processor 602 is used to implement the functions of the above-mentioned transmission module 501, acquisition module 502 and determination module 503.

[0133] It is understood that the processor in the embodiments of this application may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor may be a microprocessor or any conventional processor.

[0134] The memory in the embodiments of this application may be random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disk, portable hard disk, compact disc read-only memory (CD-ROM), or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. The storage medium may also be a component of the processor.

[0135] Based on the same technical concept, embodiments of this application also provide a computer-readable storage medium storing a computer program or instructions, which, when executed by a processor, causes the computer to perform the above-described liveness authentication method.

[0136] Based on the same technical concept, embodiments of this application also provide a computer-readable program product that, when executed, causes a computer to perform the above-described liveness authentication method.

[0137] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0138] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0139] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0140] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0141] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A liveness authentication method, characterized in that, Applied to a terminal device, the terminal device including a display screen, an inertial measurement unit, and an image acquisition unit, the method includes: In response to a liveness authentication command, optical signals are emitted through the display screen; The eye movement information of the tested object is determined by the image sequence acquired by the image acquisition unit and the optical signal. The inertial measurement unit collects the device motion information of the terminal device; Based on the eye movement information and the device movement information, a cross-correlation analysis of human biological characteristics is performed to determine whether the tested object meets the requirements for liveness certification.

2. The method according to claim 1, characterized in that, The step of performing cross-correlation analysis of human biological characteristics based on the eye movement information and the device movement information to determine whether the tested object meets the requirements for liveness certification includes: Cross-correlation analysis is performed on the eye movement information and the device movement information to obtain a phase delay parameter and a motion gain parameter; the phase delay parameter is used to characterize the time difference between the hand movement and the eye movement of the tested object; the motion gain parameter is used to characterize the amplitude relationship between the hand movement and the eye movement of the tested object. When the phase delay parameter satisfies the central nervous system conduction mechanism of the human body and the motion gain parameter satisfies the vestibular-ocular reflex mechanism, the tested object is determined to be a living organism.

3. The method according to claim 2, characterized in that, The eye movement information and the device movement information are obtained from image sequences collected within a preset authentication time window; The cross-correlation analysis of the eye movement information and the device movement information to obtain phase delay parameters and motion gain parameters includes: The eye motion information and the device motion information are cross-correlation analyzed within a sliding time window to determine the frame offset when the cross-correlation value is maximized. The phase delay parameter is determined based on the frame offset and the sampling frame interval; Motion vectors are calculated for the eye motion information and the corresponding device motion information at the time point indicated by the maximum cross-correlation value to obtain motion gain parameters.

4. The method according to claim 1, characterized in that, Also includes: If the device motion information does not meet the recognition criteria, the head motion information of the tested object is determined through the image sequence; Based on the eye movement information and the head movement information, a cross-correlation analysis of the vestibular-ocular reflex mechanism is performed to determine whether the tested object meets the requirements for liveness detection.

5. The method according to any one of claims 1 to 4, characterized in that, Before performing the cross-correlation analysis of human biological characteristics based on the eye movement information and the device movement information, the method further includes: Based on the optical signal corresponding to any image frame and a preset multi-layer media optical model of the eye, the retro-ray of the optical signal after passing through the eye is determined. Based on the backlight, the theoretical imaging point in the image frame after the eye region of the tested object reacts to the optical signal is determined; Based on the actual imaging points and the theoretical imaging points in the image frame, it is determined that the tested object has a living eyeball.

6. The method according to any one of claims 1 to 4, characterized in that, Before emitting optical signals through the display screen, the following is also included: The terminal device emits a high-frequency micro-vibration sequence; The vibration response information of the terminal device is determined by the inertial measurement unit, and the state of the terminal device is determined based on the vibration response information. The state includes a handheld state and a fixed state. The step of performing cross-correlation analysis of human biological characteristics based on the eye movement information and the device movement information to determine whether the tested object meets the requirements for liveness certification includes: When the terminal device is in a handheld state, a cross-correlation analysis of human biological characteristics is performed based on the eye movement information and the device movement information to determine whether the tested object meets the requirements for liveness authentication.

7. The method according to claim 6, characterized in that, The method further includes: The terminal device is in a fixed state: Based on the image sequence, determine the optical flow displacement information of the image captured by the image acquisition unit caused by the vibration of the terminal device; Based on the device motion information and the optical flow displacement information, it is determined whether the tested object meets the requirements for liveness authentication.

8. The method according to any one of claims 1 to 4, characterized in that, The emission of optical signals through the display screen includes: A pseudo-random spatial encryption matrix is ​​generated based on a one-time random number; the one-time random number becomes invalid after the liveness authentication is completed. Optical signals conforming to the pseudo-random spatial encryption matrix are emitted through the sub-pixel array set in the display screen.

9. A computer-readable storage medium, characterized in that, It stores a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of any of the methods described in claims 1 to 8.

10. A computer program product, characterized in that, When it is run on a computer, it causes the computer to perform the method as described in any one of claims 1 to 8.

Images