Method and device for time-frequency statistical detection and suppression of radar active jamming

Abstract

The application provides a radar active jamming time-frequency statistical detection and suppression method and device, and relates to the technical field of radar anti-jamming. The method comprises the following steps: sequentially performing short-time Fourier transform, modulus value taking and logarithm processing on a radar echo signal to determine a time-frequency pixel matrix, so as to realize time-frequency analysis of the radar echo signal in each PRT; and performing positioning processing and time-domain reconstruction on the jamming signal in the time-frequency domain according to each estimated parameter, false alarm rate and detection threshold, so that the jamming signal can be determined for jamming suppression, without needing to search for a jamming area and accurately estimate parameters on a range-Doppler plane, so that jamming suppression can be performed when a range-Doppler plane target and jamming are highly overlapped; and the jamming signal can be sequentially subjected to positioning processing and time-domain reconstruction in the time-frequency domain, so that the division of a jamming signal area is realized, the use of a neural network for distinguishing the jamming signal is avoided, a large amount of jamming data set is not needed, and the method is suitable for engineering application.

Description

Technical Field

[0001] This invention relates to the field of radar anti-jamming technology, and in particular to a method and apparatus for time-frequency statistical detection and suppression of active radar interference. Background Technology

[0002] Radar, with its all-weather, long-range, and 24 / 7 capabilities, is widely used in target detection and guidance. Radar jamming is a significant factor in radar target detection, disrupting and interfering with its operation. Radar jamming can be categorized into active and passive jamming. Active jamming, in particular, is more targeted and destructive, posing a severe challenge to radar target detection. Because active jamming significantly reduces the signal-to-noise ratio of radar echo signals, hindering subsequent target detection, and because radar receivers struggle to eliminate active jamming, effective methods for suppressing active radar jamming at the receiver are crucial.

[0003] Currently, active radar jamming suppression is achieved through two main methods: One method involves searching for the jamming region and accurately estimating its parameters in the range-Doppler dimension; then, using the estimated parameters to design a Doppler filter bank for jamming suppression. However, if the target and jamming overlap in both the range and Doppler domains—meaning the jamming suppresses the target with high energy in both the range and Doppler dimensions—then using the estimated jamming parameters to design a Doppler filter bank for jamming suppression results in significant loss of target signal, affecting subsequent target signal processing. This method fails, rendering jamming suppression impossible. Another method involves testing a trained classification network using a test dataset to obtain labels; replacing the positions corresponding to jamming terms in the labels with random numbers to obtain the replaced data; and performing a short-time Fourier transform on the replaced data to obtain the jammed signal. However, using a classification network to distinguish jamming terms requires a large dataset as prior information, which is difficult to obtain in current radar operating environments. Therefore, this method is not suitable for engineering applications. Summary of the Invention

[0004] The purpose of this invention is to provide a method and apparatus for the time-frequency statistical detection and suppression of active radar interference, which solves the problems of existing technologies failing and being unable to suppress interference when the target and interference overlap in the range-Doppler domain, and the difficulty in obtaining a large amount of interference data in the current radar operating environment, making it unsuitable for engineering applications.

[0005] To address the aforementioned technical problems, the embodiments of the present invention provide the following technical solutions:

[0006] The first aspect of this invention provides a method for time-frequency statistical detection and suppression of active radar interference, the method comprising:

[0007] The radar echo signal is subjected to short-time Fourier transform, modulus value taking and logarithmic processing in sequence to determine the time-frequency pixel matrix. The radar echo signal includes interference signal.

[0008] A Gaussian mixture model is established for the time-frequency pixel matrix. The Gaussian mixture model includes multiple parameters.

[0009] Based on the expectation-maximization algorithm, the parameters in the Gaussian mixture model are estimated to obtain the corresponding estimated parameters;

[0010] Based on the estimated parameters, false alarm rate and detection threshold, the interference signal is sequentially processed in the time and frequency domains for localization and time domain reconstruction to obtain the reconstructed interference signal.

[0011] Time-domain cancellation is performed on the radar echo signal and the reconstructed interference signal to suppress the interference of the radar echo signal.

[0012] A second aspect of this application provides a radar active interference time-frequency statistical detection and suppression device, the device comprising:

[0013] The determination module is used to sequentially perform short-time Fourier transform, modulus value taking, and logarithmic processing on the radar echo signal to determine the time-frequency pixel matrix. The radar echo signal includes interference signals.

[0014] The module is used to build a Gaussian mixture model of the time-frequency pixel matrix. The Gaussian mixture model includes multiple parameters.

[0015] The estimation module is used to estimate the parameters in the Gaussian mixture model according to the expectation-maximization algorithm, and obtain the corresponding estimated parameters.

[0016] The localization and reconstruction module is used to perform localization processing in the time and frequency domain and time domain reconstruction on the interference signal in sequence according to the estimated parameters, false alarm rate and detection threshold, so as to obtain the reconstructed interference signal.

[0017] The time-domain cancellation module is used to perform time-domain cancellation on radar echo signals and reconstructed interference signals to achieve interference suppression of radar echo signals.

[0018] Compared to existing technologies, the radar active interference time-frequency statistical detection and suppression method and apparatus provided by this invention sequentially performs short-time Fourier transform, modulus value taking, and logarithmic processing on the radar echo signal to determine the time-frequency pixel matrix, thereby realizing PRT-by-Phase time-frequency analysis of the radar echo signal. Based on various estimated parameters, false alarm rate, and detection threshold, the interference signal can be located and reconstructed in the time-frequency domain, allowing for interference suppression without the need for range Doppler search and precise parameter estimation. This enables interference suppression even when the range Doppler target and interference highly overlap. Furthermore, the method sequentially locates and reconstructs the interference signal in the time-frequency domain to delineate the region where the interference signal is located, avoiding the use of neural networks to distinguish interference signals. It also eliminates the need for large interference datasets, making it suitable for engineering applications. Attached Figure Description

[0019] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, with the same or corresponding reference numerals denoteing the same or corresponding parts, wherein:

[0020] Figure 1 The flowchart of the time-frequency statistical detection and suppression method for radar active interference is illustrated schematically. Figure 1 ;

[0021] Figure 2 The time-frequency diagram of intermittent sampling and forwarding interference with a single pulse repetition time (PRT) is schematically shown.

[0022] Figure 3 The flowchart of the time-frequency statistical detection and suppression method for radar active interference is illustrated schematically. Figure 2 ;

[0023] Figure 4 The time-frequency diagram of interference localization for single PRT intermittent sampling forwarding interference is schematically shown;

[0024] Figure 5 A schematic diagram of a 3D moving target detection (MTD) without interference is shown.

[0025] Figure 6 A schematic diagram of the MTD (Mean Transmission Target) under disturbance is shown.

[0026] Figure 7 A schematic diagram of the MTD after interference suppression is shown.

[0027] Figure 8A schematic diagram of pulse compression without interference is shown.

[0028] Figure 9 A schematic diagram of pulse compression under interference is shown.

[0029] Figure 10 A schematic diagram of pulse compression after interference suppression is shown.

[0030] Figure 11 The schematic diagram shows the structure of a radar active interference time-frequency statistical detection and suppression device. Detailed Implementation

[0031] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art.

[0032] It should be noted that, unless otherwise stated, the technical or scientific terms used in this invention should have the ordinary meaning as understood by one of ordinary skill in the art.

[0033] The methods described in the embodiments of the present invention will be explained in detail below.

[0034] Figure 1 The flowchart of the radar active interference time-frequency statistical detection and suppression method in an embodiment of the present invention is illustrated schematically. Figure 1 See Figure 1 As shown, the method may include:

[0035] S101. Perform short-time Fourier transform, modulus taking, and logarithmic processing on the radar echo signal in sequence to determine the time-frequency pixel matrix.

[0036] Radar echo signals include interference signals, target signals, and background noise.

[0037] The radar echo signal x of length H can be processed sequentially by performing Short-Time Fourier Transform (STFT), modulus taking, and logarithmic transformation to obtain the time-frequency pixel matrix y. m,n , where m is the number of frequency sampling points and n is the number of distance sampling points.

[0038] S102. Establish a Gaussian mixture model for the time-frequency pixel matrix.

[0039] The Gaussian mixture model includes multiple parameters.

[0040] Several parameters include the weights of the mixed components, the mean of the mixed components, and the variance of the mixed components.

[0041] S103. Based on the expectation-maximization algorithm, estimate the parameters in the Gaussian mixture model to obtain the corresponding estimated parameters.

[0042] The Expectation-Maximum (EM) algorithm is used to iteratively estimate the parameters in the Gaussian mixture model of the time-frequency pixel matrix, yielding the corresponding estimated parameters. These estimated parameters are the final parameters obtained from the iterative estimation and include the weights, mean, and variance of the estimated mixture components.

[0043] S104. Based on the estimated parameters, false alarm rate and detection threshold, the interference signal is sequentially processed in the time and frequency domains for localization and in the time domain for reconstruction to obtain the reconstructed interference signal.

[0044] Radar echo signals are present in the time-frequency images of intermittent sampling and forwarding interference. Figure 2 The diagram schematically illustrates the time-frequency plot of intermittent sampling and forwarding interference using a single-pulse repetitive periodic PRT. The horizontal axis represents the distance sampling point, the vertical axis represents the frequency sampling point, the bright yellow area represents the interference signal area, the light green area represents the short-time Fourier transform energy leakage area, and the long blue diagonal line represents the target signal. It can be observed that the interference signal, the target signal, and the background noise are severely coupled in the time-frequency domain, which seriously affects the accuracy of interference parameter extraction. Therefore, it is necessary to perform localization processing on the interference signal in the time-frequency domain.

[0045] S105. Perform time-domain cancellation on the radar echo signal and the reconstructed interference signal to achieve interference suppression of the radar echo signal.

[0046] The radar echo signal and the reconstructed interference signal are canceled in the time domain to obtain the radar echo signal after interference suppression, so as to achieve interference suppression of the radar echo signal.

[0047] Based on the above Figure 1As can be seen from the implementation method, the embodiments of the present invention sequentially perform short-time Fourier transform, modulus taking, and logarithmic processing on the radar echo signal to determine the time-frequency pixel matrix. The radar echo signal includes interference signals. A Gaussian mixture model of the time-frequency pixel matrix is ​​established, which includes multiple parameters. According to the expectation-maximization algorithm, each parameter in the Gaussian mixture model is estimated to obtain the corresponding estimated parameters. Based on each estimated parameter, false alarm rate, and detection threshold, the interference signal is sequentially subjected to time-frequency domain localization processing and time-domain reconstruction to obtain the reconstructed interference signal. The radar echo signal and the reconstructed interference signal are time-domain canceled to achieve interference suppression of the radar echo signal. In this way, by sequentially performing short-time Fourier transform, modulus taking, and logarithmic transformation on the radar echo signal, the time-frequency pixel matrix is ​​determined, enabling PRT-based time-frequency analysis of the radar echo signal. Based on the estimated parameters, false alarm rate, and detection threshold, the interference signal can be located and reconstructed in the time-frequency domain, allowing for interference suppression without the need for range Doppler region search and precise parameter estimation. This enables interference suppression even when the range Doppler target and interference highly overlap. The sequential location and reconstruction of the interference signal in the time-frequency domain allows for the division of the interference signal's region, avoiding the use of neural networks to distinguish interference signals. It also eliminates the need for large interference datasets, making it suitable for engineering applications.

[0048] As a refinement and extension of the above embodiments, Figure 3 The following is a flowchart of a radar active interference time-frequency statistical detection and suppression method in an embodiment of the present invention. Figure 2 See Figure 3 As shown in the figure, the radar active interference time-frequency statistical detection and suppression method provided by the embodiment of the present invention may include:

[0049] S301. Perform short-time Fourier transform, modulus taking, and logarithmic processing on the radar echo signal in sequence to determine the time-frequency pixel matrix.

[0050] Radar echo signals include interference signals, target signals, and background noise.

[0051] Let M and N be the number of frequency sampling points and the number of range sampling points, respectively. Perform STFT on the radar echo signal x of length H to obtain the time-frequency complex matrix tfr. m,n Then, the modulus of the time-frequency complex matrix is ​​taken and logarithmic transformation is performed to obtain the time-frequency pixel matrix tfr. m,n Where m is the number of frequency sampling points, n is the number of distance sampling points, 1≤m≤M, 1≤n≤N, 1≤n≤N, N≥H, in this embodiment M=7658, N=64.

[0052] Specifically, the radar echo signal is sequentially subjected to short-time Fourier transform, modulus taking, and logarithmic transformation to determine the time-frequency pixel matrix, including:

[0053] Step A1: Perform a short-time Fourier transform on the radar echo signal to obtain the time-frequency complex matrix corresponding to the radar echo signal.

[0054] Specifically, the short-time Fourier transform of the radar echo signal of length H is performed using the following formula to calculate the time-frequency complex matrix tfr of the radar echo signal. m,n :

[0055]

[0056] Where ζ represents the summation factor, h represents the analysis window function, exp(·) represents the exponential function, m is the number of frequency sampling points, n is the number of distance sampling points, x is the radar echo signal, and s is the complex unit. In this embodiment of the invention, a Hamming window is used because the Hamming window has a large main lobe width and small side lobes, which can effectively suppress high-frequency interference, reduce spectral leakage, and has high spectral resolution.

[0057] Step A2: Using the following first formula, successively take the modulus and logarithm of the time-frequency complex matrix to obtain the time-frequency pixel matrix:

[0058] y m,n =10*log 10 [abs(tfr m,n )+1];

[0059] Among them, y m,n This is a time-frequency pixel matrix, where m is the number of frequency sampling points, n is the number of distance sampling points, and log... 10 (·) is the logarithmic function to base 10, abs(·) is the modulus function, and tfr m,n Let be a time-frequency complex matrix. The increment operation in the first formula above is to avoid an illegal value where the logarithm does not exist.

[0060] S302. Establish a Gaussian mixture model for the time-frequency pixel matrix.

[0061] The Gaussian mixture model includes multiple parameters.

[0062] Specifically, based on the number of components K of the Gaussian mixture distribution, the following second formula is used to establish a Gaussian mixture model of the time-frequency pixel matrix:

[0063]

[0064] Wherein, P(y m,n |Θ) represents the Gaussian mixture model of the time-frequency pixel matrix, y m,nThis is a time-frequency pixel matrix, where m is the number of frequency sampling points, n is the number of distance sampling points, K is the number of components in the Gaussian mixture distribution, k is the number of components in the Gaussian mixture distribution, and ω is the value of ω. k Let Θ be the weight of the k-th mixture component, where Θ = (θ1, θ2, ..., θ) K ) represents the parameter vector of each mixture component, φ k For the k-th mixture component, θ k =(μ k ,σ 2 k ) is the parameter vector of a single mixed component, μ k Let σ be the mean of the k-th mixture component. 2 k Let be the variance of the k-th mixture component.

[0065] S303. Based on the expectation-maximization algorithm, estimate the parameters in the Gaussian mixture model to obtain the corresponding estimated parameters.

[0066] Set the initial parameters of the Gaussian mixture model, and use the EM algorithm to iteratively estimate the parameters of the Gaussian mixture model of the time-frequency pixel matrix until the difference between the parameter vectors of two consecutive iterations is small enough, i.e., less than the convergence threshold, and then stop to obtain the values ​​of each parameter.

[0067] Specifically, based on the expectation-maximization algorithm, the parameters in the Gaussian mixture model are estimated to obtain the corresponding estimated parameters, including:

[0068] Step B1: Sort the time-frequency pixel matrix by column to obtain multiple column data of the time-frequency pixel matrix.

[0069] Specifically, the time-frequency pixel matrix is ​​sorted by column to obtain multiple columns of data for the time-frequency pixel matrix, including:

[0070] Using the third formula below, the time-frequency pixel matrix is ​​sorted by column to obtain multiple column data of the time-frequency pixel matrix:

[0071]

[0072] Among them, y sort For multiple columns of data in the time-frequency pixel matrix, y m,n This is a time-frequency pixel matrix. `reshape(·)` reshapes the matrix column-wise. `m` represents the number of frequency sampling points, `n` represents the number of distance sampling points, `M` represents the number of frequency sampling points, `N` represents the number of distance sampling points, `J` represents the number of data points in each column, and `y1` represents the first column of the time-frequency pixel matrix. J This represents the data in the Jth column of the time-frequency pixel matrix.

[0073] Arranging the time-frequency pixel matrix into a single column facilitates iterative calculations by the EM algorithm.

[0074] Step B2: Based on the clustering algorithm, cluster the data in each column of the time-frequency pixel matrix to obtain the initial parameters.

[0075] Specifically, the K-means clustering algorithm is used to perform preliminary clustering on each column of the time-frequency pixel matrix to obtain the initial values ​​of the Gaussian mixture model parameters:

[0076] [ω k,0 ,μ k,0 ,σ 2 k,0 ] = Kmean(y sort );

[0077] Where, ω k,0 ,μ k,0 ,σ 2 k,0 ω represents the initial parameters of the k-th mixture component. k,0 μ is the weight of the initial mixed component. k,0 σ is the mean of the initial mixed components. 2 k,0 y is the variance of the initial mixed components. sort The data consists of multiple columns of the time-frequency pixel matrix. Kmean represents the K-means clustering algorithm. Because the K-means algorithm is simple in principle, easy to implement, has a fast convergence speed, and relatively good clustering effect, its clustering results can be used as the initial parameters of the Gaussian mixture model.

[0078] Step B3: Calculate the corresponding preset responsivity based on the initial parameters, the mixed components corresponding to the initial parameters, and the observed data in each column, and calculate the parameters for the current round based on the preset responsivity.

[0079] Specifically, using the following formula, the corresponding preset responsivity γ is calculated based on the initial parameters, the mixed components corresponding to the initial parameters, and the observed data in each column of data. j，k :

[0080]

[0081] Where K is the number of components in the Gaussian mixture distribution, k is the possible values ​​for the number of components in the Gaussian mixture distribution, j is the number of data in each column, and ω k,0 φ represents the initial weights of the mixed components. k For the k-th mixture component, θ k,0 =(μ k,0 ,σ 2 k,0 ) is the parameter vector of a single mixed component, μ k,0 σ is the mean of the initial mixed components. 2 k,0The variance of the initial mixed components.

[0082] The parameters of the current round include ω k ,μ k ,σ 2 k ω k μ represents the weight of the current round's mixed components. k σ is the mean of the current round's mixed components. 2 k This represents the variance of the current round's mixed components.

[0083] The parameters for the current round are calculated using the following formula based on the preset responsiveness:

[0084]

[0085] Where, γ j，k To preset the response, y j For the observed data, J represents the number of data points in each column.

[0086] Step B4: Calculate the response of the next round based on the parameters of the current round, the mixed components corresponding to the parameters of the current round, and the observation data.

[0087] Specifically, using the following formula, based on the parameter ω of the current wheel... k ,μ k ,σ 2 k The response of the next round is calculated using the mixed components corresponding to the parameters of the current round and the observed data. That is, calculate the k-th mixture component for the observed data y. j The responseness, which reflects the observed data y j The probability of belonging to the k-th mixed component:

[0088]

[0089] Where K is the number of components in the Gaussian mixture distribution, k is the possible values ​​for the number of components in the Gaussian mixture distribution, j is the number of data in each column, and ω k φ represents the weight of the k-th mixture component, which is also the weight of the mixture component in the current round. k For the k-th mixture component, θ k =(μ k ,σ 2 k ) is the parameter vector of a single mixed component, μ k Let σ be the mean of the k-th mixture component, which is also the mean of the mixture components in the current round. 2 k The variance of the k-th mixture component is the variance of the mixture component in the current round.

[0090] Step B5: Calculate the parameters for the next round based on the response rate of the next round.

[0091] The parameters for the next round include the weights of the next round's mixed components, the mean of the next round's mixed components, and the variance of the next round's mixed components.

[0092] Specifically, using the following formula, based on the response rate in the next round... Calculate the parameters for the next round.

[0093]

[0094] in, The weights for the next round of mixing components. This is the mean of the next round of mixing components. Let J be the variance of the next round of mixed components, and J be the number of data in each column.

[0095] Step B6: Repeat steps B4 and B5 until the parameters in the next round are less than the convergence threshold, then stop the loop and obtain the estimated parameters.

[0096] The estimated parameters include the weights of the estimated mixture components, the mean of the estimated mixture components, and the variance of the estimated mixture components. The variance of the estimated mixture components is the square of the standard deviation of the estimated mixture components.

[0097] Specifically, repeat steps B4 and B5 until the mean vector of the next round of mixed components and the mean vector of the current round of mixed components, as well as the variance vector of the next round of mixed components and the variance vector of the current round of mixed components, are all less than the convergence threshold. Then stop the loop and obtain the estimated parameters.

[0098] The convergence threshold ranges from 10e-3 to 10e-6, and in this embodiment of the invention, a convergence threshold of 10e-6 is selected.

[0099] Steps S304-S306 below are specific operations for performing time-frequency domain localization processing and time-domain reconstruction on the interference signal according to each estimated parameter, false alarm rate and detection threshold, to obtain the reconstructed interference signal.

[0100] Set the false alarm rate P fa And calculate the detection threshold P fa Using the binary decision principle, interference is assessed for each pixel in the time-frequency pixel matrix, resulting in a time-frequency complex matrix containing only interference information. Then, the interference signal is reconstructed in the time domain by inverse short-time Fourier transform, and the reconstructed interference signal J is calculated. recov .

[0101] S304. Determine the detection threshold based on the estimated weights of the mixed components, the estimated mean of the mixed components, the estimated standard deviation of the mixed components, and the false alarm rate.

[0102] Specifically, the detection threshold is determined based on the estimated weights of the mixed components, the estimated mean of the mixed components, the estimated standard deviation of the mixed components, and the false alarm rate, including:

[0103] Step C1: Based on the estimated weights, mean, and standard deviation of the mixed components, and the observed data, determine the relationship between the detection threshold and the false alarm rate using the following fourth formula:

[0104]

[0105] Among them, P fa ω represents the false alarm rate, I represents the detection threshold, and ω represents the false alarm rate. k1 To estimate the weights of the mixed components, μ k1 To estimate the mean of the mixed components, σ k1 To estimate the standard deviation of the mixed components, y j For observational data.

[0106] After step C1, based on the properties of the normal distribution, we can conclude... It follows the mean ω1μ1+ω2μ2+…+ω K μ K The variance is (ω1σ1). 2 +(ω2σ2) 2 +…+(ω K σ K ) 2 Since the detection threshold follows a normal distribution, the relationship between the detection threshold and the false alarm probability can be written as:

[0107]

[0108] set up The above relation can also be written as:

[0109]

[0110] make The above relation can also be written as:

[0111]

[0112] in, This represents the complement function of the error function, so the fifth formula for the detection threshold I in step C2 below can be calculated.

[0113] Step C2: Based on the relationship and the normal distribution of the false alarm probability, determine the detection threshold using the following fifth formula:

[0114]

[0115] Among them, erfc -1 (·) is the inverse function of the complement of the error function, and erfc(·) is the complement of the error function.

[0116] S305. Based on the time-frequency pixel matrix and the detection threshold, the time-frequency complex matrix is ​​located in the time-frequency domain using the binary decision method to obtain the interference time-frequency complex matrix.

[0117] Among them, the interference time-frequency complex matrix is ​​a time-frequency complex matrix that contains only the interference signal.

[0118] Specifically, based on the time-frequency pixel matrix and the detection threshold, a binary decision method is used to locate the time-frequency complex matrix in the time-frequency domain, resulting in the interference time-frequency complex matrix, including:

[0119] Based on the time-frequency pixel matrix and the detection threshold, the interference time-frequency complex matrix is ​​processed in the time-frequency domain using the binary decision method and the sixth formula below, to obtain the interference time-frequency complex matrix:

[0120]

[0121] in, For the interference time-frequency complex matrix, tfr m,n Let y be a time-frequency complex matrix. m,n Let I be the time-frequency pixel matrix, where I is the detection threshold, m is the number of frequency sampling points, and n is the number of distance sampling points.

[0122] S306. Based on the inverse short-time Fourier transform, the time-domain reconstruction of the interference time-frequency complex matrix is ​​performed to obtain the reconstructed interference signal.

[0123] Specifically, using the following formula, the time-domain reconstruction of the interference time-frequency complex matrix is ​​performed according to the inverse short-time Fourier transform to obtain the reconstructed interference signal J. recov :

[0124]

[0125] in, The interference time-frequency complex matrix is ​​given by M, where M is the number of frequency sampling points, m is the possible values ​​for the number of frequency sampling points, and n is the possible values ​​for the number of distance sampling points. This indicates that the sampling point pair is M with frequency M as the reference. Performing the inverse short-time Fourier transform yields a one-dimensional vector with G points, where R represents the Fourier transform of the window function and δ represents the integration factor.

[0126] S307. Time-domain cancellation is performed on the radar echo signal and the reconstructed interference signal to achieve interference suppression of the radar echo signal.

[0127] Specifically, time-domain cancellation is performed on the radar echo signal and the reconstructed interference signal to achieve interference suppression of the radar echo signal, including:

[0128] In the time domain, subtract the reconstructed interference signal J from the radar echo signal x. recov The interference-suppressed radar echo signal Anti_jamming is obtained to achieve interference suppression of the radar echo signal. The calculation formula is as follows:

[0129] Anti_jamming=xJ recov .

[0130] This invention performs point-to-point (PRT) time-frequency analysis and interference localization on radar echo signals, eliminating the need for interference region search and precise parameter estimation in the range Doppler domain. This avoids the problem of interference suppression when the target and interference highly overlap in the range Doppler domain, facilitating subsequent target detection. Furthermore, it only requires iterative calculations on the three parameters of the Gaussian mixture model, resulting in lower computational complexity compared to existing technologies and easier engineering applications. By utilizing a multi-component Gaussian mixture model to binarize the target and interference regions in the time-frequency domain on a PRT basis, it avoids the need for neural networks to distinguish interference terms. Therefore, this invention does not require large datasets as prior data, making it easy to implement and apply in engineering.

[0131] The technical effects of the present invention will be further explained below based on the actual test results:

[0132] Experimental conditions and contents:

[0133] The hardware test platform used in the simulation of this invention is as follows: CPU Intel Xeon(R) E5-1630 V4 with a main frequency of 3.70GHz, memory 64GB, graphics card NVIDIA GeForce RTX 2080Ti, and software platform Windows 7 64-bit operating system, MATLAB R2018a. The radar transmission parameters used in the simulation are shown in Table 1.

[0134] Table 1 shows the radar transmission parameters used in the simulation.

[0135]

[0136]

[0137] Because intermittent sampling and relay jamming has both deceptive and suppressive effects on radar systems, this invention uses it as an example of jamming. Figure 4The diagram schematically illustrates the time-frequency localization of interference in a single PRT intermittent sampling forwarding interference. The horizontal axis represents the number of distance sampling points, and the vertical axis represents the number of frequency sampling points. It can be seen that the time-frequency localization method for interference proposed in this invention can effectively locate the interference signal and the energy leakage generated by the short-time Fourier transform, providing a basis for the time-domain reconstruction and suppression of interference. Figure 5 The diagram schematically shows the 3D MTD plot when there is no interference. The horizontal axis represents the distance dimension, the vertical axis represents the Doppler dimension, and the height represents the amplitude. It can be seen that there is a peak when there is no interference, which is the location of the target. Figure 6 The diagram schematically illustrates a three-dimensional MTD plot under disturbance, with the horizontal axis representing the distance dimension, the vertical axis representing the Doppler dimension, and the height representing the amplitude. Figure 5 As can be seen, the target is covered by a large number of interference segments, causing moving target detection to fail and making further signal processing impossible. Figure 7 The diagram schematically illustrates a 3D plot of MTD after interference suppression. The horizontal axis represents the distance dimension, the vertical axis represents the Doppler dimension, and the height represents the amplitude. It can be seen that after interference suppression, compared to... Figure 5 At the same Doppler and distance positions, the target signal is well preserved, while the interference signal is suppressed. Although the target signal is somewhat lost, it can still lay the foundation for further target detection. Figure 8 The diagram schematically illustrates the pulse compression when there is no interference. The horizontal axis represents the distance dimension, and the vertical axis represents the amplitude. It can be seen that there is a spike at position 2440, which is the target signal. Figure 9 The diagram schematically illustrates the pulse compression pattern under interference, with the horizontal axis representing the distance dimension and the vertical axis representing the amplitude. It can be seen that intermittent sampling interference forms a large number of false target clusters at the pulse compression end, which has both suppression and deception effects, and has a strong interference capability. Figure 10 The diagram schematically illustrates the pulse compression after interference suppression, with the horizontal axis representing the distance dimension and the vertical axis representing the amplitude. It can be seen that the target at position 2440 is well preserved, while the interference signal at the interference position is well suppressed.

[0138] Analysis of measured results:

[0139] In the later embodiments of the present invention, by Figure 7 It can be seen that the target is located at a distance of 2440.

[0140] according to Figure 8 and Figure 9 It can be seen that after interference suppression, the Jammer to Signal Ratio (JSR) can reach 39dB, which has a better suppression effect on interference and is more conducive to engineering applications.

[0141] Based on the same inventive concept, as an implementation of the above-mentioned method for time-frequency statistical detection and suppression of radar active interference, this embodiment of the invention also provides a device for time-frequency statistical detection and suppression of radar active interference. Figure 11 This is a structural diagram of the device in an embodiment of the present invention. See also: Figure 11 As shown, the device may include:

[0142] The determination module 1101 is used to sequentially perform short-time Fourier transform, modulus value taking and logarithmic processing on the radar echo signal to determine the time-frequency pixel matrix. The radar echo signal includes interference signals.

[0143] Module 1102 is established to establish a Gaussian mixture model of the time-frequency pixel matrix. The Gaussian mixture model includes multiple parameters.

[0144] The estimation module 1103 is used to estimate the parameters in the Gaussian mixture model according to the expectation-maximization algorithm, and obtain the corresponding estimated parameters.

[0145] The positioning and reconstruction module 1104 is used to perform positioning processing in the time and frequency domain and time domain reconstruction on the interference signal in sequence according to each estimated parameter, false alarm rate and detection threshold, so as to obtain the reconstructed interference signal.

[0146] The time-domain cancellation module 1105 is used to perform time-domain cancellation on the radar echo signal and the reconstructed interference signal to achieve interference suppression of the radar echo signal.

[0147] The determination module 1101 is specifically used to perform a short-time Fourier transform on the radar echo signal to obtain the time-frequency complex matrix corresponding to the radar echo signal;

[0148] Using the following first formula, the time-frequency complex matrix is ​​successively moduloed and logarithmized to obtain the time-frequency pixel matrix:

[0149] y m,n =10*log 10 [abs(tfr m,n )+1];

[0150] Among them, y m,n This is a time-frequency pixel matrix, where m is the number of frequency sampling points, n is the number of distance sampling points, and log... 10 (·) is the logarithmic function to base 10, abs(·) is the modulus function, and tfr m,n It is a time-frequency complex matrix.

[0151] Module 1102 is specifically used to establish a Gaussian mixture model of the time-frequency pixel matrix based on the number of components of the Gaussian mixture distribution, using the following second formula:

[0152]

[0153] Wherein, P(y m,n |Θ) represents the Gaussian mixture model of the time-frequency pixel matrix, y m,n This is a time-frequency pixel matrix, where m is the number of frequency sampling points, n is the number of distance sampling points, K is the number of components in the Gaussian mixture distribution, k is the number of components in the Gaussian mixture distribution, and ω is the value of ω. k Let Θ be the weight of the k-th mixture component, where Θ = (θ1, θ2, ..., θ) K ) represents the parameter vector of each mixture component, φ k For the k-th mixture component, θ k =(μ k ,σ 2 k ) is the parameter vector of a single mixed component, μ k Let σ be the mean of the k-th mixture component. 2 k Let be the variance of the k-th mixture component.

[0154] The estimation module 1103 is specifically used to sort the time-frequency pixel matrix by column to obtain multiple column data of the time-frequency pixel matrix; to cluster the column data of the time-frequency pixel matrix according to the clustering algorithm to obtain initial parameters; to calculate the corresponding preset responsivity based on the initial parameters, the corresponding mixture components of the initial parameters, and the observed data in each column data, and to calculate the parameters of the current round based on the preset responsivity; the next round of responsivity calculation steps: to calculate the responsivity of the next round based on the parameters of the current round, the corresponding mixture components of the parameters of the current round, and the observed data; the next round of parameter calculation steps: to calculate the parameters of the next round based on the responsivity of the next round; repeating the next round of responsivity calculation steps and the next round of parameter calculation steps until the parameters of the next round are less than the convergence threshold, at which point the loop stops and the estimated parameters are obtained.

[0155] The estimation module 1103 sorts the time-frequency pixel matrix by column to obtain multiple column data of the time-frequency pixel matrix, including:

[0156] Using the third formula below, the time-frequency pixel matrix is ​​sorted by column to obtain multiple column data of the time-frequency pixel matrix:

[0157]

[0158] Among them, y sort For multiple columns of data in the time-frequency pixel matrix, y m,n This is a time-frequency pixel matrix. `reshape(·)` reshapes the matrix column-wise. `m` represents the number of frequency sampling points, `n` represents the number of distance sampling points, `M` represents the number of frequency sampling points, `N` represents the number of distance sampling points, `J` represents the number of data points in each column, and `y1` represents the first column of the time-frequency pixel matrix. JThis represents the data in the Jth column of the time-frequency pixel matrix.

[0159] The positioning and reconstruction module 1104 is specifically used to determine the detection threshold based on the estimated weights, mean, standard deviation, and false alarm rate of the mixed components; based on the time-frequency pixel matrix and the detection threshold, it uses a binary decision method to perform positioning processing on the time-frequency complex matrix in the time-frequency domain to obtain the target time-frequency complex matrix, which is a time-frequency complex matrix containing only the interference signal; based on the inverse short-time Fourier transform, it performs time-domain reconstruction on the target time-frequency complex matrix to obtain the reconstructed interference signal; the estimation parameters include the estimated weights, mean, and variance of the mixed components, and the variance of the mixed components is the square of the standard deviation of the mixed components.

[0160] The positioning reconstruction module 1104 determines the detection threshold based on the estimated weights, mean, standard deviation, and false alarm rate of the mixed components, including:

[0161] Based on the estimated weights, mean, and standard deviation of the mixed components, and the observed data, the relationship between the detection threshold and the false alarm rate is determined using the following fourth formula:

[0162]

[0163] Among them, P fa ω represents the false alarm rate, I represents the detection threshold, and ω represents the false alarm rate. k1 To estimate the weights of the mixed components, μ k1 To estimate the mean of the mixed components, σ k1 To estimate the standard deviation of the mixed components, y j For observational data;

[0164] Based on the relationship and the normal distribution of the false alarm probability, the detection threshold is determined using the following fifth formula:

[0165]

[0166] Among them, erfc -1 (·) is the inverse function of the complement of the error function, and erfc(·) is the complement of the error function.

[0167] The localization and reconstruction module 1104, based on the time-frequency pixel matrix and the detection threshold, uses a binary decision method to perform localization processing on the time-frequency complex matrix in the time-frequency domain, obtaining the interference time-frequency complex matrix, including:

[0168] Based on the time-frequency pixel matrix and the detection threshold, the interference time-frequency complex matrix is ​​processed in the time-frequency domain using the binary decision method and the sixth formula below, to obtain the interference time-frequency complex matrix:

[0169]

[0170] in, For the interference time-frequency complex matrix, tfr m,n Let y be a time-frequency complex matrix. m,n Let I be the time-frequency pixel matrix, where I is the detection threshold, m is the number of frequency sampling points, and n is the number of distance sampling points.

[0171] The time-domain cancellation module 1105 is specifically used to subtract the reconstructed interference signal from the radar echo signal in the time domain to obtain the radar echo signal after interference suppression, so as to achieve interference suppression of the radar echo signal.

[0172] It should be noted that the above description of the radar active interference time-frequency statistical detection and suppression device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects. For technical details not disclosed in the embodiments of the radar active interference time-frequency statistical detection and suppression device of the present invention, please refer to the description of the method embodiment of the present invention for understanding.

[0173] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for time-frequency statistical detection and suppression of active radar interference, characterized in that, The method includes: The radar echo signal is subjected to short-time Fourier transform, modulus taking, and logarithmic processing in sequence to determine the time-frequency pixel matrix. The radar echo signal includes interference signals. A Gaussian mixture model is established for the time-frequency pixel matrix. The Gaussian mixture model includes multiple parameters, including the weights of the mixture components, the mean of the mixture components, and the variance of the mixture components. Based on the expectation-maximization algorithm, the parameters in the Gaussian mixture model are estimated to obtain the corresponding estimated parameters. Based on the estimated parameters, false alarm rate and detection threshold, the interference signal is sequentially subjected to localization processing in the time and frequency domain and time domain reconstruction to obtain the reconstructed interference signal. The radar echo signal and the reconstructed interference signal are time-domain canceled to suppress interference in the radar echo signal.

2. The method according to claim 1, characterized in that, The process of sequentially performing short-time Fourier transform, modulus taking, and logarithmic transformation on the radar echo signal to determine the time-frequency pixel matrix includes: Perform a short-time Fourier transform on the radar echo signal to obtain the time-frequency complex matrix corresponding to the radar echo signal; Using the following first formula, the time-frequency complex matrix is ​​successively moduloed and logarithmized to obtain the time-frequency pixel matrix: ； in, The time-frequency pixel matrix, The number of frequency sampling points is taken as the value. The value is taken as the distance to the number of sampling points. It is a logarithmic function with base 10. For modulo function, Let be the time-frequency complex matrix.

3. The method according to claim 2, characterized in that, The establishment of the Gaussian mixture model for the time-frequency pixel matrix includes: Based on the number of components in the Gaussian mixture distribution, the Gaussian mixture model of the time-frequency pixel matrix is ​​established using the following second formula: ； in, The Gaussian mixture model of the time-frequency pixel matrix. The time-frequency pixel matrix, The number of frequency sampling points is taken as the value. The value is taken as the distance to the number of sampling points. The number of components in the Gaussian mixture distribution. The number of components in the Gaussian mixture distribution is taken as a value. For the first The weights of the mixed components, For each mixture component, (This refers to the parameter vector.) For the first A mixed component, For a single mixed component, the parameter vector, For the first The mean of the mixed components, For the first The variance of each mixed component.

4. The method according to claim 3, characterized in that, The parameters in the Gaussian mixture model are estimated using the expectation-maximization algorithm to obtain the corresponding estimated parameters, including: The time-frequency pixel matrix is ​​sorted by column to obtain multiple column data of the time-frequency pixel matrix; According to the clustering algorithm, the data in each column of the time-frequency pixel matrix are clustered to obtain the initial parameters; Based on the initial parameters, the mixed components corresponding to the initial parameters, and the observed data in each column of data, calculate the corresponding preset responsivity, and calculate the parameters of the current round based on the preset responsivity; The next round of response calculation steps: Calculate the response of the next round based on the parameters of the current round, the mixed components corresponding to the parameters of the current round, and the observed data; The next round of parameter calculation steps: Calculate the parameters for the next round based on the response rate of the next round; Repeat the response calculation step and the parameter calculation step in the next round until the parameter in the next round is less than the convergence threshold, then stop the loop and obtain the estimated parameters.

5. The method according to claim 4, characterized in that, The step of sorting the time-frequency pixel matrix by column to obtain multiple column data of the time-frequency pixel matrix includes: Using the third formula below, the time-frequency pixel matrix is ​​sorted column-wise to obtain multiple column data of the time-frequency pixel matrix: ； in, This refers to multiple columns of data in the time-frequency pixel matrix. The time-frequency pixel matrix, To arrange the time-frequency pixel matrix by columns, The number of frequency sampling points is set to a value. The number of distance sampling points is set to a value. This represents the number of frequency sampling points. This represents the number of sampling points. The number of data in each column. This refers to the data in the first column of the time-frequency pixel matrix. The first pixel in the time-frequency pixel matrix Column data.

6. The method according to claim 4, characterized in that, The estimation parameters include the weights, mean, and variance of the estimated mixed components. The variance of the estimated mixed components is the square of the standard deviation of the estimated mixed components. Based on the estimation parameters, the false alarm rate, and the detection threshold, the interference signal is sequentially subjected to time-frequency domain localization processing and time-domain reconstruction to obtain a reconstructed interference signal, including: The detection threshold is determined based on the weights of the estimated mixed components, the mean of the estimated mixed components, the standard deviation of the estimated mixed components, and the false alarm rate. Based on the time-frequency pixel matrix and the detection threshold, the time-frequency complex matrix is ​​located in the time-frequency domain using a binary decision method to obtain an interference time-frequency complex matrix, which is a time-frequency complex matrix containing only interference signals. The time-domain reconstruction of the interference time-frequency complex matrix is ​​performed based on the inverse short-time Fourier transform to obtain the reconstructed interference signal.

7. The method according to claim 6, characterized in that, The step of determining the detection threshold based on the weights of the estimated mixture components, the mean of the estimated mixture components, the standard deviation of the estimated mixture components, and the false alarm rate includes: Based on the weights of the estimated mixture components, the mean of the estimated mixture components, the standard deviation of the estimated mixture components, and the observed data, the relationship between the detection threshold and the false alarm rate is determined using the following fourth formula: ； in, The false alarm rate is... The detection threshold is... The weights for the estimated mixed components, The mean of the estimated mixed components, The standard deviation of the estimated mixture components is given. The observation data; Based on the aforementioned relationship and the normal distribution of the false alarm rate, the detection threshold is determined using the following fifth formula: ； in, It is the inverse function of the complement of the error function. This is the complement function of the error function.

8. The method according to claim 6, characterized in that, The step of locating the time-frequency complex matrix in the time-frequency domain using a binary decision method based on the time-frequency pixel matrix and the detection threshold to obtain the interference time-frequency complex matrix includes: Based on the time-frequency pixel matrix and the detection threshold, using the binary decision method and the following sixth formula, interference localization processing is performed on the time-frequency complex matrix in the time-frequency domain to obtain the interference time-frequency complex matrix: ； in, The interference time-frequency complex matrix is... Let be the time-frequency complex matrix. The time-frequency pixel matrix, The detection threshold is... The number of frequency sampling points is set to a value. The value is set to the number of distance sampling points.

9. The method according to claim 1, characterized in that, The step of performing time-domain cancellation on the radar echo signal and the reconstructed interference signal to suppress interference in the radar echo signal includes: The radar echo signal is subtracted from the reconstructed interference signal in the time domain to obtain the radar echo signal after interference suppression, thereby achieving interference suppression of the radar echo signal.

10. A radar active interference time-frequency statistical detection and suppression device, characterized in that, The device includes: The determination module is used to sequentially perform short-time Fourier transform, modulus value taking, and logarithmic processing on the radar echo signal to determine the time-frequency pixel matrix, wherein the radar echo signal includes interference signals; A module is established to establish a Gaussian mixture model of the time-frequency pixel matrix. The Gaussian mixture model includes multiple parameters, including the weights of the mixture components, the mean of the mixture components, and the variance of the mixture components. The estimation module is used to estimate the parameters in the Gaussian mixture model according to the expectation-maximization algorithm, and obtain the corresponding estimated parameters. The positioning and reconstruction module is used to perform time-frequency domain positioning processing and time-domain reconstruction on the interference signal according to the estimated parameters, false alarm rate and detection threshold, so as to obtain the reconstructed interference signal. The time-domain cancellation module is used to perform time-domain cancellation on the radar echo signal and the reconstructed interference signal to achieve interference suppression of the radar echo signal.

Images