Multi-band radar anti-interference method

Abstract

The invention provides a multi-band radar anti-interference method, which comprises the following steps of: 1, acquiring an enhanced model in an offline state through offline dynamic confrontation training; 2, obtaining an original echo signal in an online state, carrying out adaptive filtering, and carrying out space-time spectrum feature extraction by using an enhanced model; 3, performing feature fusion and target separation according to the extracted space-time spectrum features; step 4, performing optimization according to the separation result, and judging whether a termination condition is met or not; if yes, ending and outputting the multi-dimensional data packet; and if not, returning to the step 2. According to the method, through triple technical breakthrough of hardware reconstruction, a dynamic confrontation training method and closed-loop feedback, the anti-interference performance and the vital sign monitoring precision reach the industry leading level, and the method is particularly suitable for extreme scenes such as fire rescue and battlefield search and rescue. In the next step, light weight of the model can be emphatically optimized, and the deployment efficiency of the mobile platform is further improved.

Description

Technical Field

[0001] This invention relates to the field of radar anti-jamming technology, specifically to a method for multi-band radar anti-jamming. Background Technology

[0002] In recent years, radar technology has played an increasingly important role in disaster relief, military reconnaissance, and autonomous driving. However, radar signals in complex environments are susceptible to various interferences, such as electromagnetic noise, multipath effects, mechanical vibration, and human interference, leading to decreased target detection accuracy and even misjudgment. Traditional single-band radar anti-jamming methods have the following limitations:

[0003] Limited frequency band and limited anti-interference capability: Traditional radars usually operate in a fixed frequency band (such as 24GHz or 77GHz). When this frequency band is subjected to strong interference, the system performance drops sharply and lacks the ability to dynamically switch to avoid interference.

[0004] Signal processing algorithms lack generalization ability: conventional filtering methods (such as FIR filtering and static wavelet denoising) are difficult to cope with non-stationary interference, especially in the complex electromagnetic environment of disaster sites, where existing algorithms have poor adaptability.

[0005] Poor adaptability to extreme environments: Harsh conditions such as high temperature, high humidity, and smoke may cause the performance of radar hardware to degrade, while existing systems lack targeted thermal management and anti-jamming collaborative design.

[0006] Currently, some studies attempt to improve anti-interference capabilities through multi-band fusion or deep learning, but the following problems still exist:

[0007] Slow frequency band switching speed (e.g., mechanical switches have a delay of milliseconds) makes it impossible to respond to sudden interference in real time;

[0008] Feature extraction relies on manual design and lacks generalization ability for unknown types of interference.

[0009] Vital signs detection (such as breathing and heartbeat) is easily affected by environmental noise, and the error is large when the signal-to-noise ratio is low.

[0010] Therefore, there is an urgent need for a radar anti-jamming method that can dynamically reconstruct frequency bands, combine deep adversarial training, and possess high-precision multi-target separation capabilities to meet the detection requirements in complex scenarios. Summary of the Invention

[0011] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a method for anti-jamming of multi-band radar, which can solve the existing problems.

[0012] To achieve the above objectives, the technical solution of the present invention is as follows:

[0013] This invention is achieved through the following technical solution: a method for multi-band radar anti-jamming, comprising the following steps:

[0014] Step 1: Obtain an enhanced model in offline state through offline dynamic adversarial training;

[0015] Step 2: Obtain the original echo signal in the online state, perform adaptive filtering, and use the enhanced model to extract spatiotemporal spectral features;

[0016] Step 3: Perform feature fusion and target separation based on the extracted spatiotemporal spectral features;

[0017] Step 4: Optimize based on the separation results and determine whether the termination condition is met;

[0018] If so, end and output the multidimensional data packet;

[0019] If not, return to step two.

[0020] Furthermore, step one specifically includes:

[0021] 1.1 Acquire multi-dimensional environmental and target signals from the echo information; and perform signal preprocessing;

[0022] 1.2. Extract spatiotemporal spectral features from the preprocessed signal;

[0023] 1.3. Perform dynamic adversarial training on the extracted features in an offline state to obtain an enhanced model.

[0024] Furthermore, the echo signal is transmitted via a multi-band reconfiguration radar carried by a drone or robot, which connects the multi-band reconfiguration radar to a MIMO antenna array, and the echo signal is acquired using the MIMO antenna array.

[0025] Furthermore, the multi-band reconfiguration radar includes a dual-band reconfiguration radar, specifically comprising a PIN diode switch connected to a 24GHz local oscillator and a 77GHz local oscillator; a shared mixer connected to the PIN diode switch; a power amplifier connected to the shared mixer; and a transmitting antenna connected to the power amplifier.

[0026] Furthermore, the dynamic adversarial training includes:

[0027] Noise injection: Randomly add metal vibration, electromagnetic pulse and smoke scattering noise to the training data to simulate post-disaster interference;

[0028] Adaptive filtering optimization: Combining wavelet threshold denoising and Kalman filtering, the optimal denoising strategy is selected based on the real-time signal-to-noise ratio;

[0029] Adversarial training: Generative adversarial networks are used to generate adversarial samples to determine the enhanced model.

[0030] Furthermore, the spatiotemporal spectral feature extraction includes:

[0031] Three types of features are extracted from the echo information: spatial angle (AoA): the target azimuth angle is calculated by MIMO beamforming (accuracy ±0.5°); micro-Doppler spectrum: short-time Fourier transform (STFT) generates a spectrum to capture the respiratory (0.1-0.5Hz) and heartbeat (0.8-3Hz) frequency bands; temporal phase change: based on the phase difference of I / Q signals, the micro-motion trajectory of the chest cavity is tracked (resolution 0.1mm).

[0032] Furthermore, the feature fusion and target separation include:

[0033] The three types of features—spatial angle, micro-Doppler spectrum, and temporal phase change—are normalized and concatenated into a three-dimensional tensor (spatial × spectrum × temporal), which is then input into the LSTM-Attention enhanced network model. Through a gating mechanism, key features are dynamically weighted to determine the respiratory and heart rate and location coordinates of multiple targets.

[0034] Furthermore, the adaptive filtering includes combining wavelet threshold denoising with Kalman filtering, and selecting the optimal denoising strategy based on the real-time signal-to-noise ratio.

[0035] Furthermore, the optimization includes switching the local oscillator source via a PIN diode switch, dynamically adjusting the wavelet threshold, updating the Kalman covariance, and using feature weights for feature fusion.

[0036] Furthermore, the termination condition includes a preset number of iterations or the target confidence level reaching a preset range; the multidimensional data packet includes the respiratory and heart rate and location coordinates of multiple targets.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The multi-band radar anti-jamming method of this invention employs a dynamic adversarial training mechanism and multi-modal interference simulation: adversarial samples such as metal vibration and electromagnetic pulses are generated through FGSM / PGD attacks, improving the model's breathing detection accuracy and reducing false alarm rate even at a low signal-to-noise ratio of -20dB. Furthermore, this invention employs a closed-loop optimization mechanism: by dynamically adjusting feature weights, the efficiency of multi-band data fusion is improved.

[0039] The method of this invention achieves industry-leading performance in anti-interference and vital sign monitoring accuracy through triple technological breakthroughs in hardware reconstruction (dual-band + MIMO), algorithm innovation (dynamic adversarial training), and system optimization (closed-loop feedback), making it particularly suitable for extreme scenarios such as fire rescue and battlefield search and rescue. The next step could focus on optimizing model lightweighting to further improve the deployment efficiency on mobile platforms. Attached Figure Description

[0040] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0041] Figure 1 This is a schematic flowchart of a multi-band radar anti-jamming method according to the present invention;

[0042] Figure 2 This is a schematic diagram of the anti-interference training process in an embodiment of the present invention;

[0043] Figure 3 This is a flowchart illustrating steps two and three in an embodiment of the present invention. Detailed Implementation

[0044] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0045] This invention provides a method for anti-jamming multi-band radar, the process of which is as follows: Figure 1 As shown, it includes the following steps:

[0046] Step 1: Obtain an enhanced model in offline state through offline dynamic adversarial training;

[0047] Specifically, step one includes:

[0048] 1.1 Acquire multi-dimensional environmental and target signals from the echo information; and perform signal preprocessing;

[0049] The echo signal is carried by a drone or robot through a multi-band reconfiguration radar, which connects the multi-band reconfiguration radar to a MIMO antenna array, and the echo signal is obtained using the MIMO antenna array.

[0050] The multi-band reconfiguration radar includes a dual-band reconfiguration radar, specifically comprising a PIN diode switch connected to a 24GHz local oscillator and a 77GHz local oscillator; a shared mixer connected to the PIN diode switch; a power amplifier connected to the shared mixer; and a transmitting antenna connected to the power amplifier.

[0051] The MIMO antenna array is used to receive signals transmitted by the multi-band reconfiguration radar module; the MIMO antenna array includes...

[0052] Antenna element, the antenna element being connected to the transmitting antenna,

[0053] Butler matrix grid cells, wherein the Butler matrix grid cells are connected to antenna cells;

[0054] A beamforming control unit, wherein the beamforming control unit is connected to a Butler matrix grid cell;

[0055] A directional beamforming unit, which is connected to a beamforming control unit.

[0056] Specifically, the MIMO antenna array adopts an 8×8 foldable patch antenna structure, and the unfolded size of the MIMO antenna array is 30cm×30cm, and the folded size is 15cm×15cm. The unit spacing is 6.25mm for 24GHz and 1.95mm for 77GHz in half wavelength.

[0057] In the MIMO antenna array design, an 8×8 foldable patch antenna structure is adopted, and the antenna elements are folded and stored using a flexible PCB substrate. The spacing between each antenna element is half a wavelength (6.25mm at 24GHz; 1.95mm at 77GHz). Beamforming is achieved through Butler matrix grid elements, generating 64 independent beams (azimuth coverage ±60°, elevation ±30°). Field tests show that the array achieves a spatial resolution of ±0.5° in 77GHz mode, distinguishing multiple targets with a spacing ≥0.5m, representing a 4-fold performance improvement over the traditional 4×4 array. The beamforming control unit can dynamically adjust the beam pointing, achieving a spatial resolution of ±0.5° in 77GHz mode. The MIMO antenna array integrates high-temperature resistant packaging (ceramic substrate + aluminum alloy shell) and an adaptive heat dissipation design, allowing it to operate continuously for more than 8 hours in environments ranging from -20℃ to 70℃, meeting the needs of extreme post-disaster scenarios. Specifically, to adapt to high-temperature scenarios such as fires and explosions, the radar module adopts a multi-layer composite packaging solution: Material selection: The core circuit board is made of aluminum nitride ceramic (thermal conductivity 170W / m·K), the outer shell is made of 6061 aluminum alloy (melting point ≥600℃), and the interior is filled with silicone gel (temperature resistance -40℃~200℃) to achieve heat insulation and buffering; Heat dissipation design: Integrated micro vortex fan and heat pipe heat conduction structure to ensure that the module can work continuously for 8 hours at an ambient temperature of 70℃ with a temperature rise ≤15℃ (actual measured data); Sealing test: Passed IP67 protection level certification, dustproof and waterproof, and adaptable to extreme weather such as rainstorms and sandstorms.

[0058] The preprocessing includes, but is not limited to: 1. Removing DC components: eliminating static background interference in radar signals through a high-pass filter; 2. Motion compensation: dynamically correcting phase shift caused by vibration of rescue machinery based on Doppler frequency shift estimation of the Doppler motion supplement unit; 3. Adaptive denoising: combining wavelet threshold denoising of the wavelet threshold denoising unit with Kalman filtering algorithm of Kalman dynamic filtering unit, dynamically adjusting parameters according to the environmental noise spectrum, improving the signal-to-noise ratio by 3dB.

[0059] 1.2. Extract spatiotemporal spectral features from the preprocessed signal;

[0060] The spatiotemporal spectral feature extraction includes

[0061] Three types of features are extracted from the echo information: spatial angle (AoA): the target azimuth angle is calculated by MIMO beamforming (accuracy ±0.5°); micro-Doppler spectrum: short-time Fourier transform (STFT) generates a spectrum to capture the respiratory (0.1-0.5Hz) and heartbeat (0.8-3Hz) frequency bands; temporal phase change: based on the phase difference of I / Q signals, the micro-motion trajectory of the chest cavity is tracked (resolution 0.1mm).

[0062] 1.3. Perform dynamic adversarial training on the extracted features in an offline state to obtain an enhanced model.

[0063] The dynamic adversarial training includes:

[0064] Noise injection: Randomly add metal vibration, electromagnetic pulse and smoke scattering noise to the training data to simulate post-disaster interference;

[0065] Adaptive filtering optimization: Combining wavelet threshold denoising and Kalman filtering, the optimal denoising strategy is selected based on the real-time signal-to-noise ratio;

[0066] Adversarial training: Generative adversarial networks (GANs) are used to generate adversarial examples to determine the enhanced model. This enhances the model's generalization ability against unknown interference. For example, adversarial example generation strategies include, but are not limited to, FGSM / PGD attack strength. The anti-interference training process is as follows: Figure 2 As shown.

[0067] Step 2: Obtain the original echo signal in the online state, perform adaptive filtering, and use the enhanced model to extract spatiotemporal spectral features;

[0068] Specifically, the original echo signal in the online state is obtained by using a multi-band reconfiguration radar mounted on a drone or robot, which connects the multi-band reconfiguration radar to a MIMO antenna array, and the MIMO antenna array is used to acquire the echo signal online.

[0069] The adaptive filtering combines wavelet threshold denoising with Kalman filtering, selecting the optimal denoising strategy based on the real-time signal-to-noise ratio (SNR). For example, the SNR threshold judgment logic includes: if SNR < 5dB: wavelet denoising is used preferentially; if 5dB ≤ SNR < 15dB: Kalman filtering + wavelet filtering are used together; otherwise: only Kalman filtering is used.

[0070] The spatiotemporal spectral feature extraction includes

[0071] Three types of features are extracted from the echo information: spatial angle (AoA): the target azimuth angle is calculated by MIMO beamforming (accuracy ±0.5°); micro-Doppler spectrum: short-time Fourier transform (STFT) generates a spectrum to capture the respiratory (0.1-0.5Hz) and heartbeat (0.8-3Hz) frequency bands; temporal phase change: based on the phase difference of I / Q signals, the micro-motion trajectory of the chest cavity is tracked (resolution 0.1mm).

[0072] Step 3: Perform feature fusion and target separation based on the extracted spatiotemporal spectral features;

[0073] The feature fusion and target separation include:

[0074] The three types of features—spatial angle, micro-Doppler spectrum, and temporal phase change—are normalized and concatenated into a three-dimensional tensor (spatial × spectrum × temporal), which is then input into an LSTM-Attention enhanced network model. Through a gating mechanism, key features are dynamically weighted to determine the respiratory and heart rate and location coordinates of multiple targets. The flowcharts for steps two and three are as follows: Figure 3 As shown.

[0075] Step 4: Optimize based on the separation results and determine whether the termination condition is met;

[0076] If so, end and output the multidimensional data packet;

[0077] If not, return to step two.

[0078] Specifically, the optimization includes switching the local oscillator source via a PIN diode switch, dynamically adjusting the wavelet threshold, updating the Kalman covariance, and using feature weights for feature fusion. For example, the PIN diode switch for switching the local oscillator source prioritizes 77GHz for ranging and 24GHz for wide-area search.

[0079] The termination conditions include a preset number of iterations or a target confidence level reaching a preset range; the multidimensional data package includes the respiratory and heart rate and location coordinates of multiple targets. For example, the maximum number of iterations is N = 50; the target confidence threshold is: respiratory detection P ≥ 0.95, and the positioning error ≤ 0.3m.

[0080] This invention employs a dynamic adversarial training mechanism and multimodal interference simulation: adversarial samples such as metallic vibrations (0.5-5kHz) and electromagnetic pulses (rise time <1ns) are generated through FGSM / PGD attacks, enabling the model to maintain a breathing detection accuracy of over 85% and reduce the false alarm rate to 1.2% (compared to approximately 8% for traditional methods) even at a low signal-to-noise ratio of -20dB. Furthermore, this invention utilizes a closed-loop optimization mechanism: by dynamically adjusting feature weights (e.g., setting the weight coefficient for 77GHz data to 0.7 and 24GHz to 0.3), the efficiency of multi-band data fusion is improved by 40%.

[0081] Therefore, this method achieves industry-leading performance in anti-interference capabilities and vital sign monitoring accuracy through a triple technological breakthrough: hardware reconstruction (dual-band + MIMO), algorithmic innovation (dynamic adversarial training), and system optimization (closed-loop feedback). It is particularly suitable for extreme scenarios such as fire rescue and battlefield search and rescue. The next step could focus on optimizing model lightweighting to further improve deployment efficiency on mobile platforms.

[0082] 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, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this document.

[0083] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0084] In the embodiments provided herein, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, apparatuses, or units, or they may be electrical, mechanical, or other forms of connection.

[0085] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments described herein, depending on actual needs.

[0086] Furthermore, the functional units in the various embodiments of this document can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0087] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this paper, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this paper. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0088] This document uses specific embodiments to illustrate the principles and implementation methods of this document. The descriptions of the embodiments above are only for the purpose of helping to understand the methods and core ideas of this document. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this document. Therefore, the content of this specification should not be construed as a limitation of this document.

Claims

1. A method for multi-band radar anti-jamming, characterized in that, Includes the following steps: Step 1: Obtain an enhanced model in offline state through offline dynamic adversarial training; Step 2: Obtain the original echo signal in the online state, perform adaptive filtering, and use the enhanced model to extract spatiotemporal spectral features; Step 3: Perform feature fusion and target separation based on the extracted spatiotemporal spectral features; Step 4: Optimize based on the separation results and determine whether the termination condition is met; If so, end and output the multidimensional data packet; If not, return to step two.

2. The method for multi-band radar anti-jamming according to claim 1, characterized in that, Step one specifically includes: 1.1 Acquire multi-dimensional environmental and target signals from the echo information and perform signal preprocessing; 1.

2. Extract spatiotemporal spectral features from the preprocessed signal; 1.

3. Perform dynamic adversarial training on the extracted features in an offline state to obtain an enhanced model.

3. The method for multi-band radar anti-jamming according to claim 2, characterized in that, The echo signal is carried by a drone or robot through a multi-band reconfiguration radar, which connects the multi-band reconfiguration radar to a MIMO antenna array, and the echo signal is obtained using the MIMO antenna array.

4. The method for multi-band radar anti-jamming according to claim 3, characterized in that, The multi-band reconfiguration radar includes a dual-band reconfiguration radar, specifically comprising a PIN diode switch connected to a 24GHz local oscillator and a 77GHz local oscillator; a shared mixer connected to the PIN diode switch; a power amplifier connected to the shared mixer; and a transmitting antenna connected to the power amplifier.

5. The method for multi-band radar anti-jamming according to claim 4, characterized in that, The dynamic adversarial training includes: Noise injection: Randomly add metal vibration, electromagnetic pulse and smoke scattering noise to the training data to simulate post-disaster interference; Adaptive filtering optimization: Combining wavelet threshold denoising and Kalman filtering, the optimal denoising strategy is selected based on the real-time signal-to-noise ratio; Adversarial training: Generative adversarial networks are used to generate adversarial samples to determine the enhanced model.

6. The method for multi-band radar anti-jamming according to claim 5, characterized in that, The spatiotemporal spectral feature extraction includes: Three types of features are extracted from the echo information: spatial angle: the target azimuth angle is calculated by MIMO beamforming; micro-Doppler spectrum: short-time Fourier transform is used to generate a spectrum to capture the respiratory and heartbeat frequency bands; temporal phase change: the trajectory of chest cavity micro-movement is tracked based on the phase difference of I / Q signals.

7. A method for multi-band radar anti-jamming according to claim 6, characterized in that, The feature fusion and target separation include: The three types of features—spatial angle, micro-Doppler spectrum, and temporal phase change—are normalized and concatenated into a three-dimensional tensor (spatial × spectrum × temporal), which is then input into the LSTM-Attention enhanced model. Through a gating mechanism, key features are dynamically weighted to determine the respiratory and heart rate and location coordinates of multiple targets.

8. A method for multi-band radar anti-jamming according to claim 7, characterized in that, The adaptive filtering includes combining wavelet threshold denoising and Kalman filtering, and selecting the optimal denoising strategy based on the real-time signal-to-noise ratio.

9. A method for multi-band radar anti-jamming according to claim 8, characterized in that, The optimization includes switching the local oscillator source via a PIN diode switch, dynamically adjusting the wavelet threshold, updating the Kalman covariance, and using feature weights for feature fusion.

10. A method for multi-band radar anti-jamming according to any one of claims 1-9, characterized in that, The termination conditions include a preset number of iterations or the target confidence level reaching a preset range; The multidimensional data packet includes the respiratory and heart rate and location coordinates of multiple targets.

Images