A method for identifying radar towing jamming

Abstract

The present invention discloses a method for identifying radar towing interference, which comprises the following steps: S1. Obtain a radar received signal, sample and accumulate the signals within the first pulse repetition interval in multiple coherent processing intervals of the received signal to form a one-dimensional discrete sequence; S2. Decompose the one-dimensional discrete sequence in step S1 by using the Mallat algorithm of discrete wavelet transform to obtain a sequence composed of high-frequency coefficients of each layer at the J-layer scale; S3. According to the characteristics of the one-dimensional discrete sequence in step S1, circularly shift the sequence composed of high-frequency coefficients of each layer in step S2 to obtain a shifted sequence; S4. Calculate the correlation coefficients between the sequence composed of high-frequency coefficients of each layer in step S3 and the shifted sequence, and construct a multi-dimensional wavelet feature vector r; S5. Use a classifier to classify and identify the feature vector r for range towing interference, velocity towing interference, and range-velocity combined towing interference. The present invention is applicable to the radar tracking mode and greatly improves the recognition accuracy rate.

Description

Technical Field

[0001] The invention belongs to the technical field of radar signal interference recognition, and in particular relates to a radar towing type interference recognition method. Background Art

[0002] Interference identification technology is the basis for efficient real-time situational awareness and interference effectiveness evaluation technology. At present, interference identification mainly relies on manual intervention, with low levels of autonomy and intelligence. With the widespread use of digital radio frequency memory (DRFM) technology, jammers can intercept, store, modulate and forward signals in a very short time, making the interference signal and echo highly similar.

[0003] When the radar is in tracking mode, the deception jamming technology based on DRFM mainly includes three types of pull-off jamming, namely range gate pull off (RGPO), velocity gate pull off (VGPO), and range-velocity gate pull off (RVGPO). At present, there are two main ideas for this type of jamming identification technology:

[0004] 1. Using hardware quantization modulation features to identify and distinguish, mainly including: using the power amplifier characteristics in the jammer to extract singular spectrum entropy to classify and identify deceptive interference; using the harmonic component parasitics generated by the stepping step of the DRFM numerically controlled phase shifter, and analyzing the difference between speed deceptive interference and echo through singular spectrum, etc. However, with the continuous development of hardware technology, the quantitative features of DRFM are not easy to extract, and the difficulty of interference identification increases.

[0005] Second, mining the characteristic differences of signals in different domains as feature vectors for classification and identification, mainly including: extracting the third-order Rayleigh entropy and separability from the SPWVD time-frequency analysis image as feature parameters for identification; extracting entropy features and box dimensions from the radar receiving signal in the dual spectrum analysis domain as feature vectors; based on digital image processing technology, using Zernike moment features and stacked sparse autoencoders in the time-frequency domain to extract the time-frequency image detail features to form feature vectors for classification and identification. Through the existing public literature research, it is known that the above three methods can achieve an accuracy of about 80% for the three types of dragging interference when the signal-to-noise ratio (SNR) is 0dB, but no simulation estimation is performed when SNR<0dB; filtering, maximum entropy segmentation, and edge optimization processing are performed in the time-frequency grayscale image, and then a 9-dimensional feature vector is obtained through singular value decomposition for identification, but there are limitations such as complex calculation and general real-time performance. Summary of the Invention

[0006] To solve the above problems, the object of the present invention is to provide a method for identifying radar towing interference, which is applicable to the identification of towing interference types in the radar tracking mode.

[0007] To achieve the above object of the invention, the present invention adopts the following technical solutions:

[0008] A method for identifying radar towing interference includes the following steps:

[0009] S1. Obtain the radar received signal, and perform signal sampling and accumulation within the first pulse repetition interval in multiple coherent processing intervals of the received signal to form a one-dimensional discrete sequence Z n ; The operation is as follows:

[0010] One coherent processing interval of the radar processes M echo pulses. First, discrete signals are sampled. The echo signal within the first pulse repetition interval in the k-th coherent processing interval is denoted as x r_cpi_k_pri , and N consecutive ones are accumulated to form a one-dimensional discrete sequence Z n , which is expressed as:

[0011] Z n =(x r_cpi_1_pri , x r_cpi_2_pri ,..., x r_cpi_N_pri ) (1)

[0012] S2. For the one-dimensional discrete sequence Z in step S1 n , use the Mallat algorithm of discrete wavelet transform to decompose it to obtain a sequence d composed of high-frequency coefficients at each layer at the J-layer scale j (j = 1, 2,..., J); The operation is as follows:

[0013] Each mother wavelet function corresponds to a filter bank, that is, a high-pass filter h(k) and a low-pass filter l(k). The one-dimensional discrete sequence Z n is respectively filtered by the high-pass filter h(k) and the low-pass filter l(k) and then downsampled by two to obtain the first-layer high-frequency coefficient d1 and the low-frequency coefficient c1. The low-frequency coefficient c1 is filtered by the high-pass filter h(k) and the low-pass filter l(k) and downsampled by two to obtain the second-layer high-frequency coefficient d2 and the low-frequency coefficient c2, and so on until the J-th decomposition to obtain the high-frequency coefficient d J ;

[0014] S3. According to the sequence characteristics of the one-dimensional discrete sequence Z in step S1 n , perform circular shifting on the sequence d composed of high-frequency coefficients at each layer in step S2 j to obtain the shifted sequence d j ′, and the shifting amount is k j , dj ′, k j are respectively expressed as

[0015] d j ′ = circshift(d j , k j ) (2)

[0016]

[0017] where circshift(x, y) represents circularly shifting the sequence x by y points, and length(x) represents calculating the length of the sequence x;

[0018] S4. Calculate the correlation coefficient r between the sequence d composed of the high-frequency coefficients of each layer in step S3 and the shifted sequence d j ′, and construct a multi-dimensional wavelet feature vector r. The correlation coefficient r j ′ and the feature vector r are respectively expressed as: j j J r = [r1, r2, r3,..., r

[0019]

[0020] J J (5)

[0021] where cov(x, y) represents the covariance of the sequences x and y, and var(x) represents the variance of x;

[0022] S5. Use a classifier to perform classification and recognition of range gate pull-off interference, velocity pull-off interference, and range-velocity combined pull-off interference on the feature vector r.

[0023] Furthermore, in the above step S5, the classifier includes but is not limited to classification algorithms such as decision trees, support vector machines, and neural networks.

[0024] Due to the above-mentioned technical solution, the present invention has the following advantages:

[0025] This radar pull-off interference recognition method samples and accumulates the signals within the first PRI of multiple CPIs of the target echo to form a one-dimensional discrete vector, uses wavelet transform to extract multi-scale wavelet coefficients, calculates the correlation coefficient at each scale to form a feature vector for interference recognition; greatly improves the recognition accuracy rate, especially when the signal-to-noise ratio (SNR < 0 dB), and at the same time, range false targets and pull-off interference can be effectively recognized together, increasing the accuracy and rigor of recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a schematic diagram of multi-CPI pulse signal accumulation;

[0027] Figure 2 It is a schematic diagram of the correspondence between wavelet coefficients and frequency ranges;

[0028] Figure 3 It is a flow chart of the method for identifying radar towing jamming of the present invention;

[0029] Figure 4 It is a simulation result diagram of the correct recognition rate of jamming under different signal patterns realized by using the method for identifying radar towing jamming of the present invention. Specific embodiments

[0030] The technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0031] It is known from the interference mathematical model that range gate pull-off (RGPO) and velocity gate pull-off (VGPO) respectively perform time delay and Doppler frequency modulation on the echo, and range-velocity combined gate pull-off (RVGPO) combines the two. Assume that v0 represents the towing speed. The towing distance between adjacent pulses or within a coherent processing interval (CPI) is relatively small, only PRI×v0 or CPI×v0. When the sampling rate is not high, after the radar receives the echo pulse with superimposed interference and samples it into the digital domain, the echo pulses within the same coherent processing interval (CPI) are highly similar. Therefore, it is easy to be recognized as range false target interference within a short time. Similarly, range-velocity combined gate pull-off (RVGPO) is easy to be recognized as velocity gate pull-off (VGPO). Therefore, it is difficult to accurately identify different towing interferences within a single coherent processing interval (CPI) or within adjacent multiple pulse repetition intervals (PRI). Therefore, it is necessary to accumulate the pulses within a certain time. As Figure 1 shown, consider accumulating N coherent processing intervals (CPI), and each coherent processing interval (CPI) processes M pulses; CPI_N represents the signal to be processed within the Nth coherent processing interval (CPI), and x r_cpi_N_pri is the signal within the first pulse repetition interval (PRI) in the Nth coherent processing interval (CPI). Take the signal within the first pulse repetition interval (PRI) in each coherent processing interval (CPI) to construct a one-dimensional discrete sequence Z n , which is expressed as Equation (1); the one-dimensional discrete sequence Z n Since the pulse repetition interval (PRI) time of M*N times is accumulated, it is ensured that the time delay modulation information generated by the interference signal can be reflected in a certain sampling rate.

[0032] As Figure 2 shown, the entire frequency band ranges from 0 to f s, f s is the sampling frequency. It can be seen that as the decomposition level increases, the time-domain resolution becomes worse, but as the frequency band is subdivided, the frequency resolution becomes better. The high-frequency coefficient d after the m-th layer of decomposition m The corresponding frequency band is denoted as d m : [f s / 2 m f s / 2 m-1 ; Therefore, the wavelet coefficients at different scales can reflect the subtle differences in the time-domain and frequency-domain resolutions of the signal, and have high practical value in signal recognition.

[0033] As Figure 3 shown, the radar towing jamming recognition method of the present invention includes the following specific steps:

[0034] S1. Obtain the radar received signal y(t), expressed as

[0035] y(t) = x r (t) + j(t) + n(t) (6)

[0036] where x r (t) is the echo signal, j(t) is the jamming signal, and n(t) is the noise;

[0037] The received signal y(t) is the signal received by the radar when sequentially transmitting multiple carrier signals to the target object within a preset N coherent processing intervals (CPI);

[0038] The above received signal y(t) is used as the input for signal sampling and accumulation within the first pulse repetition interval (PRI) of multiple coherent processing intervals (CPI), and outputs a one-dimensional discrete sequence Z n ; The specific operation is:

[0039] One coherent processing interval (CPI) of the radar processes M echo pulses. First, a discrete signal is sampled. The echo signal within the first pulse repetition interval (PRI) of the k-th coherent processing interval (CPI) is denoted as x r_cpi_k_pri , and N consecutive ones are accumulated to form a one-dimensional discrete sequence Z n , expressed as:

[0040] Z n = (x r_cpi_1_pri , x r_cpi_2_pri ,..., x r_cpi_N_pri ) (1)

[0041] S2. Use the Mallat algorithm of the multi-scale discrete wavelet transform to decompose the one-dimensional discrete sequence Z in step S1 n to obtain a sequence d composed of high-frequency coefficients at each layer at the J-layer scalej (j = 1, 2, ..., J), that is, d1, d2, ..., d J ; The specific operation is as follows:

[0042] Each mother wavelet function corresponds to a filter bank, namely a high-pass filter h(k) and a low-pass filter l(k). The one-dimensional discrete sequence Z n After being filtered by the high-pass filter h(k) and the low-pass filter l(k) respectively and then downsampled by a factor of 2, the first-layer high-frequency coefficients d1 and low-frequency coefficients c1 are obtained. The low-frequency coefficients c1 are filtered by the high-pass filter h(k) and the low-pass filter l(k) and downsampled to obtain the second-layer high-frequency coefficients d2 and low-frequency coefficients c2, and so on until the Jth decomposition to obtain the high-frequency coefficients d J ;

[0043] S3. According to the one-dimensional discrete sequence Z in step S1 n Sequence characteristics, for the sequence d composed of the high-frequency coefficients of each layer in step S2 j Perform circular shift to obtain the shifted sequence d j ′, and the shift amount is k j ; The specific operation is as follows:

[0044] Since there are subtle feature differences in the in-pulse and inter-pulse of the one-dimensional discrete sequence Z corresponding to different interferences, using the regular feature that the one-dimensional discrete sequence Z n Accumulates N times at intervals of the pulse repetition interval (PRI) in time series, select the shift amount k n Is 1 / N of the length of the sequence d composed of high-frequency coefficients, so d j ′, k j Are respectively expressed as j ′, k j Are respectively expressed as

[0045] d j ′ = circshift(d j , k j ) (2)

[0046]

[0047] Among them, circshift(x, y) represents circular shift of sequence x by y points, and length(x) represents calculating the length of sequence x;

[0048] S4. Calculate the correlation coefficient r j Between the sequence d composed of the high-frequency coefficients of each layer in step S3 j And the shifted sequence d j ′, construct a multi-dimensional wavelet feature vector r, and the correlation coefficient r j And the feature vector r are respectively expressed as:

[0049]

[0050] r = [r1, r2, r3, ..., r J (5)

[0051] where cov(x, y) represents the covariance of sequences x and y, and var(x) represents the variance of x;

[0052] S5. Use a classifier to classify and identify the three types of towing interferences of the feature vector r, namely range towing interference (RGPO), velocity towing interference (VGPO), and range-velocity combined towing interference (RVGPO); the classifier includes, but is not limited to, classification algorithms of decision tree (DT), support vector machine (SVM), and neural network.

[0053] Construct a simulation environment where the radar emits LFM pulse signals to verify the generality of the radar towing interference recognition method of the present invention. Construct 3 types of signal parameters for comparative simulation. The signal bandwidth B, PRI, pulse width T, and towing velocity v0 parameters are shown in Table 1. The sampling frequency f s = 20 MHz, and the jamming-to-signal ratio (JSR) JSR = 3 dB; use range false target interference (denoted by RANGE) as a comparison for recognition, and select M = 16, N = 20; the distance between the false target and the true target R = 2 km, and the position of the false target is the starting position of range towing interference (RGPO) and range-velocity combined towing interference (RVGPO). Select the Daubechies series wavelet "DB2" as the mother wavelet function and select J = 4 levels of wavelet decomposition. Conduct 200 Monte Carlo tests under each signal-to-noise ratio (SNR) condition. The data of the first 100 times are used as the learning samples of the classifier, and the data of the last 100 times are used as test cases. Select the decision tree as the classifier, and the recognition accuracy rate is as Figure 4 shown.

[0054] Table 1 Simulation signal parameter settings

[0055]

[0056] From Figure 4It can be seen that the correct recognition rate of the interference of signal S3 pattern is the lowest. However, generally, the simulation results of the three signal patterns still have a 90% correct rate at SNR = 0dB, which is about 10% higher than the relevant literature of the prior art. Moreover, the correct rates of signals S1 and S2 are still higher than 90% at -5dB < SNR < 0dB, and signal S3 also has a correct rate of more than 80%. In the relevant literature of the prior art, the simulation of SNR < 0dB was not carried out. Therefore, it can be proved that the radar towing interference recognition method of the present invention has a significant improvement in the correct recognition rate; at the same time, it can effectively identify the interference (RANGE) and range gate pull-off (RGPO) interference.

[0057] The radar towing interference recognition method of the present invention can further optimize the interference recognition effect by adjusting the dimension of the feature vector and selecting a more suitable mother wavelet function.

[0058] The above are only the preferred embodiments of the present invention, and not the limitations of the present invention. Without departing from the spirit and scope of the present invention, all equal changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope of the patent protection of the present invention.

Claims

1. A method for identifying radar towing jamming, characterized in that: It includes the following steps: S1. Obtain the radar received signal, perform signal sampling and accumulation within the first pulse repetition interval among multiple coherent processing intervals on the received signal, and form a one-dimensional discrete sequence Z n ; The operation is as follows: One coherent processing interval of the radar processes M echo pulses. First, discrete signals are sampled. The echo signal within the first pulse repetition interval in the k-th coherent processing interval is denoted as x r_cpi_k_pri , and N consecutive ones are cumulatively formed into a one-dimensional discrete sequence Z n , which is expressed as: Z n = (x r_cpi_1_pri , x r_cpi_2_pri ,..., x r_cpi_N_pri )(1) S2. For the one-dimensional discrete sequence Z in step S1 n Use the Mallat algorithm of discrete wavelet transform to decompose it into a sequence d composed of high-frequency coefficients at each layer at the J-layer scale j , where j = 1, 2,..., J; The operation is as follows: Each mother wavelet function corresponds to a filter bank, namely a high-pass filter h(k) and a low-pass filter l(k), and a one-dimensional discrete sequence Z n After being filtered by the high-pass filter h(k) and the low-pass filter l(k) respectively and then downsampled by a factor of 2, the first-level high-frequency coefficients d1 and low-frequency coefficients c1 are obtained. The low-frequency coefficients c1 are filtered by the high-pass filter h(k) and the low-pass filter l(k) and then downsampled to obtain the second-level high-frequency coefficients d2 and low-frequency coefficients c2, and so on until the Jth decomposition to obtain the high-frequency coefficients d J ; S3. According to the characteristics of the one-dimensional discrete sequence Z in step S1 n sequence, for the sequence d composed of the high-frequency coefficients of each layer in step S2 j perform circular shifting to obtain the shifted sequence d j ′, where the shift amount is k j , d j ′, k j are respectively represented as d j ′ = circshift(d j , k j ) (2) where circshift(x,y) represents circularly shifting sequence x by y points, length(x) represents calculating the length of sequence x; N represents the length of the one-dimensional discrete sequence Z n ; S4. Calculate the correlation coefficient r of the sequence d composed of each layer of high-frequency coefficients in step S3 j and the sequence d j ' after displacement j , and construct a multi-dimensional wavelet feature vector r. The correlation coefficient r j , and the feature vector r are respectively expressed as: r＝[r1,r2,r3,...,r J ] (5) Among them, cov(x,y) represents the covariance of sequences x and y, and var(x) represents the variance of x; S5. Use a classifier to classify and identify the feature vector r for distance towing interference, speed towing interference, and distance-speed combined towing interference.

2. The radar towing interference recognition method according to claim 1, characterized in that: In step S5 thereof, the classifier includes, but is not limited to, classification algorithms such as decision trees, support vector machines, and neural networks.

Images