A biometric method and system

Abstract

The biometric method and system provided in this specification include a liveness detection model, a comparison model, and a liveness verification model. The liveness detection model is executed to obtain liveness features. The comparison model can use the liveness features of the liveness detection model for feature comparison. The liveness verification model can use the comparison features of the comparison model for liveness verification. In this way, liveness judgment and feature comparison are alternately performed through the three models. At the same time, the liveness judgment model and the comparison model can mutually use the feature information of each other to improve their respective performances, so as to achieve more accurate biometric recognition.

Description

Technical Field

[0001] This specification relates to the field of image processing, and particularly to a biometric identification method and system. Background Art

[0002] Biometric identification technology mainly refers to a technology for authenticating identities through human biometric features, such as fingerprints, faces, irises, etc. Taking face recognition as an example, with the breakthrough development of face recognition technology in recent years, face recognition systems such as face payment, face access control, and face attendance are becoming more and more common.

[0003] Liveness detection and biometric comparison are the cores of biometric identification technology. Liveness detection refers to determining whether the object to be identified is a live body or an attack, and biometric comparison refers to determining the identity of the object to be identified by comparing the object to be identified with a reference object. However, the method of biometric comparison is prone to false comparison, such as identifying the object to be identified A as the object to be identified B. The liveness detection method often has limited ability to intercept attacks, resulting in a low accuracy rate of existing biometric identification technology. Therefore, there is an urgent need for a biometric identification method with high accuracy. Summary of the Invention

[0004] The biometric identification method and system provided in this specification can achieve biometric identification with higher accuracy.

[0005] In a first aspect, this specification provides a biometric identification method, including: executing a liveness detection model to perform liveness detection on a biometric image corresponding to a target part of a target object and obtaining a liveness detection result, where the liveness detection result includes liveness features; and based on the liveness features, selecting and executing one from multiple sets of solutions to complete biometric identification of the target object and output an identification result, where the multiple sets of solutions include Solution 1, and Solution 1 includes: based on a comparison model, performing feature comparison on the biometric image and obtaining comparison features, where the comparison model uses the liveness features for the feature comparison, and based on the comparison features, performing liveness verification on the biometric image through a liveness verification model to obtain a first identification result of the target part.

[0006] In some embodiments, the selecting and executing one from multiple sets of solutions based on the liveness features includes: based on the liveness features, determining a first attack probability of the target part through the liveness detection model, where the liveness detection result further includes the first attack probability; and determining that the first attack probability is less than a detection threshold, and selecting and executing Solution 1.

[0007] In some embodiments, the living body feature includes at least one living body feature. Based on the living body feature, determining the first attack probability through the living body detection model includes: determining at least one first sub - attack probability corresponding to the at least one living body feature through the living body detection model; and determining the first average probability of the at least one first sub - attack probability and determining the first average probability as the first attack probability.

[0008] In some embodiments, it further includes: obtaining at least one frame of the biological image through the living body detection model; and performing living body detection on each frame of the at least one frame of biological image through the living body detection model to obtain at least one living body feature corresponding to the at least one frame of biological image.

[0009] In some embodiments, Scheme 1 further includes: comparing the comparison feature with the reference feature through the comparison model to obtain a comparison value, where the comparison value reflects the similarity between the comparison feature and the reference feature; and determining that the comparison value is between a first comparison threshold and a second comparison threshold, and based on the comparison feature, performing living body verification on the biological image through the living body verification model.

[0010] In some embodiments, the comparison feature includes at least one comparison feature. Obtaining the comparison value includes: determining the average comparison feature of the at least one comparison feature; and comparing the average comparison feature with the reference feature to obtain the comparison value.

[0011] In some embodiments, it further includes: obtaining at least one frame of the biological image and at least one living body feature corresponding to the at least one frame of biological image through the comparison model; and for each frame of the at least one frame of biological image, using the corresponding living body feature to perform feature comparison on the biological image through the comparison model to obtain the corresponding comparison feature, so as to obtain at least one comparison feature corresponding to the at least one frame of biological image.

[0012] In some embodiments, performing living body verification on the biological image through the living body verification model based on the comparison feature to obtain the first recognition result includes: performing living body verification on the biological image through the living body verification model based on the comparison feature and the living body feature to obtain a verification feature; and obtaining the first recognition result based on the verification feature.

[0013] In some embodiments, the obtaining of the verification features includes: obtaining at least one frame of the biological image through the liveness verification model, as well as at least one liveness feature and at least one comparison feature respectively corresponding to the at least one frame of biological image; for each frame of the at least one frame of biological image, using the corresponding liveness feature and the corresponding comparison feature through the liveness verification model to perform liveness verification on the biological image, and obtaining the corresponding verification feature, so as to obtain at least one verification feature corresponding to the at least one frame of biological image.

[0014] In some embodiments, the obtaining of the first recognition result based on the verification features includes: obtaining the second attack probability of the target part based on the verification features; and obtaining the first recognition result based on the second attack probability.

[0015] In some embodiments, the verification features include at least one verification feature. The obtaining of the second attack probability based on the verification features includes: determining at least one second sub-attack probability corresponding to the at least one verification feature; and determining the second average probability of the at least one second sub-attack probability, and determining the second average probability as the second attack probability.

[0016] In some embodiments, the obtaining of the first recognition result based on the second attack probability includes: determining that the second attack probability is greater than the verification threshold, then the first recognition result is a biometric recognition failure; or determining that the second attack probability is less than the verification threshold, then the first recognition result is a biometric recognition success.

[0017] In some embodiments, the multiple groups of solutions further include Solution 2, and Solution 2 includes: performing feature comparison on the biological image based on the comparison model to obtain comparison features, and the comparison model uses the liveness features for the feature comparison; and obtaining a second recognition result based on the comparison features.

[0018] In some embodiments, the obtaining of the second recognition result based on the comparison features includes: comparing the comparison features with reference features through the comparison model to obtain a comparison value, and the comparison value reflects the accuracy of the comparison of the biological image; and obtaining the second recognition result based on the comparison value.

[0019] In some embodiments, the obtaining of the second recognition result based on the comparison value includes: determining that the comparison value is less than the first comparison threshold, then the second recognition result is a biometric recognition failure; or determining that the comparison value is greater than the second comparison threshold, then the second recognition result is a biometric recognition success, and the first comparison threshold is less than the second comparison threshold.

[0020] In some embodiments, the multiple sets of solutions further include Solution Three, which includes: obtaining a third recognition result, where the third recognition result is a biometric recognition failure.

[0021] In some embodiments, the obtaining of the third recognition result includes: determining that the first attack probability is greater than a detection threshold, and obtaining the third recognition result.

[0022] In some embodiments, the liveness detection model includes ResNet18 and a binary classifier; the comparison model includes ResNet101 and a multi-classifier; and the liveness verification model includes MLP and a binary classifier.

[0023] In some embodiments, it further includes: independently training the liveness detection model, the comparison model, and the liveness verification model respectively; and jointly training the separately trained liveness detection model, the separately trained comparison model, and the separately trained liveness verification model.

[0024] In a second aspect, this specification also provides a biometric recognition system, including at least one storage medium and at least one processor. The at least one storage medium stores a liveness detection model, a comparison model, a liveness verification model, and at least one set of instruction sets for implementing biometric recognition. The at least one processor is communicatively connected to the at least one storage medium. When the biometric recognition system runs, the at least one processor reads the liveness detection model, the comparison model, the liveness verification model, and the at least one instruction set and implements the biometric recognition method described in the first aspect of this specification.

[0025] As can be seen from the above technical solutions, the biometric recognition method and system provided in this specification include a liveness detection model, a comparison model, and a liveness verification model. The liveness detection model is used to obtain liveness features. The comparison model can use the liveness features of the liveness detection model for feature comparison. The liveness verification model can use the comparison features of the comparison model for liveness verification. In this way, liveness judgment (liveness detection and liveness verification) and feature comparison are alternately performed through the three models in three stages. At the same time, the liveness judgment model and the comparison model can mutually use the feature information of each other to improve their respective performances, thereby achieving more accurate biometric recognition.

[0026] Other functions of the biometric recognition method and system provided in this specification will be partially listed in the following description. According to the description, the content introduced by the following numbers and examples will be obvious to those of ordinary skill in the art. The creative aspects of the biometric recognition method and system provided in this specification can be fully explained through practice or the use of the methods, devices, and combinations described in the detailed examples below. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] To more clearly illustrate the technical solutions in the embodiments of this specification, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of this specification. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0028] Figure 1 Fig. shows a schematic diagram of an application scenario of a biometric system 001 provided according to some embodiments of this specification;

[0029] Figure 2 Fig. shows a hardware structure diagram of a computing device 600 provided according to some embodiments of this specification;

[0030] Figure 3 Fig. shows a flowchart of a biometric method P100 provided according to some embodiments of this specification;

[0031] Figure 4 Fig. shows a flowchart of a biometric method S160 provided according to some embodiments of this specification; and

[0032] Figure 5 Fig. shows a schematic diagram of a face recognition method provided according to some embodiments of this specification. Detailed implementation manners

[0033] The following description provides specific application scenarios and requirements of this specification, aiming to enable those skilled in the art to manufacture and use the content in this specification. For those skilled in the art, various partial modifications to the disclosed embodiments are obvious, and without departing from the spirit and scope of this specification, the general principles defined here can be applied to other embodiments and applications. Therefore, this specification is not limited to the shown embodiments, but has the broadest scope consistent with the claims.

[0034] The terms used here are only for the purpose of describing specific example embodiments and are not restrictive. For example, unless otherwise clearly stated in the context, the singular forms "a", "an" and "the" used here may also include the plural forms. When used in this specification, the terms "include", "comprise" and / or "contain" mean that the associated integers, steps, operations, elements and / or components exist, but do not exclude the existence of one or more other features, integers, steps, operations, elements, components and / or groups, or the addition of other features, integers, steps, operations, elements, components and / or groups in the system / method.

[0035] In view of the following description, these features of the present specification and other features, as well as the operations and functions of the relevant elements of the structure, and the combination and manufacturing economy of the components can be significantly improved. Referring to the accompanying drawings, all of which form a part of the present specification. However, it should be clearly understood that the drawings are only for illustrative and descriptive purposes and are not intended to limit the scope of the present specification. It should also be understood that the drawings are not drawn to scale.

[0036] The flowcharts used in the present specification illustrate the operations implemented by the system according to some embodiments in the present specification. It should be clearly understood that the operations of the flowchart may not be implemented in sequence. On the contrary, the operations may be implemented in reverse order or simultaneously. In addition, one or more other operations may be added to the flowchart. One or more operations may be removed from the flowchart.

[0037] Before describing the specific embodiments of the present specification, the application scenarios of the present specification are introduced as follows:

[0038] The biometric method provided in this specification can be applied to scenarios that require identifying human body biometric features, such as face recognition, fingerprint recognition, iris recognition, etc. Common face recognition scenarios include face payment, face access control, face attendance, etc. Common fingerprint recognition scenarios include fingerprint smart door locks, fingerprint mobile phone unlocking, fingerprint software decryption, etc. Common iris recognition scenarios include access control and attendance, judicial security inspection, etc. The biometric method provided in this specification can also be applied to any other scenario that requires identifying the biometric features of the human body or animal body.

[0039] For the convenience of description, the terms that will appear in the following description are explained in this specification:

[0040] Face comparison: refers to a method of extracting face features using a face comparison model and then comparing the faces by the similarity of the features between the target face and the reserved face.

[0041] Liveness attack: Non-liveness refers to an attack means presented against a biometric system, including mobile phone screens, printed photos, high-precision masks, molds, prostheses, etc.

[0042] Liveness anti-attack: refers to the algorithms and technologies used in a biometric verification system to prevent attack means such as mobile phone screen recording attacks, paper photo attacks, mask attacks, mold attacks, prosthesis attacks, etc.

[0043] Figure 1 The application scenario schematic diagram of a biometric system 001 provided according to some embodiments of the present specification is shown. As Figure 1 shown, the system 001 may include a target object 100, a client 200, a server 300, and a network 400.

[0044] The target object 100 can be any user who uses the client 200 to collect biometric features.

[0045] The client 200 can collect the target part of the target object 100. The target part has the physiological characteristics of the target object. The target part can be, for example, a human face, fingerprint, iris, retina, hand shape, etc. Then, a biometric image of the target part can be collected, such as a face image, fingerprint image, iris image, retina image, hand shape image, and so on. Of course, the client 200 can also collect the behavioral characteristics of the target object 100 and collect behavioral information, such as voice characteristics, behavioral characteristics, gait characteristics, and so on. The client 200 or the server 300 can use the biometric recognition method described in this specification to perform biometric recognition on the biometric image and output the recognition result. In some embodiments, the biometric recognition method can be executed on the client 200. At this time, the client 200 can store the data or instructions for executing the biometric recognition method described in this specification and can execute or be used to execute the data or instructions. In some embodiments, the client 200 can include a hardware device with data information processing functions and the necessary programs for driving the hardware device to work. Such as Figure 1As shown, the client 200 can communicate with the server 300. In some embodiments, the server 300 can communicate with multiple clients 200. In some embodiments, the client 200 can interact with the server 300 via the network 400 to receive or send messages, such as receiving or sending biological images or various feature information, such as two-dimensional images / features and / or three-dimensional images / features. In some embodiments, the client 200 can include a mobile device, a tablet computer, a laptop computer, a built-in device of a motor vehicle or the like, the Dragonfly device of Alipay, a vending machine, a vending cabinet, or any combination thereof. In some embodiments, the mobile device can include a smart home device, a smart mobile device, a virtual reality device, an augmented reality device or the like, or any combination thereof. In some embodiments, the smart home device can include a smart TV, a desktop computer, etc., or any combination. In some embodiments, the smart mobile device can include a smart phone, a personal digital assistant, a gaming device, a navigation device, etc., or any combination thereof. In some embodiments, the virtual reality device or the augmented reality device may include a virtual reality helmet, virtual reality glasses, a virtual reality patch, an augmented reality helmet, augmented reality glasses, an augmented reality patch or the like, or any combination thereof. For example, the virtual reality device or the augmented reality device may include Google Glasses, a head-mounted display, VR, etc. In some embodiments, the built-in device in the motor vehicle can include an in-vehicle computer, an in-vehicle TV, etc. In some embodiments, the client 200 can include an image acquisition device for acquiring biological images, such as a face image of the target object 100. In some embodiments, the image acquisition device can be a two-dimensional image acquisition device (such as an RGB camera), or a two-dimensional image acquisition device (such as an RGB camera) and a depth image acquisition device (such as a 3D structured light camera, a laser detector, etc.). In some embodiments, the client 200 can be a device with positioning technology for positioning the location of the client 200.

[0046] In some embodiments, the client 200 may be installed with one or more applications (APPs). The APP can provide the target object 110 with the ability to interact with the outside world through the network 400 and an interface. The APP includes, but is not limited to: web browser APP programs, search APP programs, chat APP programs, shopping APP programs, video APP programs, financial management APP programs, instant messaging tools, email clients, social platform software, and so on. In some embodiments, a target APP may be installed on the client 200. The target APP can collect biometric images for the client 200. In some embodiments, the target APP can also identify biometric images. The target object 100 can trigger a biometric recognition request through the target APP. The target APP can execute a biometric recognition method in response to the biometric recognition request.

[0047] The server 300 can be a server that provides various services, such as a background server that supports the pages displayed on the client 200. In some embodiments, the biometric recognition method can be executed on the server 300. At this time, the server 300 can store data or instructions for executing the biometric recognition method described in this specification and can execute or be used to execute the data or instructions. In some embodiments, the server 300 can include a hardware device with data information processing capabilities and necessary programs for driving the hardware device to work. The server 300 can be communicatively connected to multiple clients 200 and receive data sent by the clients 200.

[0048] The network 400 is a medium for providing a communication connection between the client 200 and the server 300. The network 400 can facilitate the exchange of information or data. As Figure 1 shown, the client 200 and the server 300 can be connected to the network 400 and transmit information or data to each other through the network 400. In some embodiments, the network 400 can be any type of wired or wireless network, or a combination thereof. For example, the network 400 can include a cable network, a wired network, an optical fiber network, a telecommunications network, an intranet, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), a wide area network (WAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigBee network, a near field communication (NFC) network, or a similar network. In some embodiments, the network 400 can include one or more network access points. For example, the network 400 can include wired or wireless network access points, such as base stations or Internet exchange points, through which one or more components of the client 200 and the server 300 can be connected to the network 400 to exchange data or information.

[0049] It should be understood, Figure 1The numbers of the client 200, the server 300, and the network 400 in it are merely illustrative. According to the implementation requirements, there can be any number of clients 200, servers 300, and networks 400.

[0050] It should be noted that the biometric method can be executed entirely on the client 200, entirely on the server 300, or partially on the client 200 and partially on the server 300.

[0051] Figure 2 The hardware structure diagram of a computing device 600 provided according to some embodiments of the present specification is shown. The computing device 600 can execute the biometric method described in the present specification. The biometric method is introduced in other parts of the present specification. When the biometric method is executed on the client 200, the computing device 600 can be the client 200. When the biometric method is executed on the server 300, the computing device 600 can be the server 300. When the biometric method can be partially executed on the client 200 and partially on the server 300, the computing device 600 can be the client 200 and the server 300.

[0052] As Figure 2 shown, the computing device 600 may include at least one storage medium 630 and at least one processor 620. In some embodiments, the computing device 600 may further include a communication port 650 and an internal communication bus 610. At the same time, the computing device 600 may further include I / O components 660.

[0053] The internal communication bus 610 can connect different system components, including the storage medium 630, the processor 620, and the communication port 650.

[0054] The I / O components 660 support input / output between the computing device 600 and other components.

[0055] The communication port 650 is used for data communication between the computing device 600 and the outside world. For example, the communication port 650 can be used for data communication between the computing device 600 and the network 400. The communication port 650 can be a wired communication port or a wireless communication port.

[0056] The storage medium 630 may include a data storage device. The data storage device may be a non-transitory storage medium or a transitory storage medium. For example, the data storage device may include one or more of a magnetic disk 632, a read-only storage medium (ROM) 634, or a random access storage medium (RAM) 636. The storage medium 630 may store a live detection model, a comparison model, a live verification model, and at least a set of instruction sets for implementing biometric identification. The instructions are computer program codes, and the computer program codes may include programs, routines, objects, components, data structures, procedures, modules, etc. for executing the biometric identification method provided in this specification. Among them, the model may be one or more instruction sets stored in the storage medium 630 that execute corresponding instructions and are executed by the processor 620 in the computing device 600. The model may also be a part of the circuit, a hardware device, or a module in the computing device 600. For example, the live detection model may be a hardware device / module in the computing device 600 that implements live detection, and the comparison model may be a hardware device / module in the computing device 600 that implements feature comparison, and so on.

[0057] At least one processor 620 may be communicatively connected to at least one storage medium 630 and a communication port 650 via an internal communication bus 610. The at least one processor 620 is configured to execute the above-mentioned at least one instruction set. When the computing device 600 is running, the at least one processor 620 may read the liveness detection model, the comparison model, the liveness verification model, and the at least one instruction set, and execute the biometric identification method provided in this specification according to the instructions of the at least one instruction set. The processor 620 may execute all the steps included in the biometric identification method. The processor 620 may be in the form of one or more processors. In some embodiments, the processor 620 may include one or more hardware processors, such as a microcontroller, a microprocessor, a reduced instruction set computer (RISC), an application specific integrated circuit (ASIC), an application specific instruction set processor (ASIP), a central processing unit (CPU), a graphics processing unit (GPU), a physics processing unit (PPU), a microcontroller unit, a digital signal processor (DSP), a field programmable gate array (FPGA), an advanced RISC machine (ARM), a programmable logic device (PLD), any circuit or processor capable of executing one or more functions, etc., or any combination thereof. For illustrative purposes only, only one processor 620 is described in the computing device 600 in this specification. However, it should be noted that the computing device 600 in this specification may also include multiple processors. Therefore, the operations and / or method steps disclosed in this specification may be executed by one processor as described in this specification, or jointly executed by multiple processors. For example, if the processor 620 of the computing device 600 executes step A and step B in this specification, it should be understood that step A and step B may also be executed jointly or separately by two different processors 620 (e.g., the first processor executes step A, the second processor executes step B, or the first and second processors jointly execute steps A and B).

[0058] Figure 3 FIG. 4 shows a flowchart of a biometric identification method P100 according to some embodiments of this specification. As mentioned above, the computing device 600 may execute the biometric identification method P100 described in this specification. Specifically, the processor 620 may read the instruction set stored in its local storage medium, and then execute the biometric identification method P100 described in this specification according to the provisions of the instruction set. As Figure 3 shown, the method P100 may include:

[0059] S120: Train a biometric identification model, where the biometric identification model includes a liveness detection model, a comparison model, and a liveness verification model.

[0060] The live detection model is used to determine whether the target object is a live body or an attack. The live detection model can adopt a variety of different network structures. For example, it can be composed of ResNet18 (Residual Neural Network) and a binary classifier, or it can be CDCN (Central Difference Convolutional Networks), etc. The comparison model is used to identify the identity of the target object by comparison. The comparison model can adopt a variety of different network structures. For example, it can be composed of ResNet101 and a multi-classifier, or it can be FaceNet, etc. The live verification model is used to determine whether the target object is a live body or an attack. The live verification model can adopt a variety of different network structures. For example, it can be composed of an MLP (Multilayer Perceptron) and a binary classifier. The MLP is a 3-layer MLP, or it can be CDCN, etc. Among them, the functions of the live detection model and the live verification model are both to determine whether the target object is a live body or an attack. They can adopt the same network structure or different network structures.

[0061] The "attack" refers to the "live body attack" in the noun explanation. For face recognition, the live body refers to the face of a real person, rather than an attacking face; the attacking face can be a face photo displayed on an electronic screen (such as a mobile phone screen), a printed paper photo, a high-precision face mask, etc. For fingerprint recognition, the live body refers to a real finger, rather than an attacking fingerprint; the attacking fingerprint can be a printed fingerprint photo, a fingerprint mold made of plastic material (such as a silicone fingerprint, also known as a "false finger"), etc. For iris recognition, the live body refers to the real iris of a real person, rather than an attacking iris; the attacking iris can be a prosthetic iris, etc.

[0062] The processor can obtain the training set and test set data, and independently train the liveness detection model, comparison model, and liveness verification model in the biometric model using the training set and test set data, that is, each of the three models is trained separately, and the three models are trained into an initial state that can be initially used. In the initial state, the liveness detection model, comparison model, and liveness verification model can initially complete the biometric process, but the recognition accuracy is lower than that of the finally trained model. Subsequently, the processor can perform joint training on the three models that have completed separate training, so that the input and output of these three models can cross each other. For example, the processor can input a biometric image into the liveness detection model for training, input the biometric image and the liveness features output by the liveness detection model into the comparison model for training, and input the biometric image, liveness features, and the comparison features output by the comparison model into the liveness verification model for training. Another example is that the processor can also input only the biometric image and the comparison features output by the comparison model into the liveness verification model for training. Among them, in the joint training, the loss function of the biometric model consists of three parts, namely, the loss function of the liveness detection model, the loss function of the comparison model, and the loss function of the liveness verification model. The biometric model is continuously trained through the loss function until the model converges, thereby completing the training process. It should be noted that in addition to the above three models, the biometric model can also include a liveness detection model and a comparison model, or include a comparison model and a liveness verification model, or further include other models based on the above three models. Whatever models the biometric model includes, those models will be trained during training.

[0063] The processor can put the trained biometric model into use. In actual use, the processor can capture the target part (such as face, fingerprint, iris, etc.) of the target object through the client to obtain the biometric image of the target part (such as face image, fingerprint image, iris image, etc.), and determine whether the target object is a live body and confirm the identity of the target object through steps S140 - S160.

[0064] S140: Execute the liveness detection model to perform liveness detection on the biometric image corresponding to the target part of the target object and obtain the liveness detection result.

[0065] The processor can input the biological images collected by the client into the liveness detection model and execute the model to obtain the liveness detection result. The biological images collected by the client for the target part can be a single frame or multiple frames. If multiple biological images are collected, the processor can set the collection duration to a preset duration, such as 0.5s, 1s, etc., and collect multiple frames for the same target part within the preset duration. For example, 10 - 15 frames. The processor can input the multiple biological images into the liveness detection model, and perform liveness detection on each frame of the multiple biological images through the liveness detection model, that is, extract liveness features from each frame of the biological image, so as to output multiple liveness features. The processor can input each of the multiple liveness features into the binary classifier of the liveness detection model for classification, obtain the first sub - attack probability corresponding to each liveness feature, and then take the first average probability of the multiple first sub - attack probabilities as the first attack probability Pd of the target part. The first attack probability can reflect the probability that the target part belongs to an attack. If a single biological image is collected, the processor only needs to extract the liveness feature of the single biological image, and the attack probability of this liveness feature is the first attack probability of the target part.

[0066] The biometric recognition method based on a single - frame image can improve the speed and efficiency of biometric recognition. However, there are prone to problems of quality fluctuations in single - frame images. For example, in face recognition, changes in the user's posture, expression, or environmental light can cause the quality of the biological image to be low, resulting in a decrease in the recognition accuracy and causing false disturbances. The so - called "false disturbance" refers to misjudging a live body as an attack during the liveness judgment process, and misidentifying user A as user B during the comparison process, thus disturbing the user. Therefore, the performance of biometric recognition / decision - making based on single - frame images is often poor. The processor can adopt a multi - frame fusion decision - making method to overcome the shortcomings of single - frame decision - making.

[0067] Among them, the liveness feature can be the physiological feature of the target part. For example, the liveness features of live face detection can be information such as face posture, face rotation, breathing, red - eye effect, etc. The liveness features of live fingerprint detection can be information such as finger temperature, sweating, electrical conductivity, etc. The liveness features of live iris detection can be information such as iris tremor characteristics, movement information of eyelashes and eyelids, and pupil contraction and dilation response characteristics to visible light source intensity. The liveness feature can also be the environmental information of the target object's environment. For example, if the target object often performs biometric recognition in a certain specific environment, such as face - swiping / fingerprint recognition / iris recognition in a yellow - light scene, then to a certain extent, the environmental information can also reflect whether the target object (or target part) is a live body.

[0068] S160: Based on the liveness features, select one from multiple sets of solutions to execute, complete the biometric recognition of the target object, and output the recognition result.

[0069] The multiple sets of solutions include Solution 1, Solution 2, and Solution 3. The processor can set a detection threshold for the live detection model. For ease of description, the detection threshold will be referred to as the first detection threshold T1 hereinafter, and select a solution from the multiple sets of solutions and execute it based on Pd and T1.

[0070] Figure 4 The flowchart of a biometric method S160 provided according to some embodiments of the present specification is shown. As Figure 4 shown, the method S160 may include: S162 to S168.

[0071] When Pd > T1, Solution 3 can be executed. Solution 3 includes S162.

[0072] S162: Obtain a third recognition result.

[0073] The first attack probability Pd > T1 indicates that the target part is an attack. Then, the third recognition result can be directly output, and the third recognition result is a biometric recognition failure. At this time, the processor can reject subsequent operations of the target object. For example, it does not respond to any trigger operations of the target object on the client, and does not execute the comparison model and the live verification model.

[0074] It should be noted that in addition to setting the first detection threshold T1 for the live detection model, a second detection threshold T2 can also be set for it, and T2 < T1. When the first attack probability Pd < T1, there are two cases: Pd < T2 and Pd is between T2 and T1. The case of Pd < T2 can indicate that the target part is a live body. At this time, it can be considered that the live body judgment is completed. The purpose of further executing S164 is to authenticate the identity of the target object. The case of Pd being between T2 and T1 indicates that it is impossible to determine whether the target part is a live body or an attack. At this time, it can still enter the next step, that is, S164 to authenticate the identity of the target object.

[0075] It can be seen that when Pd < T1, the identity information of the target object can be further determined according to the biological image. That is, the feature comparison process in Solution 2 or Solution 1 is executed through step S164.

[0076] S164: Based on the comparison model, perform feature comparison on the biological image and obtain comparison features. The comparison model uses the live body features for the feature comparison.

[0077] In some embodiments, the processor may introduce liveness features into the comparison model to perform feature comparison. As described above, the processor may collect a single-frame biometric image or multiple-frame biometric images. Taking multiple-frame biometric images as an example, the processor may input the multiple-frame biometric images into the comparison model, and introduce a plurality of liveness features corresponding to the multiple-frame biometric images during the intermediate process of executing the comparison model. For example, a plurality of liveness features are introduced into a certain convolutional layer of the comparison model, and each liveness feature is used to extract comparison features from the biometric image corresponding to one frame, so as to obtain a plurality of comparison features corresponding to the multiple-frame biometric images.

[0078] This specification first executes a liveness verification model before the comparison model, so that liveness features can be introduced during the feature comparison process, making the extracted comparison features contain richer information, thereby improving the comparison performance. For example, the liveness feature is the face angle. When the target object often swipes the face at a certain specific angle, this specific angle can be considered to be bound to the identity of the target object. Then, the identity information of the target object can also be assisted in being identified through this specific angle. In addition, the comparison model often needs to compare the target object with multiple reference objects in the database, which occupies a large amount of machine resources. However, the liveness model only needs to divide two categories through a binary classifier and occupies less resources. For example, the comparison model occupies 80% of the processor resources, while the liveness model only occupies 20%. If the liveness detection model is not executed before the comparison model, but the comparison model is directly executed, then when encountering a large number of malicious attacks, the comparison model will occupy a large amount of machine resources, the time consumption will increase significantly, and even cause the entire biometric system to crash. If the liveness detection model is executed before the comparison model, since the liveness detection model occupies less resources, even when encountering a large number of malicious attacks, most of the attacks can be easily filtered out, and only a small number of attacks enter the comparison model, thereby reducing the time consumption of biometric identification, reducing the pressure on the biometric system, and avoiding the paralysis of the biometric system.

[0079] Further, the processor may determine the average comparison feature of the plurality of comparison features, and compare the average comparison feature with the reference feature in the comparison database to obtain a comparison value S, and the comparison value reflects the similarity between the comparison feature and the reference feature. In some embodiments, the processor may calculate the comparison value by calculating the similarity (such as cosine similarity) between the average comparison feature and the reference feature. Wherein, the comparison database includes a plurality of reference features, and the reference feature refers to a feature with known identity information, and the identity information of the target object can be obtained by comparing the comparison feature with the reference feature.

[0080] In some embodiments, the processor may set a third comparison threshold, such as T0, for the comparison model, and compare the comparison value S with T0. When S > T0, the second recognition result is a successful biometric recognition; when S < T0, the second recognition result is a failed biometric recognition. At this time, Solution 1 may not be included in multiple groups of solutions.

[0081] In some embodiments, the processor may also set two different comparison thresholds for the comparison model, namely the first comparison threshold T3 and the second comparison threshold T4, where T3 < T4, and T0 may be between T3 and T4. At this time, the processor may determine whether to implement the liveness verification model based on the magnitude relationship between the comparison value S and T3 and T4, that is, to determine whether to execute Solution 2 to obtain the second recognition result or execute Solution 1 to obtain the first recognition result.

[0082] If S is not between T3 and T4, the liveness verification model is not implemented, and Solution 2 is executed through S166 to obtain the second recognition result.

[0083] S166: Obtain the second recognition result based on the comparison feature.

[0084] When S < T3, it indicates that the identity recognition of the target object fails, so the second recognition result is a failed biometric recognition; when S > T4, it indicates that the identity recognition of the target object is successful, so the second recognition result is a successful biometric recognition.

[0085] As mentioned above, Pd < T1 includes two cases: Pd < T2 and Pd is between T2 and T1. When Pd < T2 and S < T3, it indicates that although the target part is a live body, the identity information is incorrect, and it is very likely that someone else (a live body) is impersonating. At this time, the second recognition result, that is, a failed biometric recognition, can be directly obtained, and the liveness detection model is no longer executed. When Pd < T2 and S > T4, it indicates that the target part is both a live body and the identity information comparison is successful, indicating that the target object is very safe and there are no problems with liveness and identity. At this time, the second recognition result, that is, a successful biometric recognition, can be directly obtained, and the liveness detection model does not need to be executed.

[0086] When Pd is between T2 and T1 and S < T3, it indicates that it is impossible to determine whether the target part is a live body or an attack, and the identity information is also incorrect. There are security risks for the live body and identity. At this time, the second recognition result can be directly obtained, that is, the biometric recognition fails, and the live body detection model is no longer executed. When Pd is between T2 and T1 and S > T4, it indicates that although it is impossible to determine whether the target part is a live body or an attack, the result of the identity information comparison is correct. Since the feature comparison process is carried out by introducing live body features, to a certain extent, the feature comparison can reflect the problem of the live body. Therefore, if S > T4, it can reflect both that the identity information comparison is correct and that the target part is a live body. Then, when Pd is between T2 and T1 and S > T4, the second recognition result can be directly obtained, that is, the biometric recognition is successful, and there is no need to execute the live body detection model.

[0087] If S is between T3 and T4, the live body verification model is implemented, and Scheme 1 is executed through S168, that is, further live body verification is performed and the first recognition result is obtained.

[0088] S168: Based on the comparison features, the live body verification model is used to perform live body verification on the biological image to obtain the first recognition result.

[0089] As mentioned above, the processor can set only one comparison threshold T0, or can set two comparison thresholds T3 and T4, where T3 < T4 and T0 is between T3 and T4. However, the method of setting one comparison threshold is too rigid and the user experience is low. The method of setting two comparison thresholds increases the upper limit of the comparison threshold (from T0 to T4) and adds T3. In this way, a small part (between T3 and T4) of the biological images smaller than T4 can be released to pass through the comparison model and enter the next stage for further verification.

[0090] And Pd < T1 includes two cases: Pd < T2 and Pd is between T2 and T1. When Pd < T2 and S is between T3 and T4, it indicates that although the target part is a live body, the identity information cannot be determined, and there are security problems with the identity information, and the recognition result cannot be obtained through Scheme 2. According to a large number of experiments, this security problem mainly comes from live body attacks. Therefore, at this time, the live body verification model can be further executed.

[0091] When Pd is between T2 and T1, and S is at T3 and T4, it indicates that neither can it be determined whether the target part is a living body or an attack, nor can the identity information of the target object be determined. At this time, it is more necessary to further execute the living body verification model. On the one hand, the living body verification model can make up for some functions of the comparison model and use the result of the living body verification as the result of the comparison model. On the other hand, the living body verification model can further perform a living body judgment on the biological image that has passed through the living body detection model. In this way, the first detection threshold of the living body detection model can be set more loosely. For example, the original first detection threshold was 0.5, and only the biological image with Pd < 0.5 could pass through the living body detection model. Now, the first detection threshold can be set to 0.7, so that the biological image with Pd < 0.7 can pass through the living body detection model, thereby increasing the passing rate of the living body detection model. In this way, by setting the living body verification model, the first detection threshold of the living body verification model can be increased, so that more data can pass through the living body verification model, and further reduce the false disturbance rate for users.

[0092] When the processor executes the living body verification model, it can introduce both the living body features and the comparison features into the living body verification model. Taking multiple frames of biological images as an example, the processor can input multiple frames of biological images into the living body verification model and introduce multiple living body features and multiple comparison features corresponding to the multiple frames of biological images during the intermediate process of executing the living body verification model. For example, these multiple living body features and these multiple comparison features are introduced in the convolutional layer. The convolutional layer introducing the living body features and the convolutional layer introducing the comparison features can be the same or different. Then, for each frame in the multiple frames of biological images, the processor can use the corresponding living body features and the corresponding comparison features through the living body verification model to extract the corresponding verification features from the biological image, so as to obtain multiple verification features corresponding to the multiple frames of biological images. In this way, by introducing the comparison features during the living body verification process, the living body verification model can make full use of the comparison information to perform the verification of living body attacks, intercept some unsafe attacks, and improve the interception ability against living body attacks. At the same time, the living body verification model can also use the living body features of the living body detection model to further improve its interception ability against living body attacks.

[0093] Furthermore, the processor can determine the second sub - attack probability corresponding to each of the multiple verification features, and use the second average probability of the multiple second sub - attack probabilities as the second attack probability Pc of the target part. The second attack probability can reflect the probability that the target part belongs to an attack. The processor can set a verification threshold T5 for the living body verification model. If Pc > T5, the first recognition result is biometric failure; if Pc < T5, the first recognition result is biometric success.

[0094] It should be noted that the detection threshold, comparison threshold, and verification threshold in this specification form several intervals, such as the interval range between T3 and T4, the interval less than T3, etc. These intervals can be closed intervals or open intervals, and this specification does not make any limitations in this regard.

[0095] Of course, the multiple sets of solutions in this specification can also include other solutions in addition to Solutions One, Two, and Three. For example, in some embodiments, a comparison verification model can be added after the live body verification model of Solution One to further verify the results of the comparison model. In some embodiments, only the live body detection model and the comparison model may be included, and the execution order of these two models is not limited. In some embodiments, when only the live body detection model and the comparison model are included, these two models can be executed alternately, that is, the comparison model receives the live body features output by the live body detection model and inputs the output comparison features into the live body detection model. In some embodiments, the live body detection model and the comparison model can be set in multiple rounds, each round including a live body detection model and a comparison model and executed alternately, and the number of rounds is determined by a cut-off condition. For example, the cut-off condition is that the accuracy rate of biometric recognition reaches a preset accuracy rate.

[0096] Among them, in the face recognition scenario, this specification can Figure 5 implement face recognition through the Figure 5 shows a schematic diagram of a face recognition method provided according to some embodiments of this specification. As Figure 5 shown, the face recognition process can include basic model training, first-stage live body detection, second-stage comparison verification, and third-stage live body verification, which can be specifically as follows:

[0097] (1) Basic model training: In the training stage, three models are trained, including a first-stage live body detection model, a second-stage comparison model, and a third-stage live body verification model;

[0098] (2) First-stage live body detection: Input the face image into the first-stage live body detection model to quickly reject simple attack samples;

[0099] (3) Second-stage comparison verification: The passed samples will be further input into the face comparison model for comparison to obtain comparison information;

[0100] (4) Third-stage live body verification: Make full use of the information in the previous two stages for third-stage live body verification.

[0101] In summary, the biometric method and system provided in this specification include a live detection model, a comparison model, and a live verification model. The live detection model is executed to obtain live features. The comparison model can use the live features of the live detection model for feature comparison. The live verification model can use the comparison features of the comparison model for live verification. In this way, through the three models in the three stages, live judgment (live detection and live verification) and feature comparison are carried out alternately, making full use of the information of the live and comparison models for live judgment and feature comparison, realizing the information exchange between live judgment and feature comparison, and having better performance compared with the independent two-stage model (a live judgment model and a comparison model, and they do not use each other's feature information). At the same time, when training the biometric model, the above three models are also jointly trained through the way of information interaction, so that the biometric model can play the role of information interaction during actual use, thereby improving the performance of the overall model. At the same time, through the three models in the three stages, risk control can be achieved, reducing the security problems of live attacks and comparisons. At the same time, the biometric decision-making method based on multi-frame fusion can reduce the decision-making performance problems caused by the quality problems of biometric images.

[0102] On the other hand, this specification provides a non-transitory storage medium storing at least one set of executable instructions for performing biometric identification. When the executable instructions are executed by a processor, the executable instructions direct the processor to implement the steps of the biometric identification method P100 described in this specification. In some possible implementations, various aspects of this specification may also be implemented in the form of a program product, which includes program code. When the program product runs on the biometric identification system 001, the program code is used to cause the biometric identification system 001 to execute the steps of the biometric identification method P100 described in this specification. The program product for implementing the above method may be a portable compact disc read-only memory (CD-ROM) including program code and may run on the biometric identification system 001. However, the program product of this specification is not limited to this. In this specification, the readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system. The program product may be any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. The computer-readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, in which the readable program code is carried. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium may also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted with any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above. The program code for performing the operations of this specification may be written in any combination of one or more programming languages, including object-oriented programming languages - such as Java, C++, etc., and also including conventional procedural programming languages - such as the "C" language or similar programming languages. The program code may be executed entirely on the biometric identification system 001, partially on the biometric identification system 001, executed as an independent software package, partially on the biometric identification system 001 and partially on a remote computing device, or entirely on a remote computing device.

[0103] The above description has been made of specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the acts or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require a particular order or a sequential order to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0104] In summary, after reading this detailed disclosure, those skilled in the art will appreciate that the foregoing detailed disclosure may be presented by way of example only and is not limiting. Although not explicitly stated herein, those skilled in the art will understand that this specification is intended to encompass various reasonable changes, improvements, and modifications to the embodiments. These changes, improvements, and modifications are intended to be made by this specification and are within the spirit and scope of the exemplary embodiments of this specification.

[0105] Furthermore, certain terms in this specification have been used to describe embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" mean that the particular features, structures, or characteristics described in connection with that embodiment may be included in at least one embodiment of this specification. Thus, it should be emphasized and understood that two or more references to "an embodiment" or "one embodiment" or "alternative embodiments" in various parts of this specification do not necessarily all refer to the same embodiment. Additionally, the particular features, structures, or characteristics may be appropriately combined in one or more embodiments of this specification.

[0106] It should be understood that in the foregoing description of the embodiments of this specification, for the purpose of helping to understand a feature and for the purpose of simplifying this specification, this specification combines various features in a single embodiment, figure, or its description. However, this does not mean that the combination of these features is necessary, and those skilled in the art may well mark out some of the devices as separate embodiments for understanding when reading this specification. That is to say, the embodiments in this specification may also be understood as an integration of multiple sub - embodiments. And it also holds when the content of each sub - embodiment contains fewer features than all the features of a single foregoing disclosed embodiment.

[0107] Each patent, patent application, published patent application, and other materials cited herein, such as articles, books, specifications, publications, documents, items, etc., may be incorporated herein by reference. The entire content for all purposes, except any prosecution file history associated therewith, any identical that may be inconsistent or in conflict with this document, or any identical prosecution file history that may have a limiting effect on the broadest scope of the claims. Now or hereafter associated with this document. By way of example, if there is any inconsistency or conflict between the description, definition, and / or use of terms associated with any of the materials incorporated herein and the terms, descriptions, definitions, and / or of this document, the terms of this document shall govern.

[0108] Finally, it should be understood that the embodiments of the application disclosed herein are illustrative of the principles of the embodiments of this specification. Other modified embodiments are also within the scope of this specification. Therefore, the embodiments disclosed in this specification are merely by way of example and not limitation. Those skilled in the art may adopt alternative configurations in accordance with the embodiments in this specification to implement the application in this specification. Therefore, the embodiments of this specification are not limited to the embodiments precisely described in the application.

Claims

1. A biometric recognition method, comprising: Executing a live detection model to perform live detection on a biometric image corresponding to a target part of a target object and obtaining a live detection result, where the live detection result includes live features and a first attack probability; and Based on the live features and the first attack probability, selecting and executing one from multiple sets of solutions to complete the biometric recognition of the target object and output a recognition result, where the multiple sets of solutions include Solution 1, and when it is determined that the first attack probability is less than a detection threshold, selecting and executing Solution 1: Based on a comparison model, performing feature comparison on the biometric image and obtaining a first comparison feature, where the comparison model uses the live features for the feature comparison, Based on the first comparison feature, performing live verification on the biometric image through a live verification model to obtain at least one verification feature and its corresponding at least one second sub - attack probability, determining a second attack probability based on the at least one second sub - attack probability, and obtaining a first recognition result of the target part based on the second attack probability.

2. The method according to claim 1, wherein the first attack probability is determined by the following method: Based on the live features, determining the first attack probability of the target part through the live detection model.

3. The method according to claim 2, wherein the live features include at least one live feature, and the determining the first attack probability of the target part based on the live features through the live detection model includes: Determining at least one first sub - attack probability corresponding to the at least one live feature through the live detection model; And Determining a first average probability of the at least one first sub - attack probability and determining the first average probability as the first attack probability.

4. The method according to claim 3, further comprising: Obtaining at least one frame of biometric image through the live detection model; And Performing live detection on each frame of the at least one frame of biometric image through the live detection model to obtain at least one live feature corresponding to the at least one frame of biometric image.

5. The method according to claim 1, wherein Solution 1 further includes: Comparing the first comparison feature with a reference feature through the comparison model to obtain a first comparison value, where the first comparison value reflects the similarity degree between the first comparison feature and the reference feature; and Determining that the first comparison value is between a first comparison threshold and a second comparison threshold, and performing live verification on the biometric image through the live verification model based on the first comparison feature.

6. The method according to claim 5, wherein the first comparison feature includes at least one first comparison feature, and the obtaining the first comparison value includes: Determining an average comparison feature of the at least one first comparison feature; And Comparing the average comparison feature with the reference feature to obtain the first comparison value.

7. The method according to claim 6, further comprising: Obtaining at least one frame of biometric image and at least one live feature corresponding to the at least one frame of biometric image through the comparison model; And For each frame of the at least one frame of biological images, use the corresponding liveness feature to perform feature comparison on the biological image through the comparison model to obtain the corresponding first comparison feature, so as to obtain at least one first comparison feature corresponding to the at least one frame of biological images.

8. The method according to claim 1, wherein the at least one verification feature is obtained by the following method: Based on the first comparison feature and the liveness feature, perform liveness verification on the biological image through the liveness verification model to obtain the at least one verification feature.

9. The method according to claim 8, wherein obtaining the at least one verification feature includes: Obtain at least one frame of biological images, as well as at least one liveness feature and at least one first comparison feature respectively corresponding to the at least one frame of biological images through the liveness verification model; and For each frame of the at least one frame of biological images, perform liveness verification on the biological image through the liveness verification model using the corresponding liveness feature and the corresponding first comparison feature to obtain the corresponding verification feature, so as to obtain at least one verification feature corresponding to the at least one frame of biological images.

10. The method according to claim 1, wherein The determining the second attack probability based on the at least one second sub - attack probability includes: Determine the second average probability of the at least one second sub - attack probability, and determine the second average probability as the second attack probability.

11. The method according to claim 1, wherein, Obtaining a first recognition result of the target part based on the second attack probability includes: Determine that the second attack probability is greater than the verification threshold, then the first recognition result is biometric recognition failure; or Determine that the second attack probability is less than the verification threshold, then the first recognition result is biometric recognition success.

12. The method according to claim 1, wherein the multiple sets of solutions further include Solution 2, and Solution 2 includes: Based on the comparison model, perform feature comparison on the biological image to obtain a second comparison feature, and the comparison model uses the liveness feature for the feature comparison; and Obtain a second recognition result based on the second comparison feature.

13. The method according to claim 12, wherein obtaining the second recognition result based on the second comparison feature includes: Compare the second comparison feature with a reference feature through the comparison model to obtain a second comparison value, and the second comparison value reflects the accuracy of the biological image comparison; and Obtain the second recognition result based on the second comparison value.

14. The method according to claim 13, wherein obtaining the second recognition result based on the second comparison value includes: Determine that the second comparison value is less than a first comparison threshold, then the second recognition result is biometric recognition failure; or Determine that the second comparison value is greater than a second comparison threshold, then the second recognition result is biometric recognition success, and the first comparison threshold is less than the second comparison threshold.

15. The method according to claim 2, wherein the multiple sets of solutions further include Solution 3, and Solution 3 includes: Obtain a third recognition result, and the third recognition result is biometric recognition failure.

16. The method according to claim 15, wherein obtaining the third recognition result includes: Determine that the first attack probability is greater than the detection threshold, and obtain the third recognition result.

17. The method according to claim 1, wherein the liveness detection model includes ResNet18 and a binary classifier; the comparison model includes ResNet101 and a multi-classifier; and the liveness verification model includes MLP and a binary classifier.

18. The method according to claim 1, further comprising: independently training the liveness detection model, the comparison model, and the liveness verification model respectively; and jointly training the separately trained liveness detection model, the separately trained comparison model, and the separately trained liveness verification model.

19. A biometric recognition system, comprising: at least one storage medium storing a liveness detection model, a comparison model, a liveness verification model, and at least one set of instruction sets for implementing biometric recognition; and at least one processor communicatively connected to the at least one storage medium, wherein when the biometric recognition system runs, the at least one processor reads the liveness detection model, the comparison model, the liveness verification model, and the at least one set of instruction sets and implements the biometric recognition method according to any one of claims 1-18.

Images