Adaptive face recognition and verification method based on 5G communication network

Abstract

The present invention relates to the field of wireless communications, and in particular to an adaptive face recognition and verification method based on a 5G communication network. First, a dynamic face network model is constructed; then, a data compression algorithm based on the topological structure of a complex network model and chaotic dynamics is introduced, a complex network model and a complex network chaotic coding function are constructed, and a chaotic mapping function based on the Lorenz attractor dynamics equation is defined; finally, a transmission optimization protocol based on multidimensional space topology and adaptive spectrum allocation is introduced, a multidimensional space topology model is constructed, and an adaptive spectrum allocation method is proposed based on the output results of the multidimensional space topology model. The problem of serious network delay often encountered in the prior art in mobile communication networks is solved; there is a lack of effective data compression and transmission optimization technology; there is a lack of sufficient adaptability; and when processing complex models and large amounts of data, the computational efficiency is low and the energy consumption is high.

Description

Technical Field

[0001] The present invention relates to the field of wireless communications, and in particular to an adaptive face recognition and verification method based on a 5G communication network. Background Art

[0002] In the mobile communication network environment, the high bandwidth and low latency characteristics of 5G communication networks provide an ideal foundation for real-time face recognition. This means that in various scenarios, such as public safety, smart cities, and industrial automation, devices can quickly acquire, transmit, and process face image data to support real-time identity authentication and monitoring. In the mobile communication network environment, the wide coverage and large capacity characteristics of 5G communication networks mean that face recognition solutions can be deployed on a wider range. This is particularly important for large-scale events, traffic management, and telemedicine, which can better serve people's needs.

[0003] Adaptive face recognition for 5G communication networks helps improve security and privacy protection. By using encryption and authentication features, users' identities and data can be better protected. This is critical for financial transactions, personal privacy, and network security.

[0004] my country's patent application number: CN202110778238.7, publication date: 2021.8.31, discloses a face recognition device based on 5G architecture, including a face recognition public terminal, a 5G network transmission component, a high-performance cloud computing platform and a fast cloud computing platform. The face recognition public terminal is used to collect face data and encapsulate and process the data; the face recognition public terminal interacts with the high-performance cloud computing platform and the fast cloud computing platform through the 5G network transmission component. The face recognition device based on the 5G architecture simulates a dynamic 3D stereo face model based on the information collected by the information collection component, so as to facilitate the acquisition of a collection of simulated facial photos, and compares it with a collection of real facial photos to prevent users from taking photos for face recognition, which has an anti-counterfeiting effect. The face recognition public terminal requests query information, and takes precedence over sending it to the fast cloud computing platform with the same number. In conjunction with the 5G communication network, the search range can be narrowed and the target information can be quickly found.

[0005] However, the above-mentioned technologies have at least the following technical problems: the existing technologies often encounter serious network delay problems in mobile communication networks, affecting user experience and the real-time response capability of the system; due to the lack of effective data compression and transmission optimization technology, the data transmission efficiency is low and cannot meet the transmission requirements of large data volumes and high speeds; the existing models often lack sufficient adaptability and cannot dynamically adjust model parameters and weights according to different learning scenarios, resulting in unstable performance of the models in different scenarios and unable to meet diversified and personalized needs; when processing complex models and large data volumes, the computing efficiency is low and the energy consumption is high, which is particularly obvious on mobile devices, limiting its application and development in mobile communication networks. Summary of the invention

[0006] The embodiment of the present application solves the serious network delay problem that the prior art often encounters in mobile communication networks, affecting user experience and the real-time response capability of the system by providing an adaptive face recognition and verification method based on 5G communication networks; due to the lack of effective data compression and transmission optimization technology, the data transmission efficiency is low and cannot meet the transmission requirements of large data volumes and high speeds; the existing models often lack sufficient adaptability and cannot dynamically adjust model parameters and weights according to different learning scenarios, resulting in unstable performance of the model in different scenarios and unable to meet the needs of diversification and personalization; when processing complex models and large data volumes, the computational efficiency is low and the energy consumption is high, which is particularly obvious on mobile devices, limiting its application and development in mobile communication networks. The adaptive face recognition and verification method based on 5G communication networks is implemented, and through deep learning models and optimization algorithms, the balance between real-time and accuracy, network delay, data transmission efficiency and model adaptability are effectively solved, thereby improving the performance and application scope of face recognition and verification.

[0007] This application provides an adaptive face recognition and verification method based on 5G communication network, which specifically includes the following technical solutions:

[0008] The adaptive face recognition and verification method based on 5G communication network includes the following steps:

[0009] S100: collecting facial image data through a mobile communication device, building a dynamic facial network model, inputting the facial image data into the dynamic facial network model, and outputting a category label;

[0010] S200: Introduce the data compression algorithm based on the topological structure of complex network model and chaotic dynamics, construct complex network model and complex network chaos coding function, and define the chaos mapping function based on the Lorenz attractor dynamics equation;

[0011] S300: Introduce a transmission optimization protocol based on multi-dimensional space topology and adaptive spectrum allocation, construct a multi-dimensional space topology model, and propose an adaptive spectrum allocation method based on the output results of the multi-dimensional space topology model.

[0012] Preferably, the S100 specifically includes:

[0013] The dynamic face network model includes: an adaptive feature extraction layer, a dynamic weight adjustment layer, a multi-dimensional optimization output layer, a fully connected layer and a normalization layer.

[0014] Preferably, the S100 further includes:

[0015] The adaptive feature extraction layer extracts facial features based on facial feature distribution; the dynamic weight adjustment layer adopts a gradient-based weight adjustment mechanism and introduces dynamic adjustment items to adjust the weights of the facial features extracted by the adaptive feature extraction layer.

[0016] Preferably, the S100 further includes:

[0017] The multi-dimensional optimized output layer introduces a threshold-based output optimization mechanism, and combines two or more inputs and weight matrices to optimize the output through nonlinear activation functions and quadratic terms.

[0018] Preferably, the S100 further includes:

[0019] The initial scores of the current face image under different categories are obtained through the fully connected layer; the predicted probability of the current face image for each category is calculated through the normalization layer, and the category of the current face image is divided according to the predicted probability.

[0020] Preferably, the S200 specifically includes:

[0021] The topological structure and weight matrix of the complex network model are introduced, the complex network chaotic coding function is defined, the optimization algorithm is used to solve the parameter matrix and weight matrix of the complex network chaotic coding, and the loss function is optimized.

[0022] Preferably, the S300 specifically includes:

[0023] A multi-dimensional space topology model is constructed, and network state variables and data flow parameters are introduced; the multi-dimensional space topology model dynamically adjusts network resources according to the real-time state of the network and data flow.

[0024] Preferably, the S300 further includes:

[0025] According to the output results of the multidimensional space topology model, an adaptive spectrum allocation method is proposed to adjust the allocation of spectrum resources in real time; spectrum resource allocation parameters are introduced to define a spectrum resource allocation matrix, which is constrained by the user's spectrum resource upper limit; an optimization objective function is defined and the optimization objective function is maximized.

[0026] Beneficial effects:

[0027] The multiple technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0028] 1. By constructing an authentication method based on a dynamic face network model, the real-time and high-precision face recognition and verification are ensured in a complex mobile communication network environment, especially in a 5G network environment, ensuring efficient data transmission and processing; the design of the adaptive feature extraction layer and the dynamic weight adjustment layer enables the model to extract facial features more accurately and quickly adapt to different learning scenarios, thereby accelerating the convergence of the model while maintaining accuracy; through multi-dimensional optimization of the output layer, multiple inputs and weight matrices are combined to ensure that high recognition accuracy can be maintained even in an unstable network environment, while reducing network latency;

[0029] 2. The adoption of a transmission optimization protocol based on multi-dimensional space topology and adaptive spectrum allocation has significantly reduced network latency and significantly improved data transmission reliability, ensuring rapid data transmission in mobile communication networks;

[0030] 3. By introducing a data compression algorithm based on complex network theory and chaotic dynamics, efficient nonlinear compression of facial image data is achieved, which greatly reduces network delays during data transmission, ensures rapid data transmission, and further optimizes network performance.

[0031] 4. The technical solution of this application can effectively solve the serious network delay problem often encountered in the existing technology in the mobile communication network, which affects the user experience and the real-time response capability of the system; due to the lack of effective data compression and transmission optimization technology, the data transmission efficiency is low and cannot meet the transmission requirements of large data volume and high speed; the existing models often lack sufficient adaptability and cannot dynamically adjust the model parameters and weights according to different learning scenarios, resulting in unstable performance of the model in different scenarios and unable to meet the diversified and personalized needs; when processing complex models and large data volumes, the computing efficiency is low and the energy consumption is high, which is particularly obvious on mobile devices, limiting its application and development in mobile communication networks. An adaptive face recognition and verification method based on 5G communication network is realized. Through deep learning models and optimization algorithms, the balance between real-time and accuracy, network delay, data transmission efficiency and model adaptability are effectively solved, and the performance and application scope of face recognition and verification are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a flow chart of the adaptive face recognition and verification method based on 5G communication network described in this application;

[0033] Figure 2 This is a structural diagram of the dynamic face network model described in this application. DETAILED DESCRIPTION

[0034] The embodiments of the present application provide an adaptive face recognition and verification method based on a 5G communication network, thereby solving the problem that the prior art often encounters serious network delay problems in mobile communication networks, which affects user experience and the real-time response capability of the system; due to the lack of effective data compression and transmission optimization technology, the data transmission efficiency is low and cannot meet the transmission requirements of large data volumes and high speeds; the existing models often lack sufficient adaptability and cannot dynamically adjust model parameters and weights according to different learning scenarios, resulting in unstable performance of the model in different scenarios and unable to meet diversified and personalized needs; when processing complex models and large data volumes, the computing efficiency is low and the energy consumption is high, which is particularly obvious on mobile devices, limiting its application and development in mobile communication networks.

[0035] The technical solution in the embodiment of the present application is to solve the above problems, and the overall idea is as follows:

[0036] By constructing an authentication method based on the dynamic face network model (ADFN), the real-time and high-precision face recognition and verification are ensured in a complex mobile communication network environment, especially in a mobile network environment, ensuring efficient data transmission and processing; the design of the adaptive feature extraction layer and the dynamic weight adjustment layer enables the model to extract facial features more accurately and quickly adapt to different learning scenarios, thereby accelerating the convergence of the model while maintaining accuracy; through the multi-dimensional optimization of the output layer, multiple inputs and weight matrices are combined to ensure that high recognition accuracy can be maintained even in an unstable network environment, while reducing network latency; the use of a transmission optimization protocol based on multi-dimensional space topology and adaptive spectrum allocation has achieved a significant reduction in network latency and a significant improvement in the reliability of data transmission, ensuring rapid data transmission in mobile communication networks; by introducing a data compression algorithm based on complex network theory and chaotic dynamics, efficient nonlinear compression of face image data is achieved, which greatly reduces network latency during data transmission, ensures rapid data transmission, and further optimizes network performance.

[0037] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0038] See attached Figure 1 The adaptive face recognition and verification method based on 5G communication network described in this application comprises the following steps:

[0039] S100: collecting facial image data through a mobile communication device, building a dynamic facial network model, inputting the facial image data into the dynamic facial network model, and outputting a category label;

[0040] In order to ensure the real-time and high-precision of face recognition and verification in 5G communication networks, an authentication method based on the dynamic face network model (ADFN) is constructed. The ADFN model adopts a structure and algorithm to solve the balance between real-time and accuracy, the optimization of network delay and data transmission, and the optimization of algorithm calculation efficiency and energy consumption in a complex mobile communication network environment.

[0041] For details, please refer to the attached Figure 2 , the ADFN model consists of five main parts: adaptive feature extraction layer, dynamic weight adjustment layer, multi-dimensional optimization output layer, fully connected layer and normalization layer. This model takes into account the data transmission and processing issues in the mobile network environment. Face image data is transmitted in the mobile communication network and input into the ADEN model. The face image data usually comes from various devices in the mobile communication network, such as smart phones, tablets, etc., and contains single or multiple faces. After processing and optimization by the ADEN model, it is finally quickly determined whether the verification is passed in the communication network.

[0042] In order to enhance the robustness of the model in the 5G communication environment, an adaptive feature extraction layer is designed to extract facial features more accurately. Based on the in-depth understanding of the distribution of facial features and the demand for model expression capabilities, the verification accuracy of the model is improved. Feature extraction is performed using the following formula:

[0043]

[0044] Among them, F(x) is the eigenvalue, w i is the weight, x i are input features, b is the bias term, n is the number of features, i∈[1,n], β and γ are learnable parameters, and δ is the coefficient of the square term.

[0045] The dynamic weight adjustment layer uses a gradient-based weight adjustment mechanism. Fixed weight update strategies may not be able to adapt to different learning scenarios. In order to dynamically balance real-time and accuracy, based on a deep understanding of learning dynamics and the need for model adaptability, a dynamic adjustment item is introduced to more flexibly adjust the weights of the features extracted by the adaptive feature extraction layer, increase the model's adaptability, speed up the model's convergence, and avoid overfitting. The dynamic weight adjustment layer adjusts weights using the following formula:

[0046]

[0047] Among them, w new is the adjusted weight, w old is the weight before adjustment, α is the learning rate, is the gradient of the loss function, L is the loss function, λ is the regularization parameter, η is the dynamic adjustment coefficient, and F j is the jth eigenvalue, and m is the number of features. The dynamic weight adjustment layer speeds up the convergence of the model while maintaining accuracy.

[0048] The multi-dimensional optimized output layer adopts a threshold-based output optimization mechanism based on the transmission characteristics of the 5G communication network to reduce network latency and improve data transmission efficiency, ensuring that high recognition accuracy can be maintained even in an unstable network environment. It combines multiple inputs and weight matrices, optimizes the output through nonlinear activation functions and quadratic terms, and optimizes the output through the following formula:

[0049] y j =σ(w new,j F j +b)·ReLU(w new,j σ(w new,j F j +b)+c)+ξ·(w new,j ReLU(wnew,j σ(w new,j F j +b)+c)) 2

[0050] Among them, y j is the jth output of the multi-dimensional optimized output layer, σ is the activation function, and w new,j is the adjusted weight of the jth feature, ReLU is the linear rectification function, b and c are bias terms, and ξ is the coefficient of the square term. Multi-dimensional optimization of the output layer can reduce network latency and improve data transmission efficiency, and maintain high recognition accuracy even in an unstable network environment.

[0051] The fully connected layer performs initial classification scoring based on the feature vector of the current face image to obtain the initial score of the current face image under different categories. The calculation formula of the fully connected layer is:

[0052]

[0053] Among them, Z k is the score of the kth category obtained based on all the features of the current face image. is the weight of the kth category, b k is the bias for the kth class.

[0054] The normalization layer divides each indexed score by the sum of all scores, calculates the predicted probability of the current face image for each category, ensures that the sum of the probabilities is 1, and divides the category of the current face image according to the predicted probability. The specific calculation formula is:

[0055]

[0056] Among them, Y is the final classification category of the current face image, a is the regularization coefficient, is a regular term, and τ is a smoothing coefficient. The output is a category label, indicating which category the model predicts the input image belongs to, thereby achieving face adaptive verification.

[0057] S200: Introduce the data compression algorithm based on the topological structure of complex network model and chaotic dynamics, construct complex network model and complex network chaos coding function, and define the chaos mapping function based on the Lorenz attractor dynamics equation;

[0058] In the 5G communication network, in order to ensure the efficient transmission of face recognition and verification data, a transmission protocol based on data compression technology and optimization is adopted to reduce network delays and ensure the rapid transmission of data in the 5G communication network.

[0059] In terms of data compression, a data compression algorithm specific to 5G communication networks is used. This algorithm uses the topological structure of complex network models and the uncertainty of chaotic dynamics to provide efficient nonlinear compression for various terminal devices in the communication network.

[0060] Specifically, let the face image data be X. In order to find a complex network chaotic coding function E, so that the compressed data can be restored to the original data through inverse coding, a complex network model is constructed, the input of which is the face image data X, and the output is the encoded data U.

[0061] In order to construct the chaotic coding function E of complex networks, a chaotic mapping function Φ is first defined, which is constructed based on the dynamic equation of the Lorenz attractor. The Lorenz attractor is a classic model in chaos theory, and its dynamic equation is:

[0062]

[0063] Among them, t1, t2, and t3 represent the three state variables of the Lorentz system, which mathematically describe the dynamic behavior of the chaotic map. is the time derivative of the state variables t1, t2, t3, that is, the rate of change of these variables, μ, ρ and θ are the parameters of the Lorentz system. By optimizing the parameters of the Lorentz system, a mapping function Φ with good chaotic properties can be obtained.

[0064] The topological structure and weight matrix of the complex network model are introduced, and the chaotic coding function E of the complex network is defined as:

[0065] U=E(X;Θ,W,b)=Φ(ΘX+b)⊙W

[0066] Among them, Θ is the parameter matrix of the complex network chaotic coding, b is the bias vector, W is the weight matrix of the complex network, and the element w of the weight matrix W is ij represents the weight from the i-th node to the j-th node, and its value is determined by the topological structure of the complex network. ⊙ represents the Hadamard product.

[0067] In order to solve the parameter matrix Θ and the weight matrix W, an optimization algorithm based on gradient descent is used. The loss function L is defined as:

[0068]

[0069] in, is the encoded data obtained by the complex network chaotic coding function E, and ω is the regularization parameter, represents the square of the Frobenius norm of Θ.

[0070] By optimizing the loss function L through the gradient descent algorithm, we can obtain the optimal parameter matrix Θ, weight matrix W and bias vector b. The update formula of gradient descent is:

[0071]

[0072]

[0073]

[0074] Among them, α is the learning rate, and t represents the number of iterations. Through the data compression algorithm, efficient nonlinear compression of face image data is achieved, which greatly reduces the network delay during data transmission.

[0075] S300: Introduce a transmission optimization protocol based on multi-dimensional space topology and adaptive spectrum allocation, construct a multi-dimensional space topology model, and propose an adaptive spectrum allocation method based on the output results of the multi-dimensional space topology model.

[0076] In terms of transmission optimization, a transmission optimization protocol specific to mobile communication networks based on multi-dimensional space topology and adaptive spectrum allocation is proposed. This protocol targets the characteristics of 5G communication networks, such as network congestion and signal strength changes. Through this protocol based on the characteristics of 5G communication networks, network delay and data transmission can be greatly optimized in face recognition and verification scenarios.

[0077] Specifically, a multidimensional space topology model is constructed, network state variables and data flow parameters are introduced, and the interaction between these variables is considered. The multidimensional space topology model can dynamically adjust network resources according to the real-time state of the network and data flow, thereby reducing network congestion and improving transmission reliability. The state variables of the network are defined as s = {s1, s2, ..., s M}, M is the number of network states, and the data flow is f = {f1, f2, ..., f Q}, Q is the amount of data traffic. The multidimensional space topology model is:

[0078]

[0079] Among them, T(s, f) is the multidimensional space topology model, α uv is the state variable s u and data flow rate f v are the interaction coefficients, ε and ∈ are the weight coefficients, u∈[1,M], v∈[1,Q].

[0080] At the same time, based on the output results of the multidimensional space topology model, an adaptive spectrum allocation method is proposed to adjust the allocation of spectrum resources in real time to achieve low-latency and high-reliability data transmission. The spectrum resource allocation matrix is ​​constrained by the user's spectrum resource upper limit. By optimizing the spectrum resource allocation, the efficiency and reliability of data transmission can be further improved. Spectrum resource allocation parameters are introduced, and the constraints between these parameters are considered. The spectrum resource allocation matrix is ​​defined as R = r uv , where r uv Subject to the following constraints:

[0081]

[0082]

[0083] Among them, r uv is the amount of spectrum resources allocated between the uth state and the vth state in the adaptive spectrum allocation process, C v and D u are the spectrum resource upper limits of the vth state and the uth state respectively.

[0084] The optimization objective function is defined as:

[0085]

[0086] Among them, O(R) is the optimization objective function, and λ and μ are weight coefficients. By maximizing the optimization objective function O(R), the optimal allocation of spectrum resources can be achieved, thereby achieving low-latency and high-reliability data transmission.

[0087] In order to optimize the multidimensional space topology model and spectrum resource allocation, the gradient descent optimization algorithm is used to solve the optimal spectrum resource allocation matrix R and the parameters of the multidimensional space topology model by minimizing the loss function:

[0088] L(T, R) = κ(T(s, f) - O(R))

[0089] Among them, κ is the regularization parameter. Iterate the optimization until the error meets the expectation.

[0090] Through the above-mentioned data compression algorithm based on complex network theory and chaotic dynamics and the transmission optimization protocol based on multi-dimensional space topology and adaptive spectrum allocation, the network delay and data transmission in the face recognition and verification scenarios in the 5G communication network are greatly optimized.

[0091] To sum up, the adaptive face recognition and verification method based on 5G communication network described in this application is completed.

[0092] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0093] 1. By constructing an authentication method based on a dynamic face network model, the real-time and high-precision face recognition and verification are ensured in a complex mobile communication network environment, especially in a mobile network environment, ensuring efficient data transmission and processing; the design of the adaptive feature extraction layer and the dynamic weight adjustment layer enables the model to extract facial features more accurately and quickly adapt to different learning scenarios, thereby accelerating the convergence of the model while maintaining accuracy; through multi-dimensional optimization of the output layer, multiple inputs and weight matrices are combined to ensure that high recognition accuracy can be maintained even in an unstable network environment, while reducing network latency;

[0094] 2. The adoption of a transmission optimization protocol based on multi-dimensional space topology and adaptive spectrum allocation has significantly reduced network latency and significantly improved data transmission reliability, ensuring rapid data transmission in mobile communication networks;

[0095] 3. By introducing a data compression algorithm based on complex network theory and chaotic dynamics, efficient nonlinear compression of facial image data is achieved, which greatly reduces network delays during data transmission, ensures rapid data transmission, and further optimizes network performance.

[0096] Effect research:

[0097] The technical solution of this application can effectively solve the serious network delay problem that the existing technology often encounters in the mobile communication network, which affects the user experience and the real-time response ability of the system; due to the lack of effective data compression and transmission optimization technology, the data transmission efficiency is low and cannot meet the transmission requirements of large data volume and high speed; the existing models often lack sufficient adaptability and cannot dynamically adjust the model parameters and weights according to different learning scenarios, resulting in unstable performance of the model in different scenarios and unable to meet the diversified and personalized needs; when processing complex models and large data volumes, the calculation efficiency is low and the energy consumption is high, which is particularly obvious on mobile devices, limiting its application and development in mobile communication networks. In addition, the above system or method has undergone a series of effect surveys and verifications, and finally realized the adaptive face recognition and verification method based on the 5G communication network. Through the deep learning model and optimization algorithm, the balance between real-time and accuracy, network delay, data transmission efficiency and model adaptability are effectively solved, and the performance and application scope of face recognition and verification are improved.

[0098] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0099] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0100] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0101] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An adaptive face recognition and verification method based on 5G communication network, characterized in that: The following steps are involved: S100: collecting facial image data through a mobile communication device, building a dynamic facial network model, inputting the facial image data into the dynamic facial network model, and outputting a category label; The dynamic face network model includes: an adaptive feature extraction layer, a dynamic weight adjustment layer, a multi-dimensional optimization output layer, a fully connected layer and a normalization layer; the adaptive feature extraction layer extracts face features based on face feature distribution, and the specific implementation formula of feature extraction is: Where F(x) is the eigenvalue; w i is the weight; x i is the input feature; b is the bias term; n is the number of features, i∈[1,n]; β and γ are learnable parameters; δ is the coefficient of the square term; The dynamic weight adjustment layer adopts a gradient-based weight adjustment mechanism and introduces a dynamic adjustment item to adjust the weights of the facial features extracted by the adaptive feature extraction layer. The specific implementation formula is: Among them, w new is the adjusted weight; w old is the weight before adjustment; α is the learning rate; is the gradient of the loss function; L is the loss function; λ is the regularization parameter; η is the dynamic adjustment coefficient; F j is the jth eigenvalue; m is the number of features; The multi-dimensional optimized output layer introduces a threshold-based output optimization mechanism, and combines two or more inputs and weight matrices to optimize the output through nonlinear activation functions and quadratic terms. The specific implementation formula is: and j =σ(in new,j F j +b)·ReLU(in new,j σ(in new,j F j +b)+c)+ξ (in new,j ReLU(in new,j σ(in new,j F j +b)+c)) 2 Among them, y j is the jth output of the multi-dimensional optimized output layer; σ is the activation function; w new,j is the adjusted weight of the jth feature; ReLU is the linear rectification function; b and c are bias terms; ξ is the coefficient of the square term; The initial scores of the current face image in different categories are obtained through the fully connected layer. The calculation formula of the fully connected layer is: Among them, Z k It is the score of the kth category obtained based on all the features of the current face image; is the weight of the kth category; b k is the bias of the kth category; The predicted probability of the current face image for each category is calculated through the normalization layer, and the category of the current face image is divided according to the predicted probability. The specific calculation formula is: Among them, Y is the category of the current face image; a is the regularization coefficient; is the regularization term; τ is the smoothing coefficient; S200: Introducing the data compression algorithm based on the topological structure and chaotic dynamics of the complex network model, constructing the complex network model and the complex network chaos coding function, defining the chaos mapping function based on the Lorenz attractor dynamics equation; introducing the topological structure and weight matrix of the complex network model, defining the complex network chaos coding function E as: U=E(X;Θ,W,b)=Φ(ΘX+b)⊙W Among them, X is the face image data; Θ is the parameter matrix of complex network chaotic coding; b is the bias vector; W is the weight matrix of the complex network; ⊙ represents the Hadamard product; Φ is the chaotic mapping function; The optimization algorithm is used to solve the parameter matrix and weight matrix of the complex network chaotic coding, and optimize the loss function. The loss function L is defined as: in, It is the encoded data obtained by the complex network chaotic coding function E; and ω are regularization parameters; represents the square of the Frobenius norm of Θ; S300: Introduce a transmission optimization protocol based on multi-dimensional space topology and adaptive spectrum allocation to build a multi-dimensional space topology model. The multi-dimensional space topology model is: Among them, T(s,f) is the multidimensional space topology model; s is the state variable of the network; f is the data flow; M is the number of network states; Q is the number of data flows; α uv is the state variable s u and data flow rate f v The interaction coefficients, u∈[1,M], v∈[1,Q]; ε and ∈ are weight coefficients; According to the output results of the multi-dimensional space topology model, an adaptive spectrum allocation method is proposed to adjust the allocation of spectrum resources in real time; spectrum resource allocation parameters are introduced to define a spectrum resource allocation matrix, which is constrained by the user's spectrum resource upper limit; The optimization objective function is defined as: Among them, O(R) is the optimization objective function; λ and μ are weight coefficients; r uv is the amount of spectrum resources allocated between the uth state and the vth state in the adaptive spectrum allocation process; C v and D u are the spectrum resource upper limits of the vth state and the uth state respectively; and maximize the optimization objective function.

2. The adaptive face recognition and verification method based on 5G communication network according to claim 1 is characterized in that: The S300 specifically includes: A multi-dimensional space topology model is constructed, and network state variables and data flow parameters are introduced; the multi-dimensional space topology model dynamically adjusts network resources according to the real-time state of the network and data flow.

Images