A radar signal simulation and anti-jamming method

Abstract

The application provides a radar signal simulation and anti-interference method, and relates to the technical field of radar signal processing.The method comprises the following steps: constructing an end-to-end deep generative adversarial network framework, including a generator, a discriminator and an anti-interference module; the generator receives a random noise vector and radar parameter coding, generates a time-domain radar signal through a U-Net and LSTM hybrid architecture; the anti-interference module is embedded in the generator decoding layer and performs dynamic frequency domain interference suppression; the discriminator receives the time-domain signal output by the generator, converts it into a short-time Fourier transform spectrum diagram, and evaluates the signal authenticity through a multi-scale convolution network and a spectrum attention module; and the adversarial loss, the signal reconstruction loss and the anti-interference loss are jointly optimized.The method of the application not only can realize high-fidelity radar signal generation and improve simulation authenticity, has dynamic self-adaptive anti-interference ability in a complex electromagnetic environment, but also can realize efficient and real-time processing through an end-to-end integrated architecture.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and more specifically to a radar signal simulation and anti-interference method. Background Technology

[0002] With the widespread application of radar technology in military reconnaissance, meteorological monitoring, and autonomous driving, anti-jamming capability in complex electromagnetic environments has become a core challenge for radar systems. Traditional radar jamming simulation relies on mathematical modeling (e.g., deterministic models such as noise frequency modulation and range gate dragging). While these methods can generate specific jamming types, they lack the ability to simulate dynamic jamming (e.g., adaptive frequency hopping and multi-false-target coordinated jamming). With the development of machine learning, algorithms such as Support Vector Machines (SVM) and Random Forests have been used for jamming classification and suppression. However, these methods rely on manual feature engineering and have limited ability to represent high-dimensional time-frequency signals. Under low signal-to-noise ratio conditions, statistical features are easily contaminated by noise, leading to a sharp drop in classification accuracy.

[0003] In existing related technologies, convolutional neural networks (CNNs) and recurrent neural networks (RNNs) have been introduced into radar signal processing for interference detection and signal separation. For example, CNN-based spectrogram classifiers can identify suppressive interference, but they rely on static datasets for training, making it difficult to adapt to the dynamic evolution of interference patterns, and they cannot generate adversarial interference samples to optimize anti-interference models. Summary of the Invention

[0004] In order to solve the technical problems in related technologies, the present invention provides a radar signal simulation and anti-interference method.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A radar signal simulation and anti-jamming method includes the following steps:

[0007] Step S1: Construct an end-to-end deep generative adversarial network framework, including a generator, a discriminator, and an anti-interference submodule;

[0008] Step S2: The generator receives a random noise vector and radar parameter encoding, and generates a time-domain radar signal through a hybrid architecture of U-Net and LSTM. The LSTM layer models the temporal dependence of the signal, the U-Net structure compresses signal features through the encoder convolutional layer, the decoder deconvolutional layer reconstructs the signal, and high-frequency details are preserved through skip connections.

[0009] Step S3: The anti-interference submodule is embedded in the generator decoding layer to perform dynamic frequency domain interference suppression;

[0010] Step S4: The discriminator receives the time-domain signal output by the generator, converts it into a short-time Fourier transform spectrum, and evaluates the signal authenticity through a multi-scale convolutional network and a spectrum attention module.

[0011] Step S5: Jointly optimize adversarial loss Signal reconstruction loss and anti-interference loss :

[0012]

[0013]

[0014]

[0015]

[0016] In the formula, For the total loss function, To counteract the loss of weight, Weights for signal reconstruction loss. To mitigate interference and reduce weights, The expected data distribution of the radar signal. Let x be the expectation of a Gaussian random noise vector distribution, and let x be the actual radar signal. Let z be the output probability of the discriminator for the true signal, and z be the noise vector. Encoding radar parameters, To generate the signal, T is the number of time-domain sampling points. Let be the output signal value of the generator at time t. The value of the actual signal at time t. The generated signal components after anti-interference processing. The generated interference components after anti-interference processing. It is a very small positive bias.

[0017] Optionally, in step S2, the generator's U-Net structure includes an encoder, a decoder, and a frequency domain attention submodule. The encoder employs four convolutional layers with a kernel size of 5×1 and a stride of 2. The decoder employs four deconvolutional layers with a kernel size of 5×1 and a stride of 2. The frequency domain attention submodule is embedded in the second and third layers of the decoder.

[0018] Optionally, step S3 specifically includes:

[0019] Step S3-1-1: Perform FFT transformation on the intermediate feature map of the decoder to divide the spectrum into multiple sub-bands;

[0020] Step S3-1-2: Calculate the energy of each sub-band and generate suppression weights. The range of values ​​for the interference frequency band weights is as follows: ;

[0021] Step S3-1-3: Reference Mask After suppressing interference components, the time-domain signal is recovered via inverse FFT, where, To output an anti-interference spectrum, To generate the signal spectrum, To suppress weights.

[0022] Optionally, step S3 specifically includes:

[0023] Step S3-2-1: Perform FFT transformation on the intermediate feature map of the decoder to divide the spectrum into 16 sub-bands;

[0024] Step S3-2-2: Generate suppression weights for each subband using a fully connected layer and a Softmax function. ;

[0025] Among them, the suppression weights satisfy the following condition: the range of values ​​for the interference subband suppression weights is... The non-interference subband suppression weight is .

[0026] Optionally, the discriminator multi-scale convolutional network in step S4 includes three parallel branches, wherein the first parallel branch is a 3×3 standard convolutional layer, the second parallel branch is a 5×5 dilated convolutional layer, and the third parallel branch is a 7×7 convolutional layer connected to a global average pooling layer; the outputs of the three parallel branches are concatenated and weighted by a channel attention layer, and finally output the authenticity probability through a fully connected layer.

[0027] Optionally, in step S5, the anti-interference loss This is achieved by maximizing the signal-to-interference ratio (SIR) of the signal after anti-interference, and its gradient is backpropagated to the generator to optimize the source anti-interference characteristics.

[0028] Optionally, step S1 further includes data preprocessing:

[0029] Step S1-1: Construct a radar signal dataset and interference signal library covering multiple scenarios;

[0030] Step S1-2: Standardize the time-domain waveform, frequency-domain spectrum, and radar parameters;

[0031] Step S1-3: Mark the type and frequency range of suppression interference and deceptive interference.

[0032] According to a second aspect of the present invention, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, is capable of implementing the steps of the radar signal simulation and anti-interference method described in any of the technical solutions of the first aspect of the present invention.

[0033] According to a third aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, is capable of implementing the steps of the radar signal simulation and anti-interference method described in any of the technical solutions of the first aspect of the present invention.

[0034] Beneficial effects:

[0035] 1. Through the above technical solution, this invention deeply integrates signal generation and interference suppression through the adversarial training mechanism of deep generative adversarial networks. Specifically, at the signal generation end, the generator adopts a hybrid architecture of U-Net and LSTM, combining radar parameter encoding and random noise input to achieve high-fidelity modeling of multi-modal radar signals (e.g., pulse, frequency-modulated, and phase-coded signals). At the anti-interference end, by embedding a frequency domain attention mechanism and an adaptive filtering module, the spectral energy distribution of the generated signal is dynamically analyzed, automatically identifying interference frequency bands and generating masks to suppress interference components. This invention, through an end-to-end integrated design, abandons the multi-stage serial process of signal generation, interference detection, and filtering suppression in traditional step-by-step processing, providing an efficient solution for real-time signal processing and electronic countermeasures in multi-channel systems such as phased array radars. Thus, the method of this invention not only achieves high-fidelity radar signal generation, improving simulation realism and possessing dynamic adaptive anti-interference capabilities in complex electromagnetic environments, but also achieves efficient and real-time processing through the end-to-end integrated architecture.

[0036] Specifically, the method of the present invention has the following advantages:

[0037] First, the radar signal of this invention is generated through adversarial training, which effectively improves the realism of the simulation. Traditional methods, based on ray tracing signal generation, struggle to model multipath effects, noise, and nonlinear modulation characteristics under complex electromagnetic environments. They require iterative solutions to wave equations, and the time consumption for generating a single frame signal is insufficient to meet real-time requirements. The generator of this invention employs a hybrid architecture of U-Net and LSTM. The skip connections of U-Net preserve signal details (e.g., pulse rising edge, frequency modulation slope), while the LSTM module models the long-term temporal dependence characteristics of the signal and the phase continuity of continuous wave signals. Through the multi-scale convolutional network and spectral attention module of the discriminator, the generator output signal is forced to approximate the statistical distribution of the real signal in both the time and frequency domains.

[0038] Secondly, this invention possesses dynamic adaptive anti-interference capabilities, enhancing robustness in complex electromagnetic environments. Traditional anti-interference techniques, such as FIR filtering and frequency domain notch filtering, require pre-setting interference types and frequency bands, relying on prior interference knowledge. They lack sufficient intelligent and agile interference suppression capabilities against unknown interference. Separate implementation of signal generation and anti-interference measures leads to severe signal distortion after interference suppression. This invention, however, utilizes a frequency domain attention mechanism: the anti-interference submodule automatically identifies and suppresses narrowband noise and deceptive interference, including false target signals, through spectrum sub-band segmentation and dynamic weight allocation. The anti-interference loss function (SIRLoss) is simultaneously optimized with the adversarial training of the generator and discriminator, ensuring the generated signal possesses anti-interference characteristics at the source, rather than relying on post-processing.

[0039] Third, this invention achieves efficient real-time processing through an end-to-end integrated architecture. Traditional system architectures involve separate steps for signal generation, interference detection, and filtering suppression, requiring multiple data conversions and intermediate storage, resulting in prolonged single-frame processing time and high resource consumption. In contrast, this invention's generation, anti-interference, and discrimination modules work collaboratively within a single DGAN framework. Through adversarial training and a dynamic loss function, signal generation and anti-interference strategies can be adjusted in real time, significantly improving the success rate of tasks in complex electromagnetic environments.

[0040] 2. Other beneficial effects or advantages of the present invention will be described in detail in the specific embodiments. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] in:

[0043] Figure 1 This is a flowchart illustrating the steps of a radar signal simulation and anti-interference method provided in an exemplary embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of a radar signal simulation and anti-jamming system architecture provided by an exemplary embodiment of the present invention;

[0045] Figure 3 This is a schematic diagram of a generator network structure provided in an exemplary embodiment of the present invention;

[0046] Figure 4 This is a flowchart illustrating an exemplary embodiment of the anti-interference submodule provided by the present invention;

[0047] Figure 5 This is a schematic diagram of a discriminator network structure provided in an exemplary embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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 some embodiments of the present invention, but not all embodiments.

[0049] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0050] Furthermore, the terms "comprising" and "having," and any variations thereof, used in the description of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices. It should also be noted that in embodiments of this invention, the words "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in embodiments of this invention should not be construed as preferred or advantageous over other embodiments or designs. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0051] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0052] Example 1

[0053] like Figures 1 to 5 As shown, this embodiment provides a radar signal simulation and anti-interference method, including the following steps:

[0054] Step S1: Construct an end-to-end deep generative adversarial network framework, including a generator, a discriminator, and an anti-interference submodule;

[0055] Step S2: The generator receives random noise vectors and radar parameter encodings, and generates time-domain radar signals through a hybrid architecture of U-Net and LSTM. The LSTM layer models the temporal dependence of the signal, the U-Net structure compresses signal features through the encoder convolutional layer, the decoder deconvolutional layer reconstructs the signal, and high-frequency details are preserved through skip connections.

[0056] Step S3: The anti-interference submodule is embedded in the generator decoding layer to perform dynamic frequency domain interference suppression;

[0057] Step S4: The discriminator receives the time-domain signal output by the generator, converts it into a short-time Fourier transform spectrum, and evaluates the signal authenticity through a multi-scale convolutional network and a spectrum attention module.

[0058] Step S5: Jointly optimize adversarial loss Signal reconstruction loss and anti-interference loss :

[0059]

[0060]

[0061]

[0062]

[0063] In the formula, For the total loss function, To counteract the loss of weight, Weights for signal reconstruction loss. To mitigate interference and reduce weights, The expected data distribution of the radar signal. Let x be the expectation of a Gaussian random noise vector distribution, and let x be the actual radar signal. Let z be the output probability of the discriminator for the true signal, and z be the noise vector. Encoding radar parameters, To generate the signal, T is the number of time-domain sampling points. Let be the output signal value of the generator at time t. The value of the actual signal at time t. The generated signal components after anti-interference processing. The generated interference components after anti-interference processing. It is a very small positive bias.

[0064] Through the above technical solution, this invention deeply integrates signal generation and interference suppression via the adversarial training mechanism of deep generative adversarial networks. Specifically, at the signal generation end, the generator employs a hybrid architecture of U-Net and LSTM, combining radar parameter encoding and random noise input to achieve high-fidelity modeling of multimodal radar signals (e.g., pulse, frequency-modulated, and phase-coded signals). At the anti-interference end, by embedding a frequency domain attention mechanism and an adaptive filtering module, the spectral energy distribution of the generated signal is dynamically analyzed, automatically identifying interference frequency bands and generating masks to suppress interference components. This invention, through an end-to-end integrated design, abandons the multi-stage serial process of signal generation, interference detection, and filtering suppression in traditional step-by-step processing, providing an efficient solution for real-time signal processing and electronic countermeasures in multi-channel systems such as phased array radars. Thus, the method of this invention not only achieves high-fidelity radar signal generation, improving simulation realism and providing dynamic adaptive anti-interference capabilities in complex electromagnetic environments, but also achieves efficient and real-time processing through its end-to-end integrated architecture.

[0065] Specifically, the method of the present invention has the following advantages:

[0066] First, the radar signal of this invention is generated through adversarial training, which effectively improves the realism of the simulation. Traditional methods, based on ray tracing signal generation, struggle to model multipath effects, noise, and nonlinear modulation characteristics under complex electromagnetic environments. They require iterative solutions to wave equations, and the time consumption for generating a single frame signal is insufficient to meet real-time requirements. The generator of this invention employs a hybrid architecture of U-Net and LSTM. The skip connections of U-Net preserve signal details (e.g., pulse rising edge, frequency modulation slope), while the LSTM module models the long-term temporal dependence characteristics of the signal and the phase continuity of continuous wave signals. Through the multi-scale convolutional network and spectral attention module of the discriminator, the generator output signal is forced to approximate the statistical distribution of the real signal in both the time and frequency domains.

[0067] Secondly, this invention possesses dynamic adaptive anti-interference capabilities, enhancing robustness in complex electromagnetic environments. Traditional anti-interference techniques, such as FIR filtering and frequency domain notch filtering, require pre-setting interference types and frequency bands, relying on prior interference knowledge. They lack sufficient intelligent and agile interference suppression capabilities against unknown interference. Separate implementation of signal generation and anti-interference measures leads to severe signal distortion after interference suppression. This invention, however, utilizes a frequency domain attention mechanism: the anti-interference submodule automatically identifies and suppresses narrowband noise and deceptive interference, including false target signals, through spectrum sub-band segmentation and dynamic weight allocation. The anti-interference loss function (SIRLoss) is simultaneously optimized with the adversarial training of the generator and discriminator, ensuring the generated signal possesses anti-interference characteristics at the source, rather than relying on post-processing.

[0068] Third, this invention achieves efficient real-time processing through an end-to-end integrated architecture. Traditional system architectures involve separate steps for signal generation, interference detection, and filtering suppression, requiring multiple data conversions and intermediate storage, resulting in prolonged single-frame processing time and high resource consumption. In contrast, this invention's generation, anti-interference, and discrimination modules work collaboratively within a single DGAN framework. Through adversarial training and a dynamic loss function, signal generation and anti-interference strategies can be adjusted in real time, significantly improving the success rate of tasks in complex electromagnetic environments.

[0069] The present invention will now be described with reference to an exemplary embodiment.

[0070] In one exemplary embodiment, the present invention provides a radar signal simulation and anti-jamming method (the overall system architecture of which can be found in [reference]). Figure 2 Based on an end-to-end deep generative adversarial network (DGAN), the following steps may be included:

[0071] Step 1: Organize the radar signal dataset, covering various scenarios including ground object reflection, moving targets, and electromagnetic interference, including time-domain waveforms, frequency-domain spectra, and corresponding radar parameters, carrier frequency, pulse width, and modulation type. Construct an interference signal library, collecting suppression interference (including broadband noise and narrowband sweep frequency) and deceptive interference (including false targets and range drag samples), labeling the interference type and frequency band range.

[0072] Step 2: Preprocess and standardize the acquired data to prepare for subsequent data segmentation and processing. Clean the raw data to remove noise, errors, and duplicate data, and handle missing and outlier values ​​to ensure data quality and integrity.

[0073] Step 3: The generator concatenates noise and radar parameters into a latent vector through a fully connected layer, then captures the temporal features of the signal through an LSTM layer; finally, it stacks deconvolutional layers containing anti-interference attention modules to output the temporal signal.

[0074] Step 4: The discriminator receives the time-domain signal output from the generator, converts it into a short-time Fourier transform spectrum, and then extracts features through a multi-scale convolutional network. The channel attention layer of the anti-interference submodule enhances the discrimination capability of key frequency bands. It outputs the probability of authenticity.

[0075] Step 5: Construct a custom dataset, extract the data, and randomly shuffle it to obtain a training batch of data, which is then input into the network for training.

[0076] Step Six: Train the model on the dataset using the designed deep learning network, loss function. These include adversarial loss, and the discriminator's classification error of the generated signal. x represents the actual radar signal. Let z be the output probability of the discriminator for the true signal, and z be the noise vector. Encoding radar parameters, The signal to be generated; signal reconstruction loss; mean square error between the generated signal and the true signal. ; and anti-interference loss, maximizing the signal-to-noise ratio of the signal after anti-interference:

[0077] Once the loss function converges, the model weight file is obtained.

[0078] Step 7: Load the trained weight file into the input target radar parameters through the designed deep learning network, and the generator outputs a high-fidelity radar signal; the anti-interference submodule synchronously suppresses interference and outputs time-domain waveform and spectrum.

[0079] Step 8: After the task is completed, stop the task and reclaim all resources.

[0080] It should be noted that this invention provides an integrated method for radar signal simulation and anti-jamming based on end-to-end deep generative adversarial networks (DGANs). The generator in the deep generative adversarial network realizes end-to-end generation of radar signals, such as... Figure 3 As shown. Its input layer design consists of noise input. This serves as a random seed for generating diversity. Parameter encoding: Radar parameter encoding is mapped to latent features through a fully connected layer. The generator network structure first uses LSTM layers to model the temporal continuity of the signal, outputting a temporal feature sequence. Then, it synthesizes the signal using U-Net, including an encoder: 4 convolutional layers (kernel size 5×1, stride 2), compressing the signal to a low-dimensional feature space. The decoder: 4 deconvolutional layers (kernel size 5×1, stride 2), progressively reconstructing the temporal signal, with each layer's output skipping connections to the corresponding encoder layer to preserve high-frequency details. Frequency domain attention submodules are embedded in the 2nd and 3rd layers of the decoder as an anti-interference design (structure as follows). Figure 4 This involves dynamically suppressing interference frequency bands. Its design principle is as follows: spectrum segmentation and calculation. An FFT is performed on the intermediate feature map of the decoder to divide the spectrum into 16 sub-bands, and the energy of each sub-band is calculated. Subband suppression weights are generated through fully connected layers and Softmax, and the weights of the interference frequency bands approach 0 (for example). ), application mask To suppress interference, the inverse FFT is used to recover the time-domain signal. The discriminator guides the generator to optimize signal fidelity, such as... Figure 5As shown, the input processing converts the generated signal and the real signal into an STFT spectrum. The multi-scale convolutional network has three branches: branch 1: 3×3 convolution, branch 2: 5×5 dilated convolution, and branch 3: 7×7 convolution (256 channels) + global average pooling. The outputs of the three branches are concatenated, passed through a channel attention layer, and finally input into the global average pooling layer. The output then passes through a fully connected layer to determine the realism probability, ultimately yielding an anti-interference analog radar signal.

[0081] According to a second aspect of the present invention, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it is able to implement the steps of the radar signal simulation and anti-interference method in any of the technical solutions of the first aspect of the present invention.

[0082] It is understood that in this embodiment, the memory may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; furthermore, the memory may include combinations of the above types of memory. The present invention does not specifically limit this.

[0083] Similarly, the processor can implement or execute the various exemplary logical steps described in conjunction with the disclosure of this invention. The processor can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logical steps described in conjunction with the disclosure of this invention. The processor can also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0084] According to a third aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, characterized in that, when executed by a processor, the computer program is capable of implementing the steps of the radar signal simulation and anti-interference method in any of the technical solutions of the first aspect of the present invention.

[0085] In this embodiment, the computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), a register, a hard disk, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof, or any other form of computer-readable storage medium known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium may also be a component of the processor. The processor and the storage medium may reside in an application-specific integrated circuit (ASIC). In embodiments of the present invention, a computer-readable storage medium may be any tangible medium that contains or stores a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0086] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A radar signal simulation and anti-interference method, characterized in that, Includes the following steps: Step S1: Construct an end-to-end deep generative adversarial network framework, including a generator, a discriminator, and an anti-interference submodule; Step S2: The generator receives a random noise vector and radar parameter encoding, and generates a time-domain radar signal through a hybrid architecture of U-Net and LSTM. The LSTM layer models the temporal dependence of the signal, the U-Net structure compresses signal features through the encoder convolutional layer, the decoder deconvolutional layer reconstructs the signal, and high-frequency details are preserved through skip connections. Step S3: The anti-interference submodule is embedded in the generator decoding layer to perform dynamic frequency domain interference suppression; Step S4: The discriminator receives the time-domain signal output by the generator, converts it into a short-time Fourier transform spectrum, and evaluates the signal authenticity through a multi-scale convolutional network and a spectrum attention module. Step S5: Jointly optimize adversarial loss Signal reconstruction loss and anti-interference loss : In the formula, For the total loss function, To counteract the loss of weight, Weights for signal reconstruction loss. To mitigate interference and reduce weights, The expected data distribution of the radar signal. Let x be the expectation of a Gaussian random noise vector distribution, and let x be the actual radar signal. Let z be the output probability of the discriminator for the true signal, and z be the noise vector. Encoding radar parameters, To generate the signal, T is the number of time-domain sampling points. Let be the output signal value of the generator at time t. The value of the actual signal at time t. The generated signal components after anti-interference processing. The generated interference components after anti-interference processing. It is a very small positive bias.

2. The radar signal simulation and anti-interference method according to claim 1, characterized in that, In step S2, the generator's U-Net structure includes an encoder, a decoder, and a frequency domain attention submodule. The encoder uses four convolutional layers with a kernel size of 5×1 and a stride of 2. The decoder uses four deconvolutional layers with a kernel size of 5×1 and a stride of 2. The frequency domain attention submodule is embedded in the second and third layers of the decoder.

3. The radar signal simulation and anti-interference method according to claim 1, characterized in that, Step S3 specifically includes: Step S3-1-1: Perform FFT transformation on the intermediate feature map of the decoder to divide the spectrum into multiple sub-bands; Step S3-1-2: Calculate the energy of each sub-band and generate suppression weights. The range of values ​​for the interference frequency band weights is as follows: ; Step S3-1-3: Reference Mask After suppressing interference components, the time-domain signal is recovered via inverse FFT, where, To output an anti-interference spectrum, To generate the signal spectrum, To suppress weights.

4. The radar signal simulation and anti-interference method according to claim 3, characterized in that, Step S3 specifically includes: Step S3-2-1: Perform FFT transformation on the intermediate feature map of the decoder to divide the spectrum into 16 sub-bands; Step S3-2-2: Generate suppression weights for each subband using a fully connected layer and a Softmax function. ; Among them, the suppression weights satisfy the following condition: the range of values ​​for the interference subband suppression weights is... The non-interference subband suppression weight is .

5. The radar signal simulation and anti-interference method according to claim 1, characterized in that, The discriminator multi-scale convolutional network in step S4 includes three parallel branches: the first parallel branch is a 3×3 standard convolutional layer, the second parallel branch is a 5×5 dilated convolutional layer, and the third parallel branch is a 7×7 convolutional layer connected to a global average pooling layer. The outputs of the three parallel branches are concatenated, weighted by a channel attention layer, and finally output as a true probability through a fully connected layer.

6. The radar signal simulation and anti-interference method according to claim 1, characterized in that, In step S5, the anti-interference loss This is achieved by maximizing the signal-to-interference ratio (SIR) of the signal after anti-interference, and its gradient is backpropagated to the generator to optimize the source anti-interference characteristics.

7. The radar signal simulation and anti-interference method according to claim 1, characterized in that, Step S1 further includes data preprocessing: Step S1-1: Construct a radar signal dataset and interference signal library covering multiple scenarios; Step S1-2: Standardize the time-domain waveform, frequency-domain spectrum, and radar parameters; Step S1-3: Mark the type and frequency range of suppression interference and deceptive interference.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it can implement the steps of the radar signal simulation and anti-interference method according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it can implement the steps of the radar signal simulation and anti-interference method according to any one of claims 1-7.

Images