An ISRJ suppression method based on multi-model fusion

Abstract

The present invention provides an ISRJ suppression method based on multi-model fusion, which relates to the technical field of ISRJ suppression and specifically includes the following steps: constructing an ISRJ signal model; using a fusion interference detection method based on Transformer and decision tree to extract interference segments; on the basis of the extracted interference segments, using complex variational mode decomposition (CVMD) to extract the continuous component of ISRJ and decomposing the continuous component of ISRJ into band-limited intrinsic mode functions (BIMFs); based on the estimated interference signal-to-noise ratio, selecting the interference signal and the target signal, and reconstructing the selected interference signal; using multi-scale wavelet decomposition to reconstruct the interference signal; fusing the reconstructed interference signals; and using an alternating iteration method to reconstruct the target signal and the interference signal. The technical solution of the present invention overcomes the problems in the prior art that complex modulation signals cannot be accurately characterized and the target signal cannot be completely retained.

Description

Technical Field

[0001] The present invention relates to the field of ISRJ suppression, and particularly to an ISRJ suppression method based on multi-model fusion. Background Art

[0002] Intermittent Sampling Repeater Jamming (ISRJ) is a typical active main lobe jamming, which has high power, strong coherence and complex modulation characteristics. It will generate a large number of false signal peaks, masking the real target signal and resulting in incorrect detection results. Therefore, it is crucial to develop a robust main lobe ISRJ suppression method. In recent years, many researchers have been studying this problem.

[0003] ISRJ suppression methods are mainly divided into three categories: interference signal filtering, target signal extraction and interference reconstruction. In the interference signal filtering method, the interference signal is parameterized or semi-parameterized and modeled in the form of a known signal (such as a chirp or cosine signal), and a time-frequency filter is designed to remove the interference. Although this method has high computational efficiency, due to the mismatch between the complex modulation of the interference and the assumed signal form, the residual interference is relatively significant. In the target signal extraction method, the interference position is first detected and set to zero, and then the sparse recovery technology is used to extract the target signal from the interference-free area. However, this method will inevitably cause signal loss and may result in missed detection under low Signal-to-Noise Ratio (SNR) and high-density interference conditions. In the interference reconstruction method, techniques such as Principal Component Analysis (PCA) or dictionary learning are used to extract the main signal components from multiple consistent interference segments. These methods have high reconstruction accuracy for stable or slowly varying ISRJ, but their performance will significantly decline when the interference changes rapidly.

[0004] Therefore, there is a need for an ISRJ suppression method based on multi-model fusion that can accurately characterize complex modulation signals and completely retain target signals. Summary of the Invention

[0005] The main object of the present invention is to provide an ISRJ suppression method based on multi-model fusion to solve the problems in the prior art that complex modulation signals cannot be accurately characterized and target signals cannot be completely retained.

[0006] To achieve the above object, the present invention provides an ISRJ suppression method based on multi-model fusion, which specifically includes the following steps:

[0007] S1, construct an ISRJ signal model.

[0008] S2, use a fusion interference detection method based on Transformer and decision tree to extract interference segments.

[0009] S3. On the basis of the extracted interference segments, the continuous component of the ISRJ is extracted by using the complex variational mode decomposition (CVMD), and the continuous component of the ISRJ is decomposed into the band-limited intrinsic mode functions (BIMFs).

[0010] S4. Based on the estimated interference signal-to-noise ratio, the interference signal and the target signal are selected, and the selected interference signal is reconstructed.

[0011] S5. The interference signal is reconstructed by using the multi-scale wavelet decomposition.

[0012] S6. The interference signals reconstructed in steps S4 and S5 are fused.

[0013] S7. After using the interference signal obtained in step S6 to cancel the interference of the original signal, the target signal is reconstructed by using the sparse reconstruction algorithm. Then, after using the reconstructed target signal to cancel the target signal in the original signal, the interference signal is reconstructed. Such alternating iteration is performed to reconstruct the interference signal and the target signal.

[0014] Further, step S1 specifically includes the following steps:

[0015] S1.1. Assume that the radar transmission waveform is , which is modeled as a linear frequency modulation (LFM) signal; the ISRJ signal composed of segments is expressed as :

[0016] ;

[0017] where is the interference amplitude after power amplification, is the rectangular pulse function, is the storage time of the th interference segment, and is the interference segment length.

[0018] S1.2. Modulate each interference segment. After modulation, the ISRJ signal is expressed as :

[0019] ;

[0020] where is the convolution operator, and is the random modulation kernel.

[0021] S1.3. If there are targets in the detection scene, the total received signal is expressed as :

[0022] ;

[0023] Among them, is the amplitude of the th target, is the time delay of the th target echo, is Gaussian white noise; after sampling, is expressed as a vector .

[0024] Furthermore, step S3 specifically includes the following steps:

[0025] S3.1, Band-limited Intrinsic Mode Functions BIMFs:

[0026] ;

[0027] ;

[0028] Among them, is the center frequency, is the mode function, is the th BIMFs obtained by decomposition, is the number of BIMFs, is the target signal or the observed signal, is the exponential function, is the minimum value function.

[0029] S3.2, Solve the band-limited intrinsic mode functions BIMFs by using the augmented Lagrangian method of the Alternating Direction Method of Multipliers ADMM.

[0030] Furthermore, step S4 specifically includes the following steps:

[0031] S4.1, The interference signal reconstructed by using the selected BIMFs is expressed as:

[0032] ;

[0033] Among them, is the support set of the selected BIMFs, estimated as:

[0034] ;

[0035] Among them, is the number of the selected BIMFs components, is the signal length, is the power of the signal and the noise, is the threshold factor, is the square of the 2-norm of, It is a parameter minimum value function.

[0036] S4.2, solve using the greedy algorithm , the reconstructed interference signal The discrete digital signal of is represented as a vector .

[0037] Furthermore, step S5 specifically includes the following steps:

[0038] S5.1, using the interference segments obtained in step S2, the interference wavelet coefficients with different dilation and translation factors are represented as :

[0039] ;

[0040] Among them, is the complex conjugate of the mother wavelet, is the wavelet kernel with dilation factor and translation factor , and is represented as:

[0041] ;

[0042] Among them, is the mother wavelet.

[0043] S5.2, design an adaptive threshold to separate the pulse interference component and the target signal component, as follows:

[0044] ;

[0045] ;

[0046] Among them, is the threshold factor related to JSRN, is the adjustment factor, is the processed wavelet coefficient, is a non-linear function, is the sign function.

[0047] S5.3, represent the pulse component extracted by the wavelet as :

[0048] ;

[0049] The discrete digital signal of is represented as a vector .

[0050] Furthermore, step S6 specifically includes the following steps:

[0051] S6.1, propose the following weighted fusion filter :

[0052] ;

[0053] Among them, 、 are weights.

[0054] S6.2. Let , and determine the variable to be solved using the criterion of minimizing the residual sum of squares, as follows:

[0055] ;

[0056] Among them, represents the conjugate transpose of , is the optimal solution of;

[0057] is obtained through convex optimization:

[0058] ;

[0059] ;

[0060] ;

[0061] Among them, and are intermediate variables, is the conjugate operation, is the transpose of.

[0062] S6.3. By performing interference cancellation on and combining with the sparse reconstruction method to extract the target signal ,

[0063] ;

[0064] Among them, is the sparse dictionary related to the transmission waveform, is the penalty factor, is the optimal solution for solving the target signal.

[0065] S6.4. Solve by the conjugate gradient method.

[0066] The present invention has the following beneficial effects:

[0067] Based on the multi-domain decomposition and reconstruction method, the present invention analyzes the interference characteristics, can capture the complex signal structure, and realizes accurate interference signal estimation. The present invention uses the estimated interference signal to cancel the interference in the echo, thereby completely retaining the target signal structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. In the drawings:

[0069] Figure 1 The flowchart of an ISRJ suppression method based on multi-model fusion of the present invention is shown.

[0070] Figure 2 The simulation diagram of a typical ISRJ in the time domain is shown.

[0071] Figure 3 The time-frequency distribution TFD diagram of the interference is shown.

[0072] Figure 4 The simulated echo diagram in the time domain is shown.

[0073] Figure 5 The simulated echo diagram in the TFD is shown.

[0074] Figure 6 The TFD diagram of the method provided by the present invention is shown.

[0075] Figure 7 The TFD diagram using CVMD is shown.

[0076] Figure 8 The TFD diagram using wavelet is shown.

[0077] Figure 9 The result diagram after pulse compression is shown.

[0078] Figure 10 The comparison diagram of signal power loss of different methods is shown.

[0079] Figure 11 The comparison diagram of residual JSNR of different methods is shown.

[0080] Figure 12 The comparison diagram of detection probability of different methods is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0081] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0082] As Figure 1 shown, an ISRJ suppression method based on multi-model fusion specifically includes the following steps:

[0083] S1. Construct an ISRJ signal model.

[0084] S2. Use a fusion interference detection method based on Transformer and decision tree to extract interference segments.

[0085] S3. On the basis of the extracted interference segments, use complex variational mode decomposition (CVMD) to extract the continuous component of ISRJ, and decompose the continuous component of ISRJ into band-limited intrinsic mode functions (BIMFs).

[0086] S4. Based on the estimated interference signal-to-noise ratio, select the interference signal and the target signal, and reconstruct the selected interference signal.

[0087] S5. Use multi-scale wavelet decomposition to reconstruct the interference signal.

[0088] S6. Fuse the interference signals reconstructed in steps S4 and S5.

[0089] S7. After using the interference signal obtained in step S6 to eliminate the interference of the original signal, use the sparse reconstruction algorithm to reconstruct the target signal, and then use the reconstructed target signal to eliminate the target signal in the original signal and reconstruct the interference signal. Alternately iterate in this way, that is, cycle this process to alternately reconstruct the interference signal and the target signal to improve the reconstruction accuracy of the interference signal.

[0090] Specifically, step S1 specifically includes the following steps:

[0091] S1.1. The jammer intercepts the radar signal and uses the "store-and-forward" method to send intermittent sampling and forwarding interference (ISRJ) through multiple signal segments. Assume that the radar transmit waveform is , which is modeled as a linear frequency modulation (LFM) signal; the ISRJ signal composed of segments is expressed as :

[0092] ;

[0093] Among them, is the interference amplitude after power amplification, is a rectangular pulse function, is the storage time of the th interference segment, is the length of the interference segment.

[0094] S1.2, after modulation, the ISRJ signal is expressed as :

[0095] ;

[0096] wherein, is the convolution operator, is the random modulation kernel.

[0097] S1.3, if there are targets in the detection scenario, the total received signal is expressed as :

[0098] ;

[0099] wherein, is the amplitude of the th target, is the time delay of the echo of the th target, is Gaussian white noise; after sampling, is expressed as a vector .

[0100] Based on the signal model: , a typical ISRJ is simulated in the time domain, as shown in Figure 2 . It can be observed therefrom that the ISRJ has strong coherence with the radar signal, complex modulation characteristics and high power, enabling it to effectively mask the real target signal and resulting in detection failure. In the time-frequency distribution (TFD) of the interference, as shown in Figure 3 , the modulation characteristics of continuous interference and pulse interference are mixed with each other, increasing the complexity of interference analysis. In this context, accurately characterizing the complex modulation characteristics is the key to achieving effective interference suppression.

[0101] Specifically, step S2 uses the method in <Intensive Interrupted Sampling Repeater Jamming Detection Based on Transformer CFAR Fusion Detection Model> (Detection of Intensive Interrupted Sampling Repeater Jamming Based on Transformer CFAR Fusion) to accurately locate the interference segments.

[0102] Specifically, based on the extracted interference segments, complex variational mode decomposition (CVMD) is used to extract the continuous components of ISRJ. CVMD regards the extracted interference as a band-limited signal with specific sparsity and decomposes it into band-limited intrinsic mode functions (BIMFs). BIMFs have the characteristics of smooth continuity and sparsity because band-limited Gaussian smoothing and time derivative minimization are introduced during the decomposition process. Step S3 specifically includes the following steps:

[0103] S3.1, Band-limited intrinsic mode functions BIMFs:

[0104] ;

[0105] ;

[0106] where is the center frequency, is the mode function, is the th BIMFs obtained by decomposition, is the number of BIMFs, is the target signal or observed signal, is the exponential function, is the function to take the minimum value.

[0107] S3.2, Solve the band-limited intrinsic mode functions BIMFs using the augmented Lagrangian method of the alternating direction method of multipliers ADMM.

[0108] Since the interference signal and the target signal are always mixed together, if all BIMFs are used for interference reconstruction, it will lead to the loss of the target signal. To solve this problem, based on the estimated interference signal-to-noise ratio in "Intensive Interrupted Sampling Repeater Jamming Detection Based on Transformer CFAR Fusion Detection Model", a BIMFs selection criterion is designed.

[0109] Specifically, step S4 specifically includes the following steps:

[0110] S4.1, The reconstructed interference signal using the selected BIMFs is expressed as:

[0111] ;

[0112] where is the support set of the selected BIMFs, estimated as:

[0113] ;

[0114] Among them, is the number of selected BIMFs components, is the signal length, are the powers of the signal and the noise, which are estimated simultaneously during the interference localization process. is the threshold factor, usually set between 1.5 and 2, is the square of the 2-norm of corresponding to the energy of the BIMFs component,

[0115] S4.2, solve using the greedy algorithm , the reconstructed interference signal is represented as a vector .

[0116] To improve the accuracy of the reconstructed interference, the multi-scale wavelet decomposition is also used to extract the characteristics of the pulse signal. In this step, the wavelet kernel after dilation and translation is used to decompose the signal at different scales.

[0117] Specifically, step S5 specifically includes the following steps:

[0118] S5.1, using the interference segments obtained in step S2, the interference wavelet coefficients with different dilation and translation factors are expressed as :

[0119] ;

[0120] Among them, is the complex conjugate of the mother wavelet, is the wavelet kernel with dilation factor and translation factor , expressed as:

[0121] ;

[0122] Among them, is the mother wavelet.

[0123] S5.2, design an adaptive threshold to separate the pulse interference component and the target signal component, as follows:

[0124] ;

[0125] ;

[0126] Among them, is the threshold factor related to JSRN, is the adjustment factor, is the processed wavelet coefficient, is a non - linear function used to adjust wavelet coefficients greater than the threshold λ. is a sign function used to restore the sign direction of the original wavelet coefficients.

[0127] S5.3, represent the impulse component extracted by wavelet as :

[0128] ;

[0129] The discrete digital signal of is represented as a vector .

[0130] Specifically, step S6 specifically includes the following steps:

[0131] S6.1, using CVMD and wavelet decomposition, the continuous smooth and impulse components are extracted as and . To obtain an accurate interference estimate, the following weighted fusion filter is proposed:

[0132] ;

[0133] where , are weights.

[0134] S6.2, let , and use the criterion of minimizing the residual square error to determine the variable to be solved, as follows:

[0135] ;

[0136] where represents the conjugate transpose of , is 's optimal solution.

[0137] is obtained through convex optimization:

[0138] ;

[0139] ;

[0140] ;

[0141] where and are intermediate variables, is the conjugate operation, is 's transpose.

[0142] S6.3, after the fusion filtering estimation, by performing interference cancellation and combining with the sparse reconstruction method to extract the target signal ,

[0143] ;

[0144] Among them, is the sparse dictionary related to the transmission waveform, is the penalty factor, is to solve the optimal solution of the target signal .

[0145] S6.4, solve by the conjugate gradient method.

[0146] Specifically, in step S7: In order to improve the interference reconstruction accuracy and the target detection probability, an alternating iteration method is proposed to estimate the interference signal and the target signal by iteration. Using the method provided by the present invention, the complex modulation of ISRJ can be effectively suppressed by fusion filtering and target detection can be performed by sparse recovery.

[0147] In order to verify the effect of the present invention, the following simulation experiments and analyses are carried out:

[0148] The ISRJ data is simulated to verify the effectiveness of the proposed method. The experiment is divided into two parts:

[0149] (1) The process of the method provided by the present invention is verified, and the single interference reconstruction methods using CVMD or wavelet are compared to prove the improvement achieved by multi-model fusion filtering and alternating iteration.

[0150] (2) The statistical performance of the method proposed by the present invention under different JSNRs is analyzed through Monte Carlo experiments, and the typical ISRJ suppression methods are compared to verify the performance improvement of the proposed method.

[0151] The radar and interference parameters are set as follows: The carrier frequency is 5 GHz. The bandwidth is 1 MHz. The sampling frequency is 2.8 MHz. The signal-to-noise ratio is 15 dB, and the JSNR is 15 to 35 dB. The simulated echoes of ISRJ are as Figure 4 and Figure 5 shown. It can be observed that the ISRJ with larger power covers the Figure 4 target signal and shows a complex time-frequency distribution (TFD) with continuous smooth and pulse characteristics in Figure 5 . Then, the method proposed by the present invention, as well as CVMD interference suppression and wavelet interference suppression, are applied to process these simulated interference data. The processing results are as Figures 6 - 9As shown. It can be seen that the proposed method effectively eliminates the interference and generates a clear time-frequency curve in the TFD, as Figure 6 shown. After pulse compression, the target can be detected. As Figure 7 and Figure 8 shown, due to the incomplete representation of the interference, the results using the CVMD and wavelet methods still show obvious interference residuals, and the target is still obscured by the residual interference after pulse compression. In addition, the residual JSNR (RJSNR), signal loss (SL), and interference reconstruction error (JRE) are quantitatively calculated in Table 1. The results show that the proposed method achieves the best reconstruction accuracy, the smallest interference residuals, and the smallest target signal loss. The significant improvement can be attributed to the proposed method, which uses a fusion filtering framework with multiple models to address the continuous smoothing and pulse characteristics and combines the alternating iteration method. This method improves the interference reconstruction accuracy and target detection performance.

[0152] Table 1 Performance Statistics

[0153]

[0154] This embodiment evaluates the interference suppression performance under different JSNR conditions. The JSNR is set between 5 dB and 35 dB, and 500 Monte Carlo experiments are conducted to statistically analyze the residual interference and signal loss after interference suppression. For comparison, Method 1: <Band pass filter design against interrupted-sampling repeater jamming based on time-frequency analysis>; Method 2: <Interrupted-sampling repeater jamming suppression with one-dimensional semi-parametric signal decomposition> and Method 3: <ISRJ Suppression Algorithm Based on Wavelet Transform and Compressed Sensing Reconstruction> are selected as the benchmark methods. The results are as Figures 10 - 12As shown, it indicates that the method proposed by the present invention is superior to other methods. This superiority is attributed to the fact that when the time-frequency filtering and wavelet filtering techniques cannot effectively distinguish between the target signal and the interference signal in the time-frequency domain when they overlap, resulting in a large interference residue. In addition, the interference cancellation method suppresses the interference signal by setting the interference area to zero, but at the same time it also suppresses the target signal. In addition, when the false alarm rate is 10 −5 , the target detection probability is as Figure 12 shown. The method proposed by the present invention maintains strong detection performance at different JSNR levels, further verifying its effectiveness in ISRJ suppression.

[0155] The present invention proposes an ISRJ suppression method based on multi-model fusion. This method can reconstruct and eliminate the interference signal with high precision for complex modulation interference. Simulation experiments show that this method has a significant improvement compared with traditional methods.

[0156] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the substantial scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. An ISRJ suppression method based on multi-model fusion, characterized in that: The specific steps include: S1, build ISRJ signal model; S2, extracts interference fragments using a fusion interference detection method based on Transformer and decision tree; S3, based on the extracted interference fragments, complex variational mode decomposition (CVMD) is used to extract the continuous components of ISRJ, and the continuous components of ISRJ are decomposed into band-limited intrinsic mode functions (BIMFs); S4, selecting an interference signal and a target signal based on the estimated interference signal-to-noise ratio, and reconstructing the selected interference signal; S5, reconstruct the interference signal using multi-scale wavelet decomposition; S6, fusing the interference signals reconstructed by steps S4 and S5; S7, using the interference signal obtained in step S6 to eliminate the interference of the original signal, reconstructing the target signal using a sparse reconstruction algorithm, and then using the reconstructed target signal to eliminate the target signal in the original signal and reconstruct the interference signal, and iterating alternately to reconstruct the interference signal and the target signal; Step S6 specifically includes the following steps: S6.1, the following weighted fusion filter is proposed : ； in, , is the weight, The reconstructed interference signal A discrete digital signal, is the pulse component extracted by wavelet Discrete digital signal; S6.2, order , using the criterion of minimizing the residual square error to determine the variables to be solved , as shown below: ； in, Express Taking the conjugate transpose, for The optimal solution of Through convex optimization, we get ； ； ； in, and is the intermediate variable, is the conjugate operation, for The transpose of S6.3, by Interference cancellation is performed and the target signal is extracted by combining the sparse reconstruction method. , ； in, is a sparse dictionary associated with the transmission waveform, is the penalty factor, To solve the target signal The optimal solution of S6.4, solved by the conjugate gradient method .

2. According to the ISRJ suppression method based on multi-model fusion according to claim 1, it is characterized in that: Step S1 specifically includes the following steps: S1.1, assuming that the radar transmit waveform is , is modeled as a linear frequency modulation LFM signal; The ISRJ signal composed of fragments is expressed as : ； in, is the interference amplitude after power amplification, is a rectangular pulse function, For the The storage time of the interference fragments, is the interference fragment length; S1.2, modulate each interference segment, and the modulated ISRJ signal is expressed as : ； in, is the convolution operator, is the random modulation kernel; S1.3, if there is targets, the total received signal is expressed as : ； in, For the The range of the target, For the The time delay of the target echo is is Gaussian white noise; after sampling, Represented as a vector .

3. According to claim 1, the ISRJ suppression method based on multi-model fusion is characterized in that: Step S3 specifically includes the following steps: S3.1, Band-limited intrinsic mode functions BIMFs: ； ； in, is the center frequency, is the modal function, The decomposition obtained BIMFs, is the number of BIMFs, is the target signal or the observed signal, is an exponential function, is the minimum value function; S3.2, the band-limited intrinsic mode functions BIMFs are solved using the augmented Lagrangian method with alternating direction multiplier method ADMM.

4. The ISRJ suppression method based on multi-model fusion according to claim 1, characterized in that: Step S4 specifically includes the following steps: S4.1, Interference signal reconstructed using selected BIMFs It is expressed as: ； in, is the support set of the selected BIMFs, estimated as: ； in, is the number of selected BIMFs components, is the signal length, are the power of the signal and the noise, is the threshold factor, for The square of the 2-norm of is the parameter minimum function; S4.2, using greedy algorithm to solve , the reconstructed interference signal The discrete digital signal is represented as a vector .

5. The ISRJ suppression method based on multi-model fusion according to claim 1, characterized in that: Step S5 specifically includes the following steps: S5.1, using the interference fragment obtained in step S2, the interference wavelet coefficients of different expansion and translation factors are expressed as : ； in, is the complex conjugate of the mother wavelet, Is the expansion factor and translation factor The wavelet kernel is expressed as: ； in, is the mother wavelet; S5.2, design an adaptive threshold to separate the pulse interference component and the target signal component as follows: ； ； in, is the threshold factor associated with JSRN, is the adjustment factor, is the processed wavelet coefficient, is a nonlinear function, is a sign function; S5.3, the pulse component extracted by wavelet is expressed as : ； The discrete digital signal is represented as a vector .

Images