Intelligent communication signal jamming system based on mobile signal transmission

Abstract

The application discloses an intelligent communication signal interference system based on mobile signal transmission, a data acquisition module: receiving air radio signals of a specified frequency band through a wideband radio frequency front end and pre-processing and converting to obtain real-time spectrum data; a feature analysis module: based on real-time spectrum data, using a deep neural network combined with an implicit space compression sensing based on a self-encoder to construct a signal fingerprint extraction model for signal fingerprint extraction, obtaining unique signal fingerprint identification for different emission sources; the signal fingerprint extraction model includes a feature extraction layer composed of a deep learning network, an implicit space compression conversion layer and a signal fingerprint extraction layer; a strategy generation module: the unique signal fingerprint identification is inverted into environmental physical parameters based on electromagnetic inverse scattering theory and a high-fidelity digital twin is constructed, and an adaptive interference timing strategy is obtained based on the high-fidelity digital twin through electromagnetic simulation.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, and in particular to an intelligent communication signal jamming system based on mobile signal transmission. Background Technology

[0002] Communication signal jamming technology has evolved from traditional broadband suppression jamming to modern intelligent targeting jamming. Early jamming systems mainly relied on high-power transmitters to indiscriminately suppress signals in specific frequency bands. Although this could achieve a certain jamming effect, it was power-intensive, easily detected, and prone to accidentally damaging non-target signals. With the maturity of digital signal processing and adaptive beamforming technologies, jamming systems have begun to possess preliminary signal recognition and direction-focusing capabilities. They can achieve spatially selective jamming through phased array antennas, improving jamming efficiency and concealment.

[0003] Currently, traditional communication signal jamming systems rely heavily on signal frequency and modulation methods, making it difficult to distinguish between different transmitters in the same frequency band and with the same modulation method. This leads to misjudgment of jamming targets and poor environmental adaptability. Furthermore, the jamming strategy does not fully consider the impact of scattering objects (such as buildings and terrain) on signal propagation in the communication environment, and the beamforming parameters do not match the actual environment, resulting in wasted jamming energy or jamming failure. Therefore, this paper proposes an intelligent communication signal jamming system based on mobile signal transmission. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above objectives, the present invention proposes the following technical solution:

[0005] Intelligent communication signal jamming systems based on mobile signal transmission include:

[0006] Data acquisition module: Receives airborne wireless signals in a specified frequency band through a broadband radio frequency front-end and performs preprocessing and conversion to obtain real-time spectrum data;

[0007] Feature parsing module: Based on real-time spectrum data, a signal fingerprint extraction model is constructed using a deep neural network combined with latent space compressed sensing based on a self-different encoder to extract signal fingerprints and obtain unique signal fingerprint identifiers for different emission sources.

[0008] The signal fingerprint extraction model includes a feature extraction layer composed of a deep learning network, a latent space compression and transformation layer, and a signal fingerprint extraction layer;

[0009] Strategy generation module: The environmental physical parameters are derived from the unique signal fingerprint identifier through electromagnetic inverse scattering theory and a high-fidelity digital twin is constructed. Based on the high-fidelity digital twin, an adaptive interference timing strategy is obtained through electromagnetic simulation.

[0010] Signal jamming module: It uses an adaptive jamming timing strategy to control the phased array antenna array to emit highly directional jamming beams, thereby achieving intelligent communication signal jamming.

[0011] The process of obtaining real-time spectrum data is as follows:

[0012] Collect wireless signals in space The frequency is in the specified frequency band. Inside, then to Preprocessing is performed, the signal is amplified by a low-noise amplifier, and then down-converted to obtain the intermediate frequency signal. ;

[0013] For intermediate frequency signals Sampling is performed to obtain the sampled discrete signal. For the discrete signal obtained by sampling Using a Fast Fourier Transform (FFT), with N sampling points, the final real-time spectrum is: Where m = 0, 1, ..., N-1, That is, real-time spectrum data.

[0014] The process of constructing a signal fingerprint extraction model is as follows:

[0015] The real-time spectrum data S[m] is input into the feature extraction layer and processed by the linear transformation and nonlinear activation function of the deep learning network to output the data. Based on output The probability distribution of the latent space is modeled through a latent space compression transformation layer, and the latent space representation z is output. The signal fingerprint extraction layer then uses the decoder part of a variational autoencoder to reconstruct the output using the latent space representation z. Obtain a unique signal fingerprint identifier.

[0016] The process of modeling the probability distribution in the latent space is as follows:

[0017] Output based on feature extraction layer The variational autoencoder part of the latent space compression transformation layer uses two fully connected layers to obtain the mean of the latent space representation. and logarithmic variance ;

[0018] By employing reparameterization techniques, a random vector following a standard normal distribution is introduced. Obtain the latent space representation z: ;

[0019] in, Standard deviation, , This indicates element-wise multiplication.

[0020] Reconstructing the output using the latent space representation z The process of obtaining a unique fingerprint identifier is as follows:

[0021] The output of the feature extraction layer is reconstructed using the latent space representation of z. The decoder's calculation is as follows: ;

[0022] in, These are the weight matrix and bias vector of the decoder. It is the activation function of the decoder. This is the output of the signal fingerprint extraction layer.

[0023] The training process for the fingerprint extraction model is as follows:

[0024] Using the variational lower bound as the loss function, the loss function Composed of reconstruction loss and KL divergence:

[0025] ;

[0026] in, Represents the reconstruction loss, measuring the output of the signal fingerprint extraction layer and... Differences Denotes the KL divergence, using the posterior distribution of the constrained latent space. Approximate prior distribution .

[0027] The process of constructing a high-fidelity digital twin is as follows:

[0028] Fingerprint identification based on unique signal Extract the scattered field , Let f represent the field point and f be the frequency. Based on the theory of electromagnetic inverse scattering, the integral equation of electromagnetic inverse scattering is used to express the relationship between the scattered field and the electromagnetic parameters of the scattering body. The volume of the scattering body in the integral equation is used as the basis for this relationship. The data is discretized into G grid cells, and initial physical parameters are set for each grid cell. ;

[0029] Based on the initial physical parameters Electromagnetic simulations were performed using the finite-difference time-domain method to directly obtain the predicted new scattering field. The objective function is defined as the sum of squared errors between the measured and predicted scattered fields. Then, the conjugate gradient method is used to minimize the objective function J, and the environmental physical parameters are iteratively updated. The environmental physical parameters are obtained until the objective function J is less than the preset threshold. The environmental physical parameters obtained by inversion and the position of the emission source are input into the three-dimensional electromagnetic simulation software to obtain a high-fidelity digital twin.

[0030] The process of obtaining the adaptive interference timing strategy is as follows:

[0031] Assume the emission source in the high-fidelity digital twin environment is The transmitted signal is Phased array antenna arrays have The array element, the first The weighting coefficients of each array element are Construct the radiation pattern function for the antenna array: ,in, It is the signal wavelength. It is the spacing between array elements. It is the azimuth angle;

[0032] Adjusting weighting coefficients through electromagnetic simulation using a high-fidelity digital twin. Change pattern function To obtain the optimal weighting coefficients in the direction of [the desired direction]. As a beamforming parameter;

[0033] Based on the optimal weighting coefficients The interference effects on target communication signals under different interference timings were simulated, and statistical analysis of different interference patterns was conducted using electromagnetic simulation. and Under the combined approach, the optimal interference timing for the target communication signal in terms of bit error rate and throughput performance is... ,in, This is a launch time series. For duration, , For the number of interferences, the optimal interference timing is... That is, the optimal interference timing strategy.

[0034] The present invention has the following beneficial effects:

[0035] In this invention, firstly, a signal fingerprint extraction model is constructed using deep neural networks and latent space compressed sensing technology based on variational autoencoders. This model can automatically learn and extract unique signal fingerprints from different emission sources from real-time spectrum data, overcoming the dependence of traditional methods on fixed feature libraries. By compressing and reconstructing deep signal features, the system can effectively distinguish subtle hardware differences and modulation characteristics, improving the accuracy of identifying specific target signals and the ability to trace their origins in complex electromagnetic environments.

[0036] Secondly, by applying electromagnetic backscattering theory to the interference strategy generation stage, and by inverting environmental physical parameters (dielectric constant, conductivity) from signal fingerprints, and constructing a high-fidelity digital twin consistent with the real environment, the system can accurately simulate the physical processes of signal propagation and scattering in the actual space. This makes the formulation of interference strategies no longer based on ideal models, but on the real physical environment, thus improving the credibility and accuracy of electromagnetic simulation and strategy optimization.

[0037] Finally, based on a high-fidelity digital twin, the beamforming parameters and interference timing strategies of the phased array antenna are dynamically optimized through electromagnetic simulation. The system can simulate the impact of different beam pointing and interference timing combinations on the target communication performance (such as bit error rate and throughput) in the digital twin environment, thereby autonomously finding the optimal interference parameters, so that the transmitted interference beam has high directivity and the interference timing can adapt to the target signal characteristics, achieving the maximum interference effect with minimum energy consumption, while improving the system's stealth. Attached Figure Description

[0038] Figure 1 This is a system block diagram of the intelligent communication signal jamming system based on mobile signal transmission proposed in this invention. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Example: Figure 1 As shown, the intelligent communication signal jamming system based on mobile signal transmission proposed in this invention includes:

[0041] Data acquisition module: Receives airborne wireless signals in a specified frequency band through a broadband radio frequency front-end and performs preprocessing and conversion to obtain real-time spectrum data;

[0042] A broadband radio frequency front-end hardware system is built to receive airborne wireless signals in a specified frequency band. The broadband radio frequency front-end includes an antenna and a radio frequency receiver component. The antenna is responsible for capturing radio electromagnetic wave signals in space, and the radio frequency receiver performs preprocessing operations such as amplification and downconversion on the signals received by the antenna.

[0043] Let the wireless signal to be received in space be... Its frequency is in the specified frequency band Inside, the radio frequency receiver... Preprocessing is performed first, the signal is amplified by a low-noise amplifier (LNA), and the amplified signal is... Next, a down-conversion operation is performed to convert the radio frequency signal into an intermediate frequency (IF) signal. ;

[0044] Then, the intermediate frequency signal is sampled, and according to the Nyquist sampling theorem, the sampling frequency is... Must meet To avoid spectral aliasing, the sampled discrete signal is represented as For the discrete signal obtained by sampling Preprocessing and transformation are performed, and the spectral characteristics of the signal are analyzed using Fast Fourier Transform (FFT). Assuming the number of sampling points is N, the final real-time spectrum is... Where m = 0, 1, ..., N-1, This indicates the Fast Fourier Transform operation. Real-time spectrum data is used as input to the subsequent feature analysis module for signal fingerprint extraction and matching analysis.

[0045] Feature parsing module: Based on real-time spectrum data, a signal fingerprint extraction model is constructed using a deep neural network combined with latent space compressed sensing based on a self-different encoder to extract signal fingerprints and obtain unique signal fingerprint identifiers for different emission sources.

[0046] Obtain real-time spectrum data Then, signal fingerprint extraction is performed using a signal fingerprint extraction model. The core of latent space compressed sensing is to represent and process the signal in a low-dimensional latent space to reduce the amount of data and retain key features.

[0047] The signal fingerprint extraction model includes a feature extraction layer composed of a deep learning network, a latent space compression and transformation layer, and a signal fingerprint extraction layer;

[0048] The feature extraction layer implementation process is as follows:

[0049] The real-time spectrum data S[m] is input into the feature extraction layer of the signal fingerprint extraction model. The feature extraction layer performs feature extraction through linear transformation and nonlinear activation function operation of the deep learning network.

[0050] Suppose that the feature extraction layer of the deep neural network consists of l sub-layers, and the output of the l-th sub-layer is... ,but: ;

[0051] in, It is the weight matrix of the l-th sub-layer. It is the bias vector of the l-th sublayer. It is a non-linear activation function;

[0052] The feature extraction layer, based on a deep neural network, can extract local features of different frequencies and time scales from real-time spectrum data, gradually uncovering subtle feature patterns of the signal. After passing through the feature extraction layer, the signal enters the latent space compression and transformation layer.

[0053] The implementation process of the latent space compression transformation layer is as follows:

[0054] Specifically, the latent space compression transformation layer is a latent space compression transformation structure based on variational autoencoder, which is achieved by modeling the probability distribution of the latent space of the output features of the feature extraction layer;

[0055] The process of modeling the probability distribution in the latent space:

[0056] Output based on feature extraction layer First, the variational autoencoder part of the latent space compression transformation layer uses two fully connected layers to obtain the mean of the latent space representation. and logarithmic variance ;

[0057] Next, in order to sample the latent space representation z from the probability distribution of the latent space, a reparameterization technique is used to introduce a random vector that follows a standard normal distribution. Obtain the latent space representation z:

[0058] ;

[0059] in, , This indicates element-wise multiplication;

[0060] Then, the signal fingerprint extraction layer is entered. The implementation process of the signal fingerprint extraction layer is as follows:

[0061] The signal fingerprint extraction layer is based on the decoder part of the variational autoencoder, and reconstructs the output of the feature extraction layer through the latent space representation z. The decoder's calculation is as follows: ;

[0062] in, These are the weight matrix and bias vector of the decoder. It is the activation function of the decoder. This is the output of the signal fingerprint extraction layer;

[0063] Specifically, the feature extraction layer has l sub-layers, so the output of the last layer (the l-th layer) of the multi-scale feature extraction layer is: Next, we move to the encoder section of the variational autoencoder. The encoder contains fully connected layers that compute the mean and log-variance; both of these layers are related to the (l+1)th layer. Then, the decoder section of the variational autoencoder reconstructs the output of the feature extraction layer from the latent space representation z. This part of the fully connected layer is numbered sequentially to the (l+2)th layer, so its weight matrix and bias vector are labeled as follows. ;

[0064] When training the signal fingerprint extraction model, a variational lower bound is used as the loss function. Composed of reconstruction loss and KL divergence:

[0065] ;

[0066] in, Represents the reconstruction loss, used to measure the difference between the output of the signal fingerprint extraction layer and the output of the signal fingerprint extraction layer. Differences Denotes the KL divergence, used to constrain the posterior distribution of the latent space. Approximate prior distribution ;

[0067] Specifically, when processing signals from different sources, the signal fingerprinting model extracts different signal features. These features reflect the unique characteristics of the source during signal generation and transmission. During the learning process, the variational autoencoder of the signal fingerprinting model encodes these unique features into the latent space representation z. When the decoder reconstructs the signal fingerprint from z... hour, This will preserve these unique characteristics, so the output of the signal fingerprint extraction layer... That is, a unique signal fingerprint identifier for different emission sources;

[0068] For example, in subsequent matching analysis, a signal fingerprint database of known emission sources will be constructed. For each known emission source, its corresponding signal fingerprint will be obtained based on the signal fingerprint extraction model. The system stores the unique signal fingerprint identifier. When a new signal is input, it obtains the fingerprint identifier and calculates the similarity (cosine similarity) between the fingerprint identifier and the existing unique signal fingerprint identifiers in the database to find the most similar emission source, thereby determining the emission source of the new signal.

[0069] Strategy generation module: The environmental physical parameters are derived from the unique signal fingerprint identifier through electromagnetic inverse scattering theory and a high-fidelity digital twin is constructed. Based on the high-fidelity digital twin, an adaptive interference timing strategy is obtained through electromagnetic simulation.

[0070] After obtaining a unique signal fingerprint, environmental physical parameters are inverted from the fingerprint using electromagnetic inverse scattering theory, and a high-fidelity digital twin is constructed.

[0071] Fingerprint identification based on unique signal Extract the scattered field , This represents the field point, i.e., the observation point, where f is the frequency;

[0072] Among them, fingerprint identification is based on unique signals. Determining the statistical field point c and frequency f of the emission source indirectly defines which scattered fields. It needs to be inverted. Only by identifying the emission source can the scattered field data corresponding to the emission source signal be extracted in a targeted manner, and then substituted into the electromagnetic inverse scattering equation to invert the environmental physical parameters.

[0073] According to the theory of electromagnetic inverse scattering, the integral equation of electromagnetic inverse scattering is used to express the relationship between the scattering field and the electromagnetic parameters of the scattering body. The integral equation is expressed as:

[0074] ;

[0075] in, Represents the Green's function and describes the source point. Arrival Point Electromagnetic propagation characteristics, Indicates the electromagnetic contrast of the scatterer. Indicates the incident electric field. Indicates the volume of the scatterer;

[0076] Suppose the volume V containing the scatterer is discretized into G grid cells, and the dielectric constant of each grid cell is... Electrical conductivity is Let the initial physical parameters be , where g represents the index of the grid cell;

[0077] Based on the initial physical parameters Electromagnetic simulations were performed using the finite-difference time-domain (FDTD) method to directly obtain the predicted new scattering field. ;

[0078] Then, the objective function is defined as the sum of squared errors between the measured and predicted scattered fields:

[0079] ,in, As a venue index, The index of frequency f;

[0080] Then, the conjugate gradient method (CG) is used to minimize the objective function J, and the environmental physical parameters are iteratively updated. First, calculate the objective function with respect to the dielectric constant. and conductivity The gradient is calculated, and then the parameters are updated according to the iterative formula of the conjugate gradient method:

[0081] ;

[0082] in, Indicates the first The step size of each iteration is changed, and the parameter update steps are repeated until the objective function J is less than the preset threshold. At this point, the dielectric constant is obtained. and conductivity This refers to the environmental physical parameters obtained through inversion (including the optimal dielectric constant after iteration). and conductivity ), g represents the index of the grid cell;

[0083] The environmental physical parameters obtained from the inversion (the final optimal dielectric constant) and conductivity The location of the transmitter and the source are input into a 3D electromagnetic simulation software (such as CSTMicrowaveStudio) to construct a high-fidelity digital twin that is identical to the actual communication environment. In the high-fidelity digital twin, the dielectric constant and conductivity of each grid cell are set to simulate the signal emission, propagation and scattering process of the transmitter and verify the accuracy of the environmental physical parameters obtained by inversion.

[0084] The process of obtaining the adaptive interference timing strategy is as follows:

[0085] Let the emission source in the high-fidelity digital twin be... Its transmitted signal is Assume the phased array antenna array has The array element, the first The weighting coefficients of each array element are The radiation pattern function for constructing the antenna array is expressed as: ,in, It is the signal wavelength. It is the spacing between array elements. It is the azimuth angle;

[0086] Then, electromagnetic simulation was performed using a high-fidelity digital twin, by adjusting the weighting coefficients. Change pattern function By statistically altering the beamforming shape, the goal is to find the weighting coefficients that optimize the beamforming effect of the interfering beam in the target direction (i.e., the propagation direction of the communication signal to be interfered with). As a beamforming parameter;

[0087] Finally, the optimal interference timing strategy is determined. In the digital twin environment, the weighting coefficients with the best beamforming effect are used. To simulate the interference effect on the target communication signal under different interference timing sequences, let the transmission time sequence of the interference signal be... ( , (Number of interferences), the duration of each interference is... Through electromagnetic simulation, observe different and Under the combined approach, the optimal interference timing for the target communication signal in terms of bit error rate and throughput performance is... ,(from start Duration), optimal interference timing That is, the optimal interference timing strategy.

[0088] Signal jamming module: It uses an adaptive jamming timing strategy to control the phased array antenna array to emit highly directional jamming beams, thereby achieving intelligent communication signal jamming;

[0089] According to the optimal interference timing strategy At the specified time Start, Duration Internally, beamforming parameters (weighting coefficients for optimal beamforming effect) are used. The interference signals emitted by each element of the phased array antenna are controlled. These interference signals superimpose in space to form a highly directional interference beam, the radiation pattern of which is determined by a specific radiation pattern function. The system determines the direction of propagation of the target communication signal precisely, thereby interfering with the intelligent communication signal.

[0090] In the application, several formulas are calculated by removing dimensions and taking their numerical values. The formulas are established by collecting a large amount of data and simulating the most recent real situation. Some coefficients or weights in the formulas are set by those skilled in the art according to the actual situation, so they will not be elaborated here.

[0091] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0092] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart communication signal jamming system based on mobile signal transmission, characterized in that, Comprise: Data acquisition module: receiving and pre-processing the air radio signal of the specified frequency band through the broadband radio frequency front end to obtain real-time spectrum data; Feature analysis module: based on real-time spectrum data, using deep neural network combined with implicit space compression sensing based on self-encoder to construct signal fingerprint extraction model for signal fingerprint extraction, and obtain unique signal fingerprint identification for different emission sources; The signal fingerprint extraction model includes a feature extraction layer composed of a deep learning network, an implicit space compression conversion layer, and a signal fingerprint extraction layer; Strategy generation module: from the unique signal fingerprint identification, the environmental physical parameters are inverted by the electromagnetic inverse scattering theory and the high-fidelity digital twin is constructed, and based on the high-fidelity digital twin, the adaptive interference timing strategy is obtained by electromagnetic simulation; The process of obtaining the adaptive interference timing strategy is: In the environment of high-fidelity digital twinning, the emission source is , the emission signal is , the phased array antenna array has array elements, the weighting coefficient of the th array element is , and a directional diagram function of the antenna array is constructed: wherein is the signal wavelength, is the array element spacing, and is the azimuth angle. Adjusting weighting coefficients by high-fidelity digital twin for electromagnetic simulation Changing a directional pattern function in direction to obtain optimal weighting coefficients as beamforming parameters; Based on the best weighting coefficient The interference effect on the target communication signal under different interference timing is simulated, and the different And The error rate and throughput performance index of the target communication signal under the best interference timing Wherein, The transmission time sequence, The duration, , The number of interference times, the best interference timing That is, the optimal interference timing strategy; Signal interference module: using the adaptive interference timing strategy to control the phased array antenna array to emit high-directivity interference beams, realizing intelligent communication signal interference.

2. The mobile signal transmission based intelligent communication signal jamming system as claimed in claim 1, wherein, The process of obtaining real-time spectrum data is: Collect wireless signals in space The frequency is in the specified frequency band. Inside, then to Preprocessing is performed, the signal is amplified by a low-noise amplifier, and then down-converted to obtain the intermediate frequency signal. ; sampling the intermediate frequency signal to obtain a discrete signal after sampling sampling the discrete signal using fast Fourier transform, assuming the number of sampling points is N, the final real-time frequency spectrum is wherein m=0, 1,.., N-1, namely, real-time frequency spectrum data.

3. The mobile signal transmission based intelligent communication signal jamming system of claim 1, wherein, The process of constructing the signal fingerprint extraction model is: The real-time spectral data S[m] is input into the feature extraction layer to output through a linear transformation and a nonlinear activation function operation of the deep learning network , based on the output The probability distribution of the latent space is modeled through the latent space compression conversion layer to output the latent space representation z, and the decoder part of the variational autoencoder is used through the signal fingerprint extraction layer to combine the latent space representation z to reconstruct and output Obtain a unique signal fingerprint identifier.

4. The mobile signal transmission based intelligent communication signal jamming system of claim 3, wherein, The process of modeling the probability distribution of the implicit space is: based on the output of the feature extraction layer , the mean and the log-variance of the latent space representation are obtained using 2 fully connected layers respectively through the variational autoencoder part of the latent space compression transformation layer; By using reparameterization trick, a random vector following standard normal distribution is introduced Obtain the latent space representation z: ; wherein is the standard deviation, , denotes the multiplication of elements.

5. The mobile signal transmission based intelligent communication signal jamming system of claim 3, wherein, Reconstructing the output in conjunction with the latent space representation z The process of obtaining a unique signal fingerprint identification is: reconstructing the output of the feature extraction layer by the latent space representation z The computation of the decoder is: ; wherein, is a weight matrix and bias vector of the decoder, is an activation function of the decoder, is an output of the signal fingerprint extraction layer.

6. The mobile signal transmission based intelligent communication signal jamming system of claim 3, wherein, The training process of the fingerprint extraction model is: Using the variational lower bound as the loss function, the loss function consists of the reconstruction loss and the KL divergence: ； wherein, represents a reconstruction loss, measuring the difference between the output of the signal fingerprint extraction layer and , represents a KL divergence, constraining the posterior distribution of the latent space to be close to the prior distribution .

7. The mobile signal transmission based intelligent communication signal jamming system of claim 6, wherein, The process of constructing the high-fidelity digital twin is: Fingerprinting based on unique signal , extracting the scattered field , represents the field point, f is the frequency, and the integral equation of electromagnetic inverse scattering is used according to the electromagnetic inverse scattering theory to represent that there is a relationship between the scattered field and the electromagnetic parameters of the scatterer, and the volume where the scatterer is located in the integral equation , is discretized into G grid units, and the initial physical parameters of each grid unit are set ); Based on the initial physical parameters (a) ), electromagnetic simulation is carried out by using the finite difference time domain method to directly obtain the predicted new scattering field , and the objective function is defined as the sum of the square errors of the measured scattering field and the predicted scattering field, and then the conjugate gradient method is used to minimize the objective function J, and the environmental physical parameters (b) ) are iteratively updated until the objective function J is less than the preset threshold to obtain the environmental physical parameters. The environmental physical parameters obtained by inversion and the position of the transmitting source are input into a three-dimensional electromagnetic simulation software to obtain a high-fidelity digital twin.

Images