Dynamic updating method and device of biological characteristics, equipment and storage medium

Abstract

The invention discloses a dynamic updating method and device for biological characteristics, equipment and a storage medium. Comprising the steps of collecting real-time authentication features of a user, and extracting historical fusion features of the user; calculating a feature quality factor, a similarity factor, a time interval weight and a historical cumulative weight based on the real-time authentication feature and the historical fusion feature; training a fusion weight prediction model based on the multi-layer perceptron neural network model; inputting the feature quality factor, the similarity factor, the time interval weight and the historical cumulative weight into a fusion weight prediction model, and outputting a dynamic fusion weight; and based on the dynamic fusion weight, fusing the real-time authentication feature and the historical fusion feature to obtain an updated fusion feature. According to the method, the newly collected features and the historical fusion features are combined to update the feature library, the feature fusion weight is dynamically adjusted through the multi-dimensional feature factors and the nonlinear model, and the complex relationship among the factors is captured through the model, so that the weight adjustment is more flexible and accurate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biometric recognition, in particular to a dynamic updating method and device for biometric features, equipment and storage medium. BACKGROUND

[0002] With the continuous progress of science and technology, biometric recognition technologies such as face recognition and fingerprint authentication have become important means of identity authentication and are widely used in access control, payment and attendance systems. These systems achieve fast and convenient identity authentication by matching real-time collected biometric features with pre-stored features in the database.

[0003] However, traditional biometric recognition systems have a significant limitation: they rely on static database features, which are generated when the user registers and remain unchanged in subsequent use. This static feature storage method cannot reflect the latest state of the user's biometric features in a timely manner, because the user's biometric features (such as appearance and fingerprints) will change over time and with environmental conditions. This not only reduces the recognition accuracy of the system, but also may cause authentication errors, affecting user experience and system security. SUMMARY

[0004] The embodiments of the present application provide a dynamic updating method, device, equipment and storage medium for biometric features to at least solve the technical problem of being difficult to accurately and flexibly update the feature library in related technologies.

[0005] According to an aspect of an embodiment of the present application, a dynamic updating method for biometric features is provided, comprising: collecting real-time authentication features of a user and extracting historical fusion features of the user; calculating a feature quality factor, a similarity factor, a time interval weight and a historical cumulative weight based on the real-time authentication features and the historical fusion features; training a fusion weight prediction model based on a multi-layer perceptron neural network model; inputting the feature quality factor, the similarity factor, the time interval weight and the historical cumulative weight into the fusion weight prediction model to output a dynamic fusion weight; fusing the real-time authentication features and the historical fusion features based on the dynamic fusion weight to obtain updated fusion features.

[0006] According to another aspect of an embodiment of the present application, a dynamic updating device for biometric features is also provided, comprising: a collection module configured to collect real-time authentication features of a user and extract historical fusion features of the user; The feature factor calculation module is used to calculate the feature quality factor, similarity factor, time interval weight, and historical cumulative weight based on the real-time authentication features and historical fusion features. The neural network prediction module is used to train a fusion weight prediction model based on a multilayer perceptron neural network model; the feature quality factor, similarity factor, time interval weight, and historical cumulative weight are input into the fusion weight prediction model, and the dynamic fusion weight is output. The dynamic fusion module is used to fuse the real-time authentication features and historical fusion features based on the dynamic fusion weight to obtain updated fusion features.

[0007] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-described method for dynamically updating biometric features through the computer program.

[0008] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer program, and the computer program is configured to execute the above-mentioned dynamic update method for biometrics at runtime.

[0009] The technical solutions provided in this application embodiment may include the following beneficial effects: This application proposes a method for dynamically updating a feature database, primarily focusing on optimizing the fusion weights. The aim is to improve the reliability and adaptability of biometric database updates. This application combines multiple factors such as feature quality, similarity, time interval, and historical cumulative weights, and dynamically adjusts the feature fusion weights through a nonlinear model, thereby optimizing the feature update process of the database.

[0010] Specifically, the effectiveness of feature quality is ensured by introducing a feature quality factor. Similarity and time interval weights are introduced to calculate the similarity between new and historical features, while the time interval weights are adjusted based on user behavior patterns. This makes the weight allocation more aligned with actual user behavior patterns. By introducing historical cumulative weights, historical information from multiple authentications is utilized to reduce the impact of early misidentification on subsequent feature updates. A multilayer perceptron model is used to dynamically calculate fusion weights by integrating various factors, capturing complex relationships between features, optimizing the feature fusion process, and improving the accuracy and adaptability of biometric database updates. This scheme is not only applicable to various biometric authentication systems but also effectively reduces the impact of misidentification, improving the system's robustness and reliability. Attached Figure Description

[0011] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of a method for dynamically updating biometric features according to an embodiment of this application; Figure 2 This is a flowchart of a method for dynamically updating biometric features according to an embodiment of this application; Figure 3 This is a schematic diagram of a biometric dynamic update device according to an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an optional electronic device according to an embodiment of this application. Detailed Implementation

[0012] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0013] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0014] Dynamic updates replenish the base database by combining newly acquired biometric features with historical fused features, thereby ensuring the reliability and accuracy of the feature database. However, some implementations of the update method have several shortcomings, necessitating a more optimized dynamic fusion strategy.

[0015] For example, during the authentication process, features from non-target users may be mistakenly introduced into the fusion, leading to feature contamination of the base database and affecting the reliability of the authentication results. Furthermore, dynamic fusion weights typically only consider simple factors (such as time interval and similarity), neglecting complex factors such as feature quality, user behavior, and the number of previous authentications, making them difficult to adapt to diverse scenarios. Historical information from multiple authentications is not effectively utilized, causing early errors to potentially continue to affect subsequent updates. Moreover, current quality assessments often rely solely on single factors such as image sharpness and lighting, failing to comprehensively consider issues such as occlusion, noise, and environmental interference.

[0016] This scheme comprehensively considers feature quality, similarity, time interval weight, and historical cumulative weight in the design of dynamic fusion weights. By combining these factors through a nonlinear model, the fusion weights are dynamically adjusted to optimize the reliability and adaptability of the base database update.

[0017] The method for dynamically updating biometric features according to embodiments of this application will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, the method mainly includes the following steps: S101 collects users' real-time authentication features and extracts users' historical fusion features.

[0018] In one implementation, the user's real-time authentication features (f) are first collected from the authentication device. These features can be various types of biometrics. These features are typically unique and stable, suitable for personal identification and verification. Examples include fingerprints, facial recognition features, iris scans, and voiceprints.

[0019] These real-time authentication features can be used individually or in combination to improve the accuracy and security of identity authentication. In practical applications, the choice of which feature to use depends on various factors, including the requirements of the authentication system, user convenience, cost, and privacy protection.

[0020] Furthermore, extract the user's historical fusion features. Extract the user's historical fusion features from the database, formed through gradual fusion and updating during multiple identity authentication processes. Simultaneously, acquire auxiliary data related to these features, such as timestamps and image quality.

[0021] Historical fusion features are comprehensive features formed by gradually fusing and updating the collected feature data during multiple identity authentication or feature collection processes. These historical features are usually a combination of features extracted in previous authentication processes and new authentication data, with the aim of improving the accuracy and robustness of the authentication system, especially in the face of changes in user biometrics, and maintaining high recognition accuracy.

[0022] S102 calculates feature quality factor, similarity factor, time interval weight, and historical cumulative weight based on real-time authentication features and historical fusion features.

[0023] In one implementation, a feature quality factor is calculated based on real-time authentication features. The feature quality factor is used to comprehensively evaluate feature quality.

[0024] Specifically, based on real-time authentication features, image sharpness factor, illumination factor, noise level, and occlusion ratio are calculated; the image sharpness factor, illumination factor, noise level, and occlusion ratio are input into a preset quality evaluation model, and the feature quality factor is output.

[0025] In one implementation, image sharpness factor, illumination factor, noise level, and occlusion ratio are calculated based on real-time authentication features.

[0026] Specifically, edge detection is performed based on real-time authentication features to obtain the image sharpness factor.

[0027] Authentication devices (such as cameras and fingerprint scanners) are used to capture users' biometric images in real time. The captured images undergo preprocessing, including grayscale conversion, noise removal, and contrast enhancement, to improve the accuracy of subsequent edge detection. Edge detection is then performed on the images to obtain an edge image. In the edge image, the intensity of edge pixels reflects the richness of image detail. The total intensity or number of edge pixels in the edge image is statistically analyzed and used as an indicator of image sharpness. For example, the more edge pixels there are, the richer the image detail, and the higher the sharpness factor.

[0028] Alternatively, an image sharpness factor can be calculated based on image contrast. By calculating the image's grayscale histogram, the brightest and darkest pixel values ​​are found, and their difference is calculated as the global contrast. This contrast value can then be normalized to a specific range (such as 0 to 1) as a reference index for the image sharpness factor.

[0029] Furthermore, a grayscale histogram is calculated based on real-time authentication features to obtain the illumination factor. The number of pixels for each grayscale value in the image is counted to obtain the grayscale histogram. The distribution of the histogram is analyzed. If the peaks of the histogram are concentrated in a certain area, it indicates uneven illumination; if the histogram is relatively flat, it indicates relatively uniform illumination.

[0030] Optionally, the image can be converted to a grayscale image, the brightness value of each pixel can be calculated, and then the variance of the brightness values ​​of the entire image can be calculated. The smaller the variance, the more uniform the illumination of the image, and the higher the LCF value; the larger the variance, the more uneven the illumination, and the lower the LCF value.

[0031] Furthermore, frequency domain analysis is performed based on real-time authentication features to obtain the noise level. The image is transformed from the spatial domain to the frequency domain to obtain a spectrogram. In the spectrogram, the low-frequency portion mainly contains smooth regions and overall structural information of the image, while the high-frequency portion mainly contains image details and noise information. The energy of the high-frequency region in the spectrogram is calculated (this can be obtained by operations such as summing the squares of the spectral values ​​of the high-frequency portion). The higher the high-frequency energy, the higher the noise level in the image. The high-frequency energy can be compared with the total energy of the image (the sum of the squares of all pixel values ​​in the spatial domain) to obtain a quantitative index of the noise level.

[0032] Furthermore, keypoint detection is performed based on real-time authentication features to obtain the occlusion ratio. Keypoint detection algorithms (such as SIFT, SURF, ORB, etc.) are used to detect keypoints in the image, obtaining the keypoint locations and descriptors. If there are two images (such as a target image and a reference image), the keypoints of the two images can be matched. The number of matched keypoints and the number of unmatched keypoints are counted. Unmatched keypoints may be due to occlusion or other reasons. The occlusion ratio can be estimated by the ratio of the number of unmatched keypoints to the total number of keypoints.

[0033] Furthermore, the image sharpness factor, illumination factor, noise level, and occlusion ratio are input into a preset quality evaluation model, and the feature quality factor Q is output.

[0034] The quality evaluation model can take the following form:

[0035] Wherein, IQF represents the image sharpness factor, LCF represents the illumination factor, Noise represents the noise level, and Occlusion represents the occlusion ratio.

[0036] The comprehensive evaluation method of the feature quality factor Q provides a holistic feature quality evaluation mechanism for biometric recognition systems by combining four dimensions: image sharpness factor, illumination factor, noise level, and occlusion ratio. This multi-dimensional evaluation approach can more accurately reflect the true quality of features, thereby improving the accuracy and robustness of the recognition system. Specifically, the image sharpness factor and illumination factor ensure image sharpness and illumination uniformity, which helps improve the accuracy of feature extraction; the evaluation of noise level helps to identify and eliminate noise interference, ensuring the purity of features; and the consideration of occlusion ratio helps to identify and handle occlusion problems in the image, reducing recognition errors. Combining these factors, the feature quality factor provides a reliable quality control means for biometric recognition, significantly improving the overall performance of the system.

[0037] In one implementation, a similarity factor is calculated based on real-time authentication features and historical fusion features, including: calculating the cosine similarity between the real-time authentication features and the historical fusion features; and determining the similarity factor based on the cosine similarity. The similarity factor S between the new feature and the historical features is calculated using the cosine similarity.

[0038] Indicates newly collected features, Indicating historical integration characteristics, , These are the norms (i.e., the modulus of the vectors) of the new features and the historical features, respectively.

[0039] In one implementation, a time interval weight is calculated based on real-time authentication features and historical fusion features. The time interval weight is calculated based on the time interval and the user behavior prediction time. The time interval weight is dynamically adjusted by combining the predicted time interval with user behavior patterns.

[0040] First, the time interval is determined based on the timestamps of real-time authentication features and historical fusion features.

[0041] Furthermore, historical authentication behavior data of users is collected, and historical authentication time features are extracted from the historical authentication behavior data; the historical authentication time features are input into the long short-term memory network model to obtain the predicted time of user behavior.

[0042] Specifically, the first step is to collect and organize the timestamp data of each user's authentication process. This data reflects the user's authentication behavior patterns at different points in time. Next, key temporal features are extracted from these timestamps, such as authentication intervals, frequency, and potential periodic patterns. Then, these temporal features are used as input to train a Long Short-Term Memory (LSTM) network model. Due to its advantages in processing time series data, the LSTM model can learn the temporal dependencies of user authentication behavior and predict the time when the user is likely to authenticate next.

[0043] Furthermore, a time weight generation model based on the exponential decay function is generated by inputting the time interval and the user behavior prediction time into the time weight generation model to obtain the time interval weight.

[0044] The time weight generation model is shown below:

[0045] in, Indicates time interval, Indicates the time frame for predicting user behavior. This represents a parameter sensitive to time intervals. As the time difference increases, the weight... The impact of historical data on the prediction result decreases exponentially; the further away the historical data is from the prediction time, the smaller its influence on the prediction result. Conversely, the smaller the time difference, the greater the weight. The closer it gets to 1.

[0046] and The smaller the difference, the greater the weight of the time interval. The closer the value is to 1, the greater the impact of historical data closer to the prediction time on the prediction result; that is, the more consistent (or habitual) the user behavior is, the greater the impact.

[0047] and The greater the difference, the greater the weight of the time interval. Exponential decay means that the further away the historical data is from the prediction time, the smaller the impact on the prediction result. In other words, the less regular (or habitual) the user behavior is, the smaller the impact.

[0048] This scheme adapts to user authentication behavior by dynamically adjusting the weight of time intervals, significantly improving the system's adaptability and prediction accuracy. When user behavior is close to the prediction time, the weight approaches 1, enhancing the impact of recent behavior on the prediction result and reflecting the regularity of user behavior. Conversely, when the behavior differs significantly from the prediction time, the weight decays exponentially, reducing the impact of abnormal or irregular behavior. This mechanism enables the system to flexibly respond to different users' authentication frequencies and habits, optimizing the user experience and enhancing the system's personalized service capabilities.

[0049] In one implementation, the historical cumulative weight is calculated based on real-time authentication features and historical fusion features, including: collecting the user's historical authentication data; determining the weight value of each authentication, the feature quality factor of each authentication, and the total number of authentications based on the historical authentication data; and inputting the weight value of each authentication, the feature quality factor of each authentication, and the total number of authentications into the cumulative weight generation model to obtain the historical cumulative weight.

[0050] By leveraging historical information from multiple authentications, the impact of early misidentifications on subsequent feature updates is reduced. The cumulative weight H is calculated based on the historical results of multiple authentications.

[0051] The weight value represents the i-th authentication result, usually a number between 0 and 1, indicating the importance or credibility of the authentication result. n is the total number of authentications. Calculate the cumulative effect of the first n-1 authentication results. This is done through the product term. The successive product is used to reflect the weakening effect of the earlier certification on the final weight. Used to reflect the weighted contribution of progressively enhanced subsequent certifications.

[0052] To further enhance the robustness of the cumulative weights, we can combine the feature quality of each authentication. This is used to correct the cumulative weights. Feature quality. It is an assessment of the effectiveness or reliability of authentication features, typically determined based on factors such as image quality and noise levels. Lower This might mean the quality of the authenticated features is poor, and therefore should not have an excessive impact on the cumulative weight. Adjusted cumulative weight. :

[0053] Wherein, the weight value of the i-th authentication The weight can be determined based on the reliability of the authentication result. For example, if the i-th authentication result is correct, it can be assigned a higher weight because this authentication provides reliable user characteristic information. If the authentication result is incorrect, it should be assigned a lower weight because this authentication may contain incorrect or inaccurate characteristic information.

[0054] By introducing historical cumulative weights and considering the results of multiple authentications, the impact of early misidentifications is reduced, thereby improving the stability and reliability of the base database features. Furthermore, to enhance the robustness of the cumulative weights, they are adjusted based on the feature quality of each authentication.

[0055] S103 is based on a multilayer perceptron neural network model, which trains a fusion weight prediction model. The model is input into the feature quality factor, similarity factor, time interval weight, and historical cumulative weight, and outputs dynamic fusion weights.

[0056] Specifically, the first step is to train a fusion weight prediction model.

[0057] First, a multidimensional dataset containing feature quality, similarity, time interval weights, and historical cumulative weights needs to be collected and preprocessed. Next, an MLP architecture is designed, including an input layer, one or more hidden layers, and an output layer. The hidden layers employ a non-linear activation function such as ReLU to enhance the model's non-linear expressive power, while the output layer uses the Sigmoid function to output fused weights between 0 and 1. Then, a loss function such as mean squared error is defined to quantify the difference between the predicted and actual weights, and an optimization algorithm such as Adam is selected for model training. During training, the model learns the complex mapping relationship between input features and fused weights, iteratively optimizing the network weights to ultimately predict fused weights to optimize the updating of the biometric database.

[0058] Furthermore, the feature quality factor, similarity factor, time interval weight, and historical cumulative weight are input into the fusion weight prediction model to output dynamic fusion weight.

[0059] Specifically, the input vector is obtained based on the feature quality factor, similarity factor, time interval weight, and historical cumulative weight; the input vector is linearly transformed using the first weight matrix and the first bias term, and then nonlinearly transformed using the ReLU activation function to obtain the output of the hidden layer; the output of the hidden layer is multiplied by the second weight matrix, the second bias term is added, and then the dynamic fusion weight is obtained by passing the Sigmoid activation function.

[0060] Specifically, input factors Multi-layer perceptron (MLP) fusion weight prediction model. Calculate dynamic fusion weights. :

[0061] Represents the weight matrix. The value represents the bias term. ReLU(x) = max(0,x) represents the non-linear activation function. The ReLU function outputs 0 for negative numbers and keeps positive numbers unchanged. It can perform non-linear transformation on the linear combination of inputs, allowing the model to capture more complex relationships. This represents the Sigmoid activation function, with its output range limited to [0,1]. Input factors include feature quality Q, similarity S, and time interval weights. and historical cumulative weight .

[0062] Optionally, integrating user contextual information into a Multilayer Perceptron (MLP) neural network model can significantly improve the accuracy of fusion weight prediction and the system's adaptability. First, user behavioral context, such as user location and authentication time, and environmental context, such as weather conditions and network status, are collected to more comprehensively evaluate authentication results. Necessary preprocessing, such as normalization and categorical variable encoding, is then performed. Next, the MLP model architecture is designed, extending the input layer to incorporate these new features and potentially adding hidden layers or neurons to accommodate more complex data processing needs. Subsequently, the model is trained using a dataset containing contextual information, selecting an appropriate loss function and optimization algorithm, and adjusting weights through backpropagation to reduce prediction errors. After evaluating model performance on a validation set, the model parameters are adjusted and optimized based on the results. Finally, the trained model is integrated and deployed into a biometric recognition system, enabling it to process contextual information in real time and predict fusion weights, continuously monitoring and adapting to changes in user behavior, thereby improving the accuracy of biometric database updates and the overall adaptability of the system.

[0063] By using a Multilayer Perceptron (MLP) model to dynamically calculate and fuse weights from various factors, complex nonlinear relationships and interactions between input features can be captured. The advantage of the MLP model lies in its multilayer structure, which allows it to learn high-level feature representations of the input data through nonlinear transformations of the hidden layers, thereby improving prediction accuracy.

[0064] S104 uses dynamic fusion weights to fuse real-time authentication features and historical fusion features to obtain updated fusion features.

[0065] Specifically, based on dynamic fusion weights Update fusion features:

[0066] Indicates the current fusion characteristics, Indicating historical integration characteristics, This indicates newly collected real-time authentication features. This indicates the dynamic fusion weights.

[0067] Save the updated fusion features to the base database. Record the current fusion weights and factor values ​​for later analysis.

[0068] To facilitate understanding of the methods provided in the embodiments of this application, the following description is in conjunction with the appendix. Figure 2 Further description. For example... Figure 2 The image illustrates a flowchart of a biometric recognition system used to update the base database features. The process begins with three external inputs: real-time authentication features, historical fusion features, and historical authentication behavior. First, factors such as image sharpness, illumination, noise, and occlusion are evaluated, and feature quality is calculated based on a quality assessment model. Then, a similarity module calculates the similarity between new and historical features, and a time interval calculation and LSTM behavior prediction module determine the time interval weights. Simultaneously, historical cumulative weights are generated through authentication frequency statistics. These weights are input into an MLP fusion weight model to dynamically calculate the fusion weights. Finally, a weighted fusion module combines the real-time authentication features and historical fusion features to update the base database features, and the results are output or stored. The entire process aims to improve the reliability and adaptability of biometric database updates.

[0069] The proposed dynamic fusion weight optimization method comprehensively considers multiple factors such as feature quality, similarity, time interval, and historical cumulative weights. Utilizing a multilayer perceptron (MLP) model, it dynamically adjusts the fusion weights, significantly improving the reliability and adaptability of biometric database updates. Feature quality assessment ensures feature effectiveness, the combination of similarity and time interval makes weight adjustments more aligned with user behavior, the introduction of historical cumulative weights reduces the impact of early false recognitions, and feature quality correction enhances robustness. Nonlinear weight calculation further improves the model's ability to capture complex relationships and optimizes the feature fusion process. Ultimately, this approach is not only applicable to various biometric authentication systems but also effectively reduces the impact of false recognitions, improves system robustness and reliability, provides solid technical support for the application of biometric recognition technology in a wider range of scenarios, and significantly improves user experience and overall system performance.

[0070] According to another aspect of the embodiments of this application, a biometric dynamic update apparatus for implementing the above-described biometric dynamic update method is also provided. For example... Figure 3 As shown, the device includes: The acquisition module 301 is used to acquire the user's real-time authentication features and extract the user's historical fusion features; The feature factor calculation module 302 is used to calculate the feature quality factor, similarity factor, time interval weight, and historical cumulative weight based on real-time authentication features and historical fusion features. The neural network prediction module 303 is used to train a fusion weight prediction model based on a multilayer perceptron neural network model; it inputs feature quality factor, similarity factor, time interval weight, and historical cumulative weight into the fusion weight prediction model and outputs dynamic fusion weight. The dynamic fusion module 304 is used to fuse real-time authentication features and historical fusion features based on dynamic fusion weights to obtain updated fusion features.

[0071] It should be noted that the above embodiments of the biometric dynamic update device are only illustrated by the division of the above functional modules when executing the biometric dynamic update method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the biometric dynamic update device and the biometric dynamic update method embodiments provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments, which will not be repeated here.

[0072] According to another aspect of the embodiments of this application, an electronic device corresponding to the dynamic updating method of biometrics provided in the foregoing embodiments is also provided, so as to execute the dynamic updating method of biometrics described above.

[0073] Please refer to Figure 4 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 4 As shown, the electronic device includes: a processor 400, a memory 401, a bus 402, and a communication interface 403. The processor 400, the communication interface 403, and the memory 401 are connected via the bus 402. The memory 401 stores a computer program that can run on the processor 400. When the processor 400 runs the computer program, it executes the dynamic updating method of biometrics provided in any of the foregoing embodiments of this application.

[0074] The memory 401 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 403 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0075] Bus 402 can be an ISA bus, PCI bus, or EISA bus, etc. Buses can be divided into address buses, data buses, control buses, etc. Memory 401 is used to store programs. After receiving execution instructions, processor 400 executes the program. The dynamic updating method for biometrics disclosed in any of the aforementioned embodiments of this application can be applied to processor 400, or implemented by processor 400.

[0076] The processor 400 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 400 or by instructions in software form. The processor 400 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 401. The processor 400 reads the information in memory 401 and, in conjunction with its hardware, completes the steps of the above method.

[0077] The electronic device provided in this application embodiment and the biometric dynamic update method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0078] According to another aspect of the present application, a computer-readable storage medium corresponding to the dynamic updating method of biometrics provided in the foregoing embodiments is also provided, wherein a computer program (i.e., a program product) is stored thereon, and when the computer program is run by a processor, it executes the dynamic updating method of biometrics provided in any of the foregoing embodiments.

[0079] It should be noted that examples of computer-readable storage media may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0080] The computer-readable storage medium provided in the above embodiments of this application and the dynamic updating method of biometrics provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0081] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0082] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for dynamically updating biometric features, characterized in that, include: Collect the user's real-time authentication features and extract the user's historical fusion features; Based on the real-time authentication features and historical fusion features, the feature quality factor, similarity factor, time interval weight, and historical cumulative weight are calculated. A fusion weight prediction model is trained based on a multilayer perceptron neural network model; the feature quality factor, similarity factor, time interval weight, and historical cumulative weight are input into the fusion weight prediction model to output dynamic fusion weights. Based on the dynamic fusion weight, the real-time authentication features and historical fusion features are fused to obtain the updated fusion features.

2. The method according to claim 1, characterized in that, Based on the real-time authentication features and historical fusion features, the time interval weight is calculated, including: The time interval is determined based on the timestamps of the real-time authentication features and the historical fusion features. Collect users' historical authentication behavior data, and extract historical authentication time features from the historical authentication behavior data; The historical authentication time features are input into the Long Short-Term Memory network model to obtain the user behavior prediction time; The time weight generation model is based on the exponential decay function. The time interval and the user behavior prediction time are input into the time weight generation model to obtain the time interval weight.

3. The method according to claim 1, characterized in that, Based on the real-time authentication features and historical fusion features, the historical cumulative weight is calculated, including: Collect users' historical authentication data; The weight value, feature quality factor, and total number of authentications for each authentication are determined based on the historical authentication data. The historical cumulative weights are obtained by inputting the weight value of each authentication, the feature quality factor of each authentication, and the total number of authentications into the cumulative weight generation model.

4. The method according to claim 1, characterized in that, Based on the real-time authentication features, the feature quality factor is calculated, including: Based on the real-time authentication features, calculate the image sharpness factor, illumination factor, noise level, and occlusion ratio; The image sharpness factor, illumination factor, noise level, and occlusion ratio are input into a preset quality evaluation model, and the feature quality factor is output.

5. The method according to claim 4, characterized in that, Based on the aforementioned real-time authentication features, the image sharpness factor, illumination factor, noise level, and occlusion ratio are calculated, including: Edge detection is performed based on the real-time authentication features to obtain the image sharpness factor; The illumination factor is obtained by calculating the grayscale histogram based on the real-time authentication features. The noise level is obtained by performing frequency domain analysis based on the real-time authentication features. The occlusion ratio is obtained by performing key point detection based on the real-time authentication features.

6. The method according to claim 1, characterized in that, Based on the real-time authentication features and historical fusion features, a similarity factor is calculated, including: Calculate the cosine similarity between the real-time authentication features and the historical fusion features; The similarity factor is determined based on the cosine similarity.

7. The method according to claim 1, characterized in that, The feature quality factor, similarity factor, time interval weight, and historical cumulative weight are input into the fusion weight prediction model, and the dynamic fusion weight is output, including: The input vector is obtained based on the aforementioned feature quality factor, similarity factor, time interval weight, and historical cumulative weight. The input vector is linearly transformed using the first weight matrix and the first bias term, and then nonlinearly transformed using the ReLU activation function to obtain the output of the hidden layer. The output of the hidden layer is multiplied by the second weight matrix, a second bias term is added, and then the dynamic fusion weights are obtained by passing the Sigmoid activation function.

8. A device for dynamically updating biometric features, characterized in that, include: The data acquisition module is used to collect the user's real-time authentication features and extract the user's historical fusion features; The feature factor calculation module is used to calculate the feature quality factor, similarity factor, time interval weight, and historical cumulative weight based on the real-time authentication features and historical fusion features. The neural network prediction module is used to train a fusion weight prediction model based on a multilayer perceptron neural network model; the feature quality factor, similarity factor, time interval weight, and historical cumulative weight are input into the fusion weight prediction model, and the dynamic fusion weight is output. The dynamic fusion module is used to fuse the real-time authentication features and historical fusion features based on the dynamic fusion weight to obtain updated fusion features.

9. An electronic device, characterized in that, It includes a processor and a memory storing program instructions, the processor being configured to perform the dynamic updating method of biometrics as described in any one of claims 1 to 7 when executing the program instructions.

10. A computer-readable medium, characterized in that, It stores computer-readable instructions that are executed by a processor to implement a method for dynamically updating biometric features as described in any one of claims 1 to 7.

Images