Method and system for verifying identity of sperm library based on comparison of multi-biological characteristic information

Abstract

The invention discloses a sperm library identity verification method and system based on multi-biological characteristic information comparison, belongs to the technical field of sperm library management, and solves the problem that the existing method adopts a single face biological characteristic to carry out face recognition, is easily influenced by environmental factors or individual states, and causes insufficient identity verification accuracy and adaptability. The system comprises an information acquisition end, an information database, an information comparison module and an identity verification module. According to the method, the sperm library knowledge graph is constructed based on the multi-biological characteristic information, so that the sperm library knowledge graph integrates various biological characteristic information, more comprehensive data support can be provided for modeling and information analysis of the information comparison model, and the multi-biological characteristic information is analyzed and verified through the information comparison model, so that the accuracy of the information comparison model is improved. Therefore, the accuracy and adaptability of identity verification are improved, the risk of misjudgment of the system is reduced, and the convenience and robustness of the system are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of sperm bank management, and in particular relates to a sperm bank identity authentication method and system based on multi-biometric information comparison. Background Art

[0002] A sperm bank refers to an institution that uses ultra-low temperature freezing technology to collect, test, preserve and provide sperm for the purpose of treating infertility and preventing genetic diseases. When managing a sperm bank, it is very necessary to authenticate the identities of personnel during the collection, testing and preservation process. The identity authentication system can automatically handle tasks such as person-ID comparison, three-dimensional face recognition, and personnel information entry for sperm donor volunteers or patients who preserve their own sperm, greatly improving work efficiency and accuracy.

[0003] Chinese patent CN111832535B discloses a face recognition method and device, which includes: obtaining an RGB image and a corresponding depth image for face recognition; selecting a target face from the RGB image; judging whether there is an interfering face in the RGB image based on the target face and the depth image; if not, performing face recognition based on the target face. However, the existing method uses a single facial biometric feature for face recognition, which is easily affected by environmental factors or individual status, resulting in insufficient identity authentication accuracy and adaptability. To address the above problems, we propose a sperm bank identity authentication method and system based on multi-biometric information comparison. Summary of the invention

[0004] The purpose of the present invention is to address the shortcomings of the prior art and provide a sperm bank identity authentication method and system based on multi-biometric information comparison, which solves the problem that the existing method uses a single facial biometric feature for face recognition, which is easily affected by environmental factors or individual status, resulting in insufficient identity authentication accuracy and adaptability.

[0005] Existing methods use a single facial biometric for face recognition, which is easily affected by environmental factors or individual status, resulting in insufficient identity authentication accuracy and adaptability. To address the above problems, we propose a sperm bank identity authentication method and system based on multi-biometric information comparison. The system consists of an information collection end, an information database, an information comparison module, and an identity authentication module. When the system is working, the information collection end first collects multi-source heterogeneous multi-biometric information in real time, and pre-processes the multi-biometric information based on the iterative extended Kalman filter algorithm. Then the information database constructs a sperm bank knowledge graph based on the multi-biometric information, and uses the modeling sample set to construct an information comparison model. The information comparison module uses the pre-processed multi-biometric information as input, executes the information comparison model, and outputs the information comparison result. Finally, the identity authentication module determines whether the user identity authentication is passed based on the information comparison result. In an embodiment of the present invention, a sperm bank knowledge graph is constructed based on multiple biometric information, so that the sperm bank knowledge graph integrates multiple biometric information, thereby providing more comprehensive data support for information comparison model modeling and information analysis, and verifying the multiple biometric information analysis through the information comparison model, thereby improving the accuracy and adaptability of identity authentication, reducing the risk of system misjudgment, and improving the convenience and robustness of the system.

[0006] The present invention is implemented in this way: a sperm bank identity authentication system based on multi-biometric information comparison, the sperm bank identity authentication system based on multi-biometric information comparison comprises:

[0007] The information collection end is used to collect multi-source heterogeneous multi-biometric information in real time, and pre-process the multi-biometric information based on the iterative extended Kalman filter algorithm, where the multi-biometric information includes facial feature information, fingerprint feature information, voiceprint feature information, and retinal feature information;

[0008] An information database is used to store the multi-biometric information collected by the information collection terminal, and to construct a sperm bank knowledge graph based on the multi-biometric information, to derive a modeling sample set based on the sperm bank knowledge graph, and to construct an information comparison model using the modeling sample set;

[0009] An information comparison module is used to load the preprocessed multi-biometric information from the sperm bank knowledge graph and load the information comparison model, take the preprocessed multi-biometric information as input, execute the information comparison model, and output the information comparison result;

[0010] The identity authentication module is used to load the information comparison results and determine whether the user identity authentication is passed based on the information comparison results.

[0011] Preferably, the information collection terminal includes:

[0012] An information input unit, used to input standard characteristic information of sperm bank-related personnel;

[0013] An information collection unit, used for distributed collection of multi-source heterogeneous multi-biometric information;

[0014] An information preprocessing unit, which preprocesses multi-biometric information based on an iterative extended Kalman filter algorithm;

[0015] The information uploading unit is connected to the information input unit, the information collection unit and the information preprocessing unit for uploading the standard feature information and the preprocessed multi-biometric feature information to the information database in real time.

[0016] Preferably, the method for preprocessing multi-biometric information based on iterative extended Kalman filter algorithm specifically includes:

[0017] Load multiple biometric information and process missing values ​​and outliers for the multiple biometric information;

[0018] Acquire multi-biometric information after missing values ​​and outliers are processed, preset the number of iterative filtering times, pre-filter the multi-biometric information based on the iterative extended Kalman filter algorithm, and obtain a pre-filter set, wherein when pre-filtering the multi-biometric information based on the iterative extended Kalman filter algorithm, use the posterior mean of the previous iteration as a new working point, and linearize the working point;

[0019] Loading the pre-filter set, performing reverse one-step smoothing on the pre-filter set, repeating the pre-filtering and reverse one-step smoothing processes based on a preset number of iterative filtering times, and outputting a reverse smoothing set;

[0020] Obtaining a reverse smooth set, performing dimensionality reduction processing on the reverse smooth set to obtain a smoothed reduced dimensionality set, and using the smoothed reduced dimensionality set as an output representation of the multi-biometric information after preprocessing;

[0021] Among them, the smoothed dimensionality reduction set is calculated by the following formula:

[0022]

[0023] Among them, f(x) represents the smoothed dimensionality reduction set, a(x), are the reverse smooth set and the variance of the reverse smooth set, q i is the weight parameter of the current biometric feature, and γ is the data dimension of the reverse smoothing set.

[0024] Preferably, the information database includes:

[0025] An information receiving unit, used for receiving pre-processed multi-biometric information;

[0026] A graph construction unit that combines multiple biological feature information to construct a sperm bank knowledge graph;

[0027] A model building unit, which exports a modeling sample set based on the sperm bank knowledge graph, uses the modeling sample set to build an information comparison model, and

[0028] The information storage unit is used to store multi-biometric information in the form of a sperm bank knowledge graph.

[0029] Preferably, the method for constructing a sperm bank knowledge graph by combining multiple biometric information specifically includes:

[0030] The Scrapy crawler framework is used to formulate regular expressions to crawl the knowledge graph related data, and the knowledge graph related data is preprocessed by data cleaning, Chinese word segmentation and data annotation, and the knowledge graph related data is saved to the information storage unit;

[0031] Define basic terms, entity attributes, entity relationships, and named entities in the sperm bank knowledge domain, create top-level entities of the sperm bank knowledge graph based on the basic terms, entity attributes, entity relationships, and named entities, and define the model layer of the sperm bank knowledge graph in combination with the top-level entities;

[0032] Load the top-level entities and pattern layers of the sperm bank knowledge graph, fill the knowledge graph associated data and multi-biometric information into the top-level entities and pattern layers, and complete the extraction of knowledge graph associated data.

[0033] Load the top-level entities and pattern layers filled with knowledge graph associated data and multi-biometric information, merge and fuse multi-biometric and multi-source knowledge data, and use the merged knowledge graph associated data and multi-biometric information as graph nodes. The merge and fuse multi-biometric and multi-source knowledge data are performed in the following ways: attribute alignment, entity alignment, and knowledge merging.

[0034] Among them, when the attributes are aligned, the attribute alignment formula is expressed as:

[0035]

[0036] In formula (2), sin(A,B) represents the similarity coefficient between entity A and entity B, |A∩B| represents the number of identical characters between entity A and entity B, and |A∪B| represents the number of common characters between entity A and entity B.

[0037] Load the graph nodes, top-level entities, and pattern layers, store the graph nodes, top-level entities, and pattern layers in the Neo4j graph database, complete the construction of the sperm bank knowledge graph, and store the sperm bank knowledge graph in the information database.

[0038] Preferably, the method of constructing an information comparison model using a modeling sample set specifically includes:

[0039] Obtain a modeling sample set, preprocess the modeling samples in the modeling sample set, use a minimum-maximum ratio to eliminate differences in sample dimensions and parameter units during modeling sample preprocessing, and divide the preprocessed modeling samples into a training set and a test set;

[0040] The preprocessing formula is expressed as:

[0041]

[0042] Among them, f(t) is the preprocessed modeling sample, t is the original modeling sample, and t max , t min are the maximum and minimum values ​​of the original modeling samples, respectively;

[0043] Taking BP neural network as the initial model of the information comparison model, the generative adversarial network and transfer learning strategy are introduced into the BP neural network to improve the initial model.

[0044] The facial feature information, fingerprint feature information, voiceprint feature information, and retinal feature information are used as the four input parameters of the BP neural network, and the comparison similarity is used as the output value of the BP neural network. The number of input units of the BP neural network is 4, the number of output units is 1, and the number of neurons in the hidden layer of the BP neural network is determined by the following formula:

[0045] S hid =2N input ±n(4)

[0046] Among them, S hid is the number of neurons in the hidden layer, N input Indicates the number of input units, n is the preset increase value;

[0047] Obtain the training set, preset the training error target, maximum number of training steps, and loss function of the initial model, use the Bayesian regularization algorithm based on the training set to iteratively train the initial model until convergence, and output the converged initial model;

[0048] Load the test set, use the test set as input, execute the initial model, output the test results, and determine whether the test results meet the preset accuracy threshold. If they meet the preset accuracy threshold, output the converged information comparison model.

[0049] Preferably, the method of executing the information comparison model using the pre-processed multi-biometric information as input specifically comprises:

[0050] Acquire the preprocessed multi-biometric information, extract the smoothed reduced dimension set from the multi-biometric information, optimize the smoothed reduced dimension set using a Bayesian regularization algorithm, and output the optimized result of the smoothed reduced dimension set;

[0051] Load the optimization result of the smooth dimension reduction set, perform frequency domain bandpass filtering on the optimized smooth dimension reduction set, and output the frequency domain filter set;

[0052] The spatial convolution layer in the BP neural network performs multi-layer dilated convolution on the frequency domain filter set, and the dilated convolution results are combined based on the fully connected layer to output the feature adjacency matrix.

[0053] Combined with the transfer learning strategy, different weights are assigned to different matrix vectors in the multi-scale adjacency matrix and learned to obtain the key point matrix, and the identity key points are determined based on the key point matrix;

[0054] Load identity key points, fuse the identity key points using the information comparison model, calculate the comparison similarity based on the Gaussian distribution function, and output the comparison similarity.

[0055] Preferably, the optimization result of the smoothed dimensionality reduction set is expressed as:

[0056]

[0057] in, represents the optimization result of the smooth dimensionality reduction set, L is the amount of data in the smooth dimensionality reduction set, q i is the weight parameter of the current biometric feature, α and β are regularization coefficients, θ(x) represents the transfer function of the hidden layer of the information comparison model, f(x), They are the smoothed reduced dimension set and the mean of the smoothed reduced dimension set respectively;

[0058] The comparative similarity is calculated by the following formula:

[0059]

[0060] sim(x,W x ) represents the comparison similarity, τ is the Gaussian distribution function convergence factor, A x is the expected value of the Gaussian distribution function, W x is the identity key point input vector, and υ is the number of Gaussian distribution function iterations.

[0061] On the other hand, the present invention also provides a sperm bank identity authentication method based on multi-biometric information comparison, the sperm bank identity authentication method based on multi-biometric information comparison specifically includes:

[0062] Collect multi-source heterogeneous multi-biometric information in real time, and pre-process the multi-biometric information based on the iterative extended Kalman filter algorithm;

[0063] Build a sperm bank knowledge graph based on multi-biometric information, derive a modeling sample set based on the sperm bank knowledge graph, and use the modeling sample set to build an information comparison model;

[0064] Load the preprocessed multi-biometric information from the sperm bank knowledge graph, and load the information comparison model, use the preprocessed multi-biometric information as input, execute the information comparison model, and output the information comparison result;

[0065] Load the information comparison result, and determine whether the user identity authentication is passed based on the information comparison result.

[0066] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0067] In an embodiment of the present invention, a sperm bank knowledge graph is constructed based on multiple biometric information, so that the sperm bank knowledge graph integrates multiple biometric information, thereby providing more comprehensive data support for information comparison model modeling and information analysis, and verifying the multiple biometric information analysis through the information comparison model, thereby improving the accuracy and adaptability of identity authentication, reducing the risk of system misjudgment, improving the convenience and robustness of the system, and overcoming the problem that a single biometric may be affected by environmental factors or individual status, resulting in insufficient accuracy and adaptability of identity authentication.

[0068] In the embodiment of the present invention, the multi-biometric information is pre-filtered based on the iterative extended Kalman filter algorithm, which can significantly reduce the noise and error in the multi-biometric information by optimizing the state estimation through multiple iterations. This improves the accuracy of the data and provides a more reliable basis for subsequent identity authentication.

[0069] In an embodiment of the present invention, a sperm bank knowledge graph is constructed by combining multiple biometric information, and a Scrapy crawler framework is used to formulate regular expressions when constructing the sperm bank knowledge graph. Through automated data crawling and processing processes, manual intervention can be reduced and the speed and accuracy of data processing can be improved. By defining top-level entities and model layers, data from different sources can be effectively integrated to ensure data consistency and integrity, thereby effectively integrating biometric information from different sources to form a unified knowledge network. This integration helps the information comparison model to access and utilize data more quickly.

[0070] In an embodiment of the present invention, the information comparison model uses a BP neural network as an initial model, and introduces a generative adversarial network and a transfer learning strategy, and adopts a Bayesian regularization algorithm to iteratively train the initial model, so that the existing relevant task knowledge can be used through transfer learning to improve the learning effect of new tasks, reduce the training time and data requirements of the information comparison model, and improve the accuracy of the information comparison model. The adversarial training mechanism in the generative adversarial network enables the information comparison model to maintain high stability and robustness in the face of various attacks and disturbances. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a structural schematic diagram of a sperm bank identity authentication system based on multi-biometric information comparison provided by the present invention.

[0072] Figure 2 The figure shows a schematic diagram of the implementation process of a method for preprocessing multi-biometric information based on an iterative extended Kalman filter algorithm.

[0073] Figure 3 The figure shows a schematic diagram of the implementation process of the method for constructing a sperm bank knowledge graph by combining multiple biometric information.

[0074] Figure 4 A schematic diagram of the implementation process of a method for constructing an information comparison model using a modeling sample set is shown.

[0075] Figure 5 The figure shows a schematic diagram of the implementation flow of the information comparison model method using preprocessed multi-biometric information as input.

[0076] Figure 6 The figure shows a schematic diagram of the implementation process of the sperm bank identity authentication method based on multi-biometric information comparison.

[0077] In the figure: 100-information acquisition terminal, 110-information entry unit, 120-information acquisition unit, 130-information preprocessing unit, 140-information upload unit, 200-information database, 210-information receiving unit, 220-map construction unit, 230-model construction unit, 240-information storage unit, 300-information comparison module, 400-identity authentication module. DETAILED DESCRIPTION

[0078] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of this application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0079] Existing methods use a single facial biometric feature for face recognition, which is easily affected by environmental factors or individual status, resulting in insufficient identity authentication accuracy and adaptability. To address the above problems, we propose a sperm bank identity authentication method and system based on multi-biometric information comparison. The system consists of an information acquisition terminal 100, an information database 200, an information comparison module 300 and an identity authentication module 400. When the system is working, the information acquisition terminal 100 first collects multi-source heterogeneous multi-biometric information in real time, and pre-processes the multi-biometric information based on an iterative extended Kalman filter algorithm. Then, the information database 200 constructs a sperm bank knowledge graph based on the multi-biometric information, and uses a modeling sample set to construct an information comparison model. The information comparison module 300 uses the pre-processed multi-biometric information as input, executes the information comparison model, and outputs the information comparison result. Finally, the identity authentication module 400 determines whether the user identity authentication is passed based on the information comparison result. In an embodiment of the present invention, a sperm bank knowledge graph is constructed based on multiple biometric information, so that the sperm bank knowledge graph integrates multiple biometric information, thereby providing more comprehensive data support for information comparison model modeling and information analysis, and verifying the multiple biometric information analysis through the information comparison model, thereby improving the accuracy and adaptability of identity authentication, reducing the risk of system misjudgment, improving the convenience and robustness of the system, and overcoming the problem that a single biometric may be affected by environmental factors or individual status, resulting in insufficient accuracy and adaptability of identity authentication.

[0080] The embodiment of the present invention provides a sperm bank identity authentication system based on multi-biometric information comparison. Figure 1 A schematic diagram of a sperm bank identity verification system based on multi-biometric information comparison is shown, wherein the sperm bank identity verification system based on multi-biometric information comparison specifically includes:

[0081] The information collection terminal 100 is used to collect multi-source heterogeneous multi-biometric information in real time and pre-process the multi-biometric information based on an iterative extended Kalman filter algorithm;

[0082] Among them, multi-biometric information includes but is not limited to facial feature information, fingerprint feature information, voiceprint feature information, and retinal feature information.

[0083] The information collection terminal 100 includes:

[0084] An information input unit 110 is used to input standard characteristic information of sperm bank-related personnel;

[0085] It should be noted that the information entry unit 110 can be a fingerprint recognition device, a facial recognition camera, a voiceprint recognition device, a pulse wave sensor, and an infrared detector, and the sperm bank-related personnel can include sperm donation volunteers, fertilized persons, sperm bank service personnel, and sperm bank management personnel.

[0086] The information collection unit 120 is used for collecting multi-source heterogeneous multi-biometric information in a distributed manner;

[0087] An information preprocessing unit 130 preprocesses the multi-biometric information based on an iterative extended Kalman filter algorithm;

[0088] The information uploading unit 140 is in communication with the information input unit 110 , the information collecting unit 120 , and the information preprocessing unit 130 , and is used to upload the standard feature information and the preprocessed multi-biometric feature information to the information database 200 in real time.

[0089] The information database 200 is used to store the multi-biometric information collected by the information collection terminal 100, and to construct a sperm bank knowledge graph based on the multi-biometric information, to derive a modeling sample set based on the sperm bank knowledge graph, and to construct an information comparison model using the modeling sample set;

[0090] The information comparison module 300 is used to load the pre-processed multi-biometric information from the sperm bank knowledge graph, and load the information comparison model, take the pre-processed multi-biometric information as input, execute the information comparison model, and output the information comparison result;

[0091] The identity authentication module 400 is used to load the information comparison result and determine whether the user identity authentication is passed based on the information comparison result.

[0092] In this embodiment, the information collection terminal 100, the information database 200, the information comparison module 300, and the identity authentication module 400 are connected by Bluetooth, 5G or DTU communication, and the information database 200 is connected to the information collection terminal 100 in a one-to-N manner.

[0093] In an embodiment of the present invention, a sperm bank knowledge graph is constructed based on multiple biometric information, so that the sperm bank knowledge graph integrates multiple biometric information, thereby providing more comprehensive data support for information comparison model modeling and information analysis, and verifying the multiple biometric information analysis through the information comparison model, thereby improving the accuracy and adaptability of identity authentication, reducing the risk of system misjudgment, improving the convenience and robustness of the system, and overcoming the problem that a single biometric may be affected by environmental factors or individual status, resulting in insufficient accuracy and adaptability of identity authentication.

[0094] The embodiment of the present invention provides a method for preprocessing multi-biometric information based on an iterative extended Kalman filter algorithm. Figure 2 The schematic diagram of the implementation process of the method for preprocessing multiple biometric information based on the iterative extended Kalman filter algorithm is shown. The method for preprocessing multiple biometric information based on the iterative extended Kalman filter algorithm specifically includes:

[0095] Step S101, loading multiple biometric information, and processing missing values ​​and abnormal values ​​of the multiple biometric information;

[0096] In this embodiment, the method of processing missing values ​​and outliers for multiple biometric information may be to delete missing values, fill missing values, multiple interpolation missing values, delete outliers, replace outliers, which are the processing methods for missing values.

[0097] Step S102, obtaining multi-biometric information after missing values ​​and outliers are processed, presetting the number of iterative filtering times, and pre-filtering the multi-biometric information based on the iterative extended Kalman filter algorithm to obtain a pre-filter set, wherein when pre-filtering the multi-biometric information based on the iterative extended Kalman filter algorithm, the posterior mean of the previous iteration is used as a new working point, and the working point is linearized;

[0098] In this embodiment, the preset number of iterative filtering times can be 3-8 times, and the multi-biometric information is pre-filtered based on the iterative extended Kalman filter algorithm. The iterative extended Kalman filter algorithm can significantly reduce the noise and error in the multi-biometric information by optimizing the state estimation through multiple iterations. This improves the accuracy of the data and provides a more reliable basis for subsequent identity authentication.

[0099] Step S103, loading the pre-filter set, performing a reverse one-step smoothing process on the pre-filter set, repeating the pre-filtering and reverse one-step smoothing process based on a preset number of iterative filtering times, and outputting a reverse smoothing set;

[0100] Step S104, obtaining a reverse smooth set, performing dimensionality reduction processing on the reverse smooth set to obtain a smoothed reduced dimensionality set, and using the smoothed reduced dimensionality set as the output representation of the pre-processed multi-biometric feature information;

[0101] Among them, the smoothed dimensionality reduction set is calculated by the following formula:

[0102]

[0103] Among them, f(x) represents the smoothed dimensionality reduction set, a(x), are the reverse smooth set and the variance of the reverse smooth set, q i is the weight parameter of the current biometric feature, γ is the data dimension of the reverse smoothing set, and in this embodiment, the data dimension can be 1-5.

[0104] In this embodiment, the iterative extended Kalman filter can more accurately approximate the real state of the nonlinear system through multiple iterations, and each iteration will correct and optimize the previous estimation result, thereby improving the accuracy of the overall data.

[0105] like Figure 1As shown, the information database 200 includes:

[0106] An information receiving unit 210, configured to receive pre-processed multi-biometric information;

[0107] A graph construction unit 220, which combines multiple biometric information to construct a sperm bank knowledge graph;

[0108] The model building unit 230 derives a modeling sample set based on the sperm bank knowledge graph, uses the modeling sample set to build an information comparison model, and

[0109] The information storage unit 240 is used to store multi-biometric information in the form of a sperm bank knowledge graph.

[0110] In this embodiment, the information receiving unit 210, the graph building unit 220, the model building unit 230, and the information storage unit 240 can be connected by a local area network or Bluetooth communication. The information database 200 can be a SQL Server database. The information database 200 uses MyBatis as a framework. The information receiving unit 210 can be a wireless communication device, a sensor, or a transponder receiving end.

[0111] The embodiment of the present invention provides a method for constructing a sperm bank knowledge graph by combining multiple biometric information. Figure 3 The schematic diagram of the implementation process of the method for constructing a sperm bank knowledge graph by combining multiple biometric information is shown. The method for constructing a sperm bank knowledge graph by combining multiple biometric information specifically includes:

[0112] Step S201, using the Scrapy crawler framework to formulate a regular expression, crawling the knowledge graph associated data, pre-processing the knowledge graph associated data by means of data cleaning, Chinese word segmentation and data annotation, and saving the knowledge graph associated data to the information storage unit 240;

[0113] It should be noted that the knowledge graph-related data includes but is not limited to the donor's age, name, family medical history, personal health information, medical examination information, and semen quality parameters. The advantage of using the Scrapy crawler framework to formulate regular expressions is that Scrapy is an efficient web crawler framework that can quickly extract required data from web pages. By using regular expressions, specific information in web pages can be matched and extracted more accurately, thereby improving the efficiency of data acquisition.

[0114] Step S202, defining basic terms, entity attributes, entity relationships and named entities in the sperm bank knowledge domain, creating top-level entities of the sperm bank knowledge graph based on the basic terms, entity attributes, entity relationships and named entities, and defining the pattern layer of the sperm bank knowledge graph in combination with the top-level entities;

[0115] In this embodiment, by defining the top-level entity and model layers, data from different sources can be effectively integrated to ensure data consistency and integrity. The use of graph databases makes multi-degree association queries much more efficient than traditional relational databases, which is especially important for sperm bank knowledge graphs that need to process a large number of entities and relationships.

[0116] Step S203, load the top-level entity and pattern layer of the sperm bank knowledge graph, fill the knowledge graph associated data and multi-biometric information into the top-level entity and pattern layer, and complete the extraction of knowledge graph associated data.

[0117] Step S204, loading the top-level entities and pattern layers filled with knowledge graph associated data and multi-biometric information, merging and fusing multi-biometric and multi-source knowledge data, and using the merged knowledge graph associated data and multi-biometric information as graph nodes, wherein the merging and fusing of multi-biometric and multi-source knowledge data is performed by attribute alignment, entity alignment, and knowledge merging;

[0118] Among them, when the attributes are aligned, the attribute alignment formula is expressed as:

[0119]

[0120] In formula (2), sin(A,B) represents the similarity coefficient between entity A and entity B, |A∩B| represents the number of identical characters between entity A and entity B, and |A∪B| represents the number of common characters between entity A and entity B.

[0121] Step S205, load the graph nodes, top-level entities, and pattern layers, store the graph nodes, top-level entities, and pattern layers in the Neo4j graph database, complete the construction of the sperm bank knowledge graph, and store the sperm bank knowledge graph in the information database 200.

[0122] In an embodiment of the present invention, a sperm bank knowledge graph is constructed by combining multiple biometric information, and a Scrapy crawler framework is used to formulate regular expressions when constructing the sperm bank knowledge graph. Through automated data crawling and processing processes, manual intervention can be reduced and the speed and accuracy of data processing can be improved. By defining top-level entities and model layers, data from different sources can be effectively integrated to ensure data consistency and integrity, thereby effectively integrating biometric information from different sources to form a unified knowledge network. This integration helps the information comparison model to access and utilize data more quickly.

[0123] The embodiment of the present invention provides a method for constructing an information comparison model using a modeling sample set. Figure 4 The following is a schematic diagram of the implementation process of a method for constructing an information comparison model using a modeling sample set. The method for constructing an information comparison model using a modeling sample set specifically includes:

[0124] Step S301, obtain a modeling sample set, preprocess the modeling samples in the modeling sample set, use the minimum-maximum ratio to eliminate the differences in sample dimensions and parameter units during modeling sample preprocessing, and divide the preprocessed modeling samples into a training set and a test set. The ratio of the training set to the test set can be 4:1. In this embodiment, 505 groups of multi-source heterogeneous modeling sample seat modeling sample sets are selected, and the modeling sample set is divided into a training set and a test set using a five-fold cross-validation method.

[0125] The preprocessing formula is expressed as:

[0126]

[0127] Among them, f(t) is the preprocessed modeling sample, t is the original modeling sample, and t max , t min are the maximum and minimum values ​​of the original modeling samples, respectively;

[0128] Step S302, using the BP neural network as the initial model of the information comparison model, introducing a generative adversarial network and a transfer learning strategy into the BP neural network to improve the initial model;

[0129] Step S303, using facial feature information, fingerprint feature information, voiceprint feature information, and retinal feature information as four input parameters of the BP neural network, and using the comparison similarity as the output value of the BP neural network. The number of input units of the BP neural network is 4, the number of output units is 1, and the number of neurons in the hidden layer of the BP neural network is determined by the following formula:

[0130] S hid =2N input ±n(4)

[0131] Among them, S hid is the number of neurons in the hidden layer, N input Indicates the number of input units, n is the preset increase value;

[0132] Step S304, obtaining a training set, presetting the training error target, maximum number of training steps, and loss function of the initial model, iteratively training the initial model using a Bayesian regularization algorithm based on the training set until convergence, and outputting a converged initial model;

[0133] In this embodiment, the training error target of the initial model can be set to 10 -5 , and the loss function can be the cosine similarity loss function, and the maximum number of training steps is 1000 steps.

[0134] Step S305, loading the test set, taking the test set as input, executing the initial model, and outputting the test results;

[0135] Step S306, determining whether the test result meets a preset accuracy threshold, where the accuracy threshold can be set to 0.9-0.95;

[0136] Step S307: if the preset accuracy threshold is met, output the converged information comparison model;

[0137] If it does not meet the preset accuracy threshold, return to step S304.

[0138] In an embodiment of the present invention, the information comparison model uses a BP neural network as an initial model, and introduces a generative adversarial network and a transfer learning strategy, and adopts a Bayesian regularization algorithm to iteratively train the initial model, so that the existing relevant task knowledge can be used through transfer learning to improve the learning effect of new tasks, reduce the training time and data requirements of the information comparison model, and improve the accuracy of the information comparison model. The adversarial training mechanism in the generative adversarial network enables the information comparison model to maintain high stability and robustness in the face of various attacks and disturbances.

[0139] The embodiment of the present invention provides a method for executing an information comparison model using pre-processed multi-biometric information as input. Figure 5 The schematic diagram of the implementation flow of the method for executing the information comparison model using the preprocessed multi-biometric information as input is shown. The method for executing the information comparison model using the preprocessed multi-biometric information as input specifically includes:

[0140] Step S401, obtaining the pre-processed multi-biometric information, extracting a smoothed reduced dimension set from the multi-biometric information, optimizing the smoothed reduced dimension set using a Bayesian regularization algorithm, and outputting an optimization result of the smoothed reduced dimension set;

[0141] In this embodiment, the optimization result of the smoothed dimension reduction set is expressed as:

[0142]

[0143] in, represents the optimization result of the smooth dimensionality reduction set, L is the amount of data in the smooth dimensionality reduction set, q i is the weight parameter of the current biometric feature, α and β are regularization coefficients. In this embodiment, the regularization coefficient is 0.1-0.5, θ(x) represents the transfer function of the hidden layer of the information comparison model, f(x), are the smoothed reduced dimension set and the mean of the smoothed reduced dimension set respectively.

[0144] Step S402, loading the optimization result of the smoothed dimensionality reduction set, performing frequency domain bandpass filtering on the optimized smoothed dimensionality reduction set, and outputting a frequency domain filtered set;

[0145] Step S403, the spatial convolution layer in the BP neural network performs multi-layer dilated convolution on the frequency domain filter set, performs feature combination on the dilated convolution result based on the fully connected layer, and outputs a feature adjacency matrix;

[0146] In an embodiment of the present invention, the spatial convolution layer performs multi-layer dilated convolution on the frequency domain filter set in the BP neural network, which can expand the receptive field of the frequency domain filter set and capture multi-scale features, and the feature combination of the dilated convolution results based on the fully connected layer can integrate global features and make classification decisions. The spatial convolution layer in the BP neural network performs multi-layer dilated convolution on the frequency domain filter set, and the feature combination of the dilated convolution results based on the fully connected layer effectively reduces the amount of parameters and calculations, while maintaining a larger receptive field.

[0147] Step S404, assigning different weights to different matrix vectors in the multi-scale adjacency matrix in combination with a transfer learning strategy and performing learning, obtaining a key point matrix, and determining identity key points based on the key point matrix;

[0148] Step S405 , loading identity key points, fusing the identity key points with the information comparison model, calculating the comparison similarity based on the Gaussian distribution function, and outputting the comparison similarity.

[0149] In this embodiment, the comparison similarity is calculated by the following formula:

[0150]

[0151] sim(x,W x ) represents the comparison similarity, τ is the Gaussian distribution function convergence factor, A x is the expected value of the Gaussian distribution function, W x is the identity key point input vector, and υ is the number of Gaussian distribution function iterations.

[0152] In this embodiment, the information comparison model fuses the identity key points. When calculating the comparison similarity based on the Gaussian distribution function, the Gaussian distribution function can capture subtle differences in the data, so that even small feature changes can be accurately identified. By fusing different identity key points, the system can evaluate identity information from multiple dimensions, reducing the risk of single feature failure.

[0153] The embodiment of the present invention provides a sperm bank identity verification method based on multi-biometric information comparison. Figure 6 The present invention shows a schematic diagram of the implementation process of a sperm bank identity verification method based on multi-biometric information comparison, wherein the sperm bank identity verification method based on multi-biometric information comparison specifically includes:

[0154] Step S10, collecting multi-source heterogeneous multi-biometric information in real time, and pre-processing the multi-biometric information based on an iterative extended Kalman filter algorithm;

[0155] Step S20, constructing a sperm bank knowledge graph based on the multi-biometric information, deriving a modeling sample set based on the sperm bank knowledge graph, and constructing an information comparison model using the modeling sample set;

[0156] Step S30, loading the preprocessed multi-biometric information from the sperm bank knowledge graph, and loading the information comparison model, taking the preprocessed multi-biometric information as input, executing the information comparison model, and outputting the information comparison result;

[0157] Step S40, loading the information comparison result, and judging whether the user identity verification is passed based on the information comparison result.

[0158] In summary, the present invention provides a sperm bank identity authentication method and system based on multi-biometric information comparison. In an embodiment of the present invention, a sperm bank knowledge graph is constructed based on multi-biometric information, so that the sperm bank knowledge graph integrates a variety of biometric information, thereby providing more comprehensive data support for information comparison model modeling and information analysis, and verifying multi-biometric information analysis through the information comparison model, thereby improving the accuracy and adaptability of identity authentication, reducing the risk of system misjudgment, and improving the convenience and robustness of the system, overcoming the problem that a single biometric may be affected by environmental factors or individual status, resulting in insufficient accuracy and adaptability of identity authentication.

[0159] It should be noted that, for the above-mentioned embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described order of actions, because according to the present invention, some steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0160] In the several embodiments provided in the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the above-mentioned units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be in the form of telecommunication or other forms.

[0161] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on these embodiments, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in this field can still combine, add, delete or make other adjustments to the features in the various embodiments of the present invention according to the circumstances without conflict, without making creative work, so as to obtain different other technical solutions that do not deviate from the concept of the present invention in essence, and these technical solutions also belong to the scope of protection of the present invention.

Claims

1. A sperm bank identity verification system based on multi-biometric information comparison, characterized by: The sperm bank identity authentication system based on multi-biometric information comparison includes: The information collection end is used to collect multi-source heterogeneous multi-biometric information in real time, and pre-process the multi-biometric information based on the iterative extended Kalman filter algorithm, where the multi-biometric information includes facial feature information, fingerprint feature information, voiceprint feature information, and retinal feature information; An information database is used to store the multi-biometric information collected by the information collection terminal, and to construct a sperm bank knowledge graph based on the multi-biometric information, to derive a modeling sample set based on the sperm bank knowledge graph, and to construct an information comparison model using the modeling sample set; An information comparison module is used to load the preprocessed multi-biometric information from the sperm bank knowledge graph and load the information comparison model, take the preprocessed multi-biometric information as input, execute the information comparison model, and output the information comparison result; The identity authentication module is used to load the information comparison results and determine whether the user identity authentication is passed based on the information comparison results.

2. The sperm bank identity authentication system based on multi-biometric information comparison as claimed in claim 1, characterized in that: The information collection terminal includes: An information input unit, used to input standard characteristic information of sperm bank-related personnel; An information collection unit, used for distributed collection of multi-source heterogeneous multi-biometric information; An information preprocessing unit, which preprocesses multi-biometric information based on an iterative extended Kalman filter algorithm; The information uploading unit is connected to the information input unit, the information collection unit and the information preprocessing unit for uploading the standard feature information and the preprocessed multi-biometric feature information to the information database in real time.

3. The sperm bank identity authentication system based on multi-biometric information comparison as claimed in claim 2, characterized in that: The method for preprocessing multi-biometric information based on iterative extended Kalman filter algorithm specifically includes: Load multiple biometric information and process missing values ​​and outliers for the multiple biometric information; Acquire multi-biometric information after missing values ​​and outliers are processed, preset the number of iterative filtering times, pre-filter the multi-biometric information based on the iterative extended Kalman filter algorithm, and obtain a pre-filter set, wherein when pre-filtering the multi-biometric information based on the iterative extended Kalman filter algorithm, use the posterior mean of the previous iteration as a new working point, and linearize the working point; Loading the pre-filter set, performing reverse one-step smoothing on the pre-filter set, repeating the pre-filtering and reverse one-step smoothing processes based on a preset number of iterative filtering times, and outputting a reverse smoothing set; Obtaining a reverse smooth set, performing dimensionality reduction processing on the reverse smooth set to obtain a smoothed reduced dimensionality set, and using the smoothed reduced dimensionality set as an output representation of the multi-biometric information after preprocessing; Among them, the smoothed dimensionality reduction set is calculated by the following formula: Among them, f(x) represents the smoothed dimensionality reduction set, a(x), are the reverse smooth set and the variance of the reverse smooth set, q i is the weight parameter of the current biometric feature, and γ is the data dimension of the reverse smoothing set.

4. The sperm bank identity authentication system based on multi-biometric information comparison as claimed in claim 1, characterized in that: The information database includes: An information receiving unit, used for receiving pre-processed multi-biometric information; A graph construction unit that combines multiple biological feature information to construct a sperm bank knowledge graph; A model building unit, which exports a modeling sample set based on the sperm bank knowledge graph, uses the modeling sample set to build an information comparison model, and The information storage unit is used to store multi-biometric information in the form of a sperm bank knowledge graph.

5. The sperm bank identity authentication system based on multi-biometric information comparison as claimed in claim 4, characterized in that: The method for constructing a sperm bank knowledge graph by combining multiple biometric information specifically includes: The Scrapy crawler framework is used to formulate regular expressions to crawl the knowledge graph related data, and the knowledge graph related data is preprocessed by data cleaning, Chinese word segmentation and data annotation, and the knowledge graph related data is saved to the information storage unit; Define basic terms, entity attributes, entity relationships, and named entities in the sperm bank knowledge domain, create top-level entities of the sperm bank knowledge graph based on the basic terms, entity attributes, entity relationships, and named entities, and define the model layer of the sperm bank knowledge graph in combination with the top-level entities; Load the top-level entities and pattern layers of the sperm bank knowledge graph, fill the knowledge graph associated data and multi-biometric information into the top-level entities and pattern layers, and complete the extraction of knowledge graph associated data.

6. The sperm bank identity verification system based on multi-biometric information comparison as claimed in claim 5, characterized in that: The method for constructing a sperm bank knowledge graph by combining multiple biometric information specifically includes: Load the top-level entities and pattern layers filled with knowledge graph associated data and multi-biometric information, merge and fuse multi-biometric and multi-source knowledge data, and use the merged knowledge graph associated data and multi-biometric information as graph nodes. The merge and fuse multi-biometric and multi-source knowledge data are performed in the following ways: attribute alignment, entity alignment, and knowledge merging. Among them, when the attributes are aligned, the attribute alignment formula is expressed as: In formula (2), sin(A,B) represents the similarity coefficient between entity A and entity B, |A∩B| represents the number of identical characters between entity A and entity B, and |A∪B| represents the number of common characters between entity A and entity B. Load the graph nodes, top-level entities, and pattern layers, store the graph nodes, top-level entities, and pattern layers in the Neo4j graph database, complete the construction of the sperm bank knowledge graph, and store the sperm bank knowledge graph in the information database.

7. The sperm bank identity authentication system based on multi-biometric information comparison as claimed in claim 4, characterized in that: The method of using a modeling sample set to construct an information comparison model specifically includes: Obtain a modeling sample set, preprocess the modeling samples in the modeling sample set, use a minimum-maximum ratio to eliminate differences in sample dimensions and parameter units during modeling sample preprocessing, and divide the preprocessed modeling samples into a training set and a test set; The preprocessing formula is expressed as: Among them, f(t) is the preprocessed modeling sample, t is the original modeling sample, and t max , t min are the maximum and minimum values ​​of the original modeling samples, respectively; Taking BP neural network as the initial model of the information comparison model, the generative adversarial network and transfer learning strategy are introduced into the BP neural network to improve the initial model. The facial feature information, fingerprint feature information, voiceprint feature information, and retinal feature information are used as the four input parameters of the BP neural network, and the comparison similarity is used as the output value of the BP neural network. The number of input units of the BP neural network is 4, the number of output units is 1, and the number of neurons in the hidden layer of the BP neural network is determined by the following formula: <h2 style=";text-align:left;direction:ltr">S<h2 style=";text-align:left;direction:ltr"> hid <h2 style=";text-align:left;direction:ltr"> <2N<h2 style=";text-align:left;direction:ltr"> input <h2 style=";text-align:left;direction:ltr"> ±n(4) Among them, S hid is the number of neurons in the hidden layer, N input Indicates the number of input units, n is the preset increase value; Obtain the training set, preset the training error target, maximum number of training steps, and loss function of the initial model, use the Bayesian regularization algorithm based on the training set to iteratively train the initial model until convergence, and output the converged initial model; Load the test set, use the test set as input, execute the initial model, output the test results, and determine whether the test results meet the preset accuracy threshold. If they meet the preset accuracy threshold, output the converged information comparison model.

8. The sperm bank identity authentication system based on multi-biometric information comparison as claimed in claim 7, characterized in that: The method of executing the information comparison model using the pre-processed multi-biometric information as input specifically includes: Acquire the preprocessed multi-biometric information, extract the smoothed reduced dimension set from the multi-biometric information, optimize the smoothed reduced dimension set using a Bayesian regularization algorithm, and output the optimized result of the smoothed reduced dimension set; Load the optimization result of the smooth dimension reduction set, perform frequency domain bandpass filtering on the optimized smooth dimension reduction set, and output the frequency domain filter set; The spatial convolution layer in the BP neural network performs multi-layer dilated convolution on the frequency domain filter set, and the dilated convolution results are combined based on the fully connected layer to output the feature adjacency matrix. Combined with the transfer learning strategy, different weights are assigned to different matrix vectors in the multi-scale adjacency matrix and learned to obtain the key point matrix, and the identity key points are determined based on the key point matrix; Load identity key points, fuse the identity key points using the information comparison model, calculate the comparison similarity based on the Gaussian distribution function, and output the comparison similarity.

9. The sperm bank identity authentication system based on multi-biometric information comparison as claimed in claim 8, characterized in that: The optimization result of the smooth dimension reduction set is expressed as: in, represents the optimization result of the smooth dimensionality reduction set, L is the amount of data in the smooth dimensionality reduction set, q i is the weight parameter of the current biometric feature, α and β are regularization coefficients, θ(x) represents the transfer function of the hidden layer of the information comparison model, f(x), They are the smoothed reduced dimension set and the mean of the smoothed reduced dimension set respectively; The comparative similarity is calculated by the following formula: sim(x,W x ) represents the comparison similarity, τ is the Gaussian distribution function convergence factor, A x is the expected value of the Gaussian distribution function, W x is the identity key point input vector, and υ is the number of Gaussian distribution function iterations.

10. A sperm bank identity authentication method based on multi-biometric information comparison, implemented by a sperm bank identity authentication system based on multi-biometric information comparison as claimed in any one of claims 1 to 9, characterized in that: The sperm bank identity authentication method based on multi-biometric information comparison specifically includes: Collect multi-source heterogeneous multi-biometric information in real time, and pre-process the multi-biometric information based on the iterative extended Kalman filter algorithm; Build a sperm bank knowledge graph based on multi-biometric information, derive a modeling sample set based on the sperm bank knowledge graph, and use the modeling sample set to build an information comparison model; Load the preprocessed multi-biometric information from the sperm bank knowledge graph, and load the information comparison model, use the preprocessed multi-biometric information as input, execute the information comparison model, and output the information comparison result; Load the information comparison result, and determine whether the user identity authentication is passed based on the information comparison result.

Images