Variable-hopping-speed frequency hopping signal sorting method based on sparse reconstruction

Abstract

The invention discloses a variable-hopping-speed frequency hopping signal sorting method based on sparse reconstruction. The method comprises the following steps: estimating frequency hopping signal frequency points by using a frequency measurement function of a channelized receiver; the incoming wave direction of the frequency hopping signal has sparsity in the airspace, and the sparse representation model of the frequency hopping signal can be established in the airspace by using an array antenna; solving the sparse representation model; estimating the incoming wave direction of the frequency hopping signals, and separating the frequency hopping signals; and performing time-frequency analysis on the separated frequency hopping signal, and estimating to obtain a frequency hopping pattern. The method is simultaneously applicable to sorting of constant-hopping-speed and variable-hopping-speed frequency hopping signals, overcomes the limitation that only the constant-hopping-speed frequency hopping signals can be sorted by the existing method, is wider in application scene, can obtain the time domain waveform, the airspace incoming wave direction and the frequency domain frequency hopping frequency set of each path of variable-hopping-speed frequency hopping signals, and enriches communication confrontation reconnaissance information.

Description

Technical Field

[0001] The invention belongs to the field of signal sorting in communication countermeasure reconnaissance, and in particular relates to a variable hopping rate frequency hopping signal sorting method based on sparse reconstruction. Background Art

[0002] Regarding the sorting of frequency hopping signals, some researchers have proposed a sorting method based on machine learning algorithms. This type of method first estimates the parameters of the frequency hopping signal in the time domain and frequency domain, such as the frequency hopping frequency, the frequency hopping dwell time, etc. These parameters are regarded as the characteristics of the frequency hopping signal, and then uses algorithms in machine learning, such as neural network algorithms and clustering algorithms, to sort multiple frequency hopping signals based on the characteristics. This type of method is only applicable to the sorting of constant hopping rate frequency hopping signals. The sorting accuracy depends on the accuracy of parameter estimation, and it can only classify the frequency hopping signals, but cannot obtain the waveforms of each frequency hopping signal.

[0003] In order to separate the waveforms of each frequency hopping signal from the aliased frequency hopping signal, Yang Yuncheng, Sun Xueli, Zhong Zhaogen and others proposed an improved independent component analysis frequency hopping network station separation method, and used independent component analysis to realize blind separation of aliased frequency hopping signals. However, the aliasing matrix in the signal model is a random real number matrix, and the actual aliasing matrix is ​​a complex number matrix containing the phase differences of each receiving channel. Therefore, it is inconsistent with the actual aliasing matrix.

[0004] For the sorting of frequency-hopping signals with variable hopping rates, some researchers have combined time-frequency single source point detection, clustering, and sparse reconstruction methods to achieve frequency-hopping station classification and source signal recovery. However, the aliasing matrix in the signal model applicable to this method does not introduce the frequency of the frequency-hopping signal, and the established signal model is inaccurate.

[0005] Wang Miao, Cai Xiaoxia, and Lei Yingke proposed a variable-speed frequency-hopping signal sorting algorithm based on frame overlap. They use the frame overlap of the receiver sampling data frames to sort the frequency-hopping signals. This method requires that the frequency points where the frequencies of the various frequency-hopping signals are concentrated cannot overlap, so the application scenarios are limited.

[0006] In summary, the existing method is only applicable to the sorting of constant jump rate network stations, and cannot sort variable jump rate network stations. Variable jump rate network stations have also been applied in practice. The existing method is not suitable for sorting variable jump rate network station signals in actual scenarios. If the signal sorting cannot be completed, the next step of network station identification cannot be carried out, and the ability to obtain communication countermeasure reconnaissance information is limited. Summary of the invention

[0007] In order to solve the problem that the existing methods cannot sort variable hopping rate frequency hopping signals, the present invention provides a variable hopping rate frequency hopping signal sorting method based on sparse reconstruction. The method is suitable for the sorting of constant hopping rate and variable hopping rate frequency hopping signals, overcoming the limitation that the existing methods can only sort constant hopping rate frequency hopping signals, and has a wider application scenario. It can obtain the time domain waveform, spatial domain incoming wave direction, and frequency domain frequency hopping frequency set of each variable hopping rate frequency hopping signal, thereby enriching the communication countermeasure reconnaissance information.

[0008] The objective of the present invention is achieved through the following technical solutions:

[0009] A variable-speed frequency-hopping signal sorting method based on sparse reconstruction comprises the following steps:

[0010] S1. Estimating the frequency of the frequency hopping signal using the frequency measurement function of the channelized receiver;

[0011] S2. Construct a sparse representation model of frequency hopping signals: The incoming wave direction of the frequency hopping signal is sparse in the spatial domain. The sparse representation model of the frequency hopping signal can be established in the spatial domain by using an array antenna;

[0012] S3. Solve the sparse representation model;

[0013] S4 estimates the frequency hopping signal wave direction, and separates each frequency hopping signal;

[0014] S5. Perform time-frequency analysis on the separated frequency hopping signal to estimate the frequency hopping pattern.

[0015] Furthermore, the step S1 comprises:

[0016] The channelized receiver divides the receiving channel bandwidth into three levels: sub-band, sub-band and sub-channel;

[0017] The instantaneous frequency coverage range f A ~f B Divided into M parts, the bandwidth of each sub-band is B M =(f B -f A ) / M;

[0018] Sub-bands further subdivide sub-bands. If each sub-band is divided into N sub-bands, the bandwidth of each sub-band is B. N =(f B -f A ) / MN;

[0019] Sub-channels further divide the sub-bands into K sub-channels, and the bandwidth of each sub-channel is B. K =(f B -f A ) / MNK;

[0020] After the received source signal enters the channelized receiver, if there is a signal output in the mth sub-band, the nth sub-band, and the kth sub-channel, the center frequency f of the channel is used. s As an estimate of the output signal carrier frequency:

[0021]

[0022] Since the carrier frequency of the frequency hopping signal is constantly changing, the received frequency hopping communication signal is output on multiple sub-channels, so the estimated value of each hopping carrier frequency is obtained from the center frequency of each sub-channel.

[0023] Furthermore, the step S2 comprises:

[0024] The array antenna is a uniform linear array, and the angle coverage interval of the entire airspace is [-π / 2, π / 2]. The airspace is divided into M equal parts, and the interval between adjacent angles is Δθ=π / M, that is, a dictionary matrix A(θ) is established in the airspace:

[0025] A(θ)=[a(θ 1 ),a(θ 2 ),…,a(θ M )]

[0026] in, f k Indicates a frequency point of the frequency hopping signal;

[0027] If the receiver receives an aliased frequency-hopping signal, then the signal received within a period of time contains multiple frequency-hopping frequencies, and the maximum frequency value f max Corresponding to the smallest wavelength λ min ,Right now c represents the speed of light, d is the array element spacing, and

[0028] Assume that these frequency hopping frequencies are [f 1 ,f 2 ,…,f K ], each frequency point corresponds to a spatial domain dictionary matrix, thereby constructing the dictionary matrix D(θ) in the complete sparse representation model:

[0029] D(θ)=[A 1 (θ),A 2 (θ),…,A K (θ)]

[0030] Since the frequency hopping signal only exists at a certain frequency point and direction at a certain moment, the dictionary matrix D(θ) is used to construct a sparse representation model of the frequency hopping signal:

[0031] Y(t)=D(θ)S(t)

[0032] Where Y(t) is the array antenna receiving signal matrix, S(t) = [s 1 (t),s 2 (t),…,s n (t)] T is the frequency hopping source signal matrix.

[0033] Furthermore, the step S3 comprises:

[0034] The sparse representation model is solved using a convex optimization algorithm, which transforms the solution of the sparse representation model into a convex optimization problem:

[0035]

[0036] Among them, ||Y(t)-D(θ)S(t)|| F To reconstruct the residual, It means that after taking the 2-norm of each row of the source signal matrix, the resulting vector is taken to the 1-norm. This term is used to measure the sparsity of the solution vector. λ is the regularization factor that balances the reconstruction residual and sparsity.

[0037] The optimization problem is solved using the interior point method in convex optimization.

[0038] Furthermore, the step S4 comprises:

[0039] The corresponding atomic vector basis is found from the location of the larger value of the sparse solution vector. The direction of the frequency hopping signal is estimated according to the direction information contained in the atomic vector basis. The steering matrix B(θ) can be constructed from the direction of the wave. The pseudo-inverse of the steering matrix B(θ) is calculated to restore the source signal:

[0040]

[0041] Furthermore, the step S5 comprises:

[0042] Performing time-frequency analysis on each frequency hopping signal separated in step S4 by short-time Fourier transform;

[0043] The time-frequency diagram obtained by short-time Fourier transform shows the relationship between frequency jump and time, and the jump speed can be estimated from the duration of each frequency point;

[0044] Further statistical analysis is performed on all frequency points of the time-frequency diagram within a period of time to estimate the frequency hopping pattern of the frequency hopping signal.

[0045] In order to reduce the size of the sparse representation model and reduce the computational complexity of the model solution, preferably, the received source signal is first processed in segments; for each segment of the signal, a sparse representation model is constructed through step S2, and the model is solved through step S3, and then the incoming direction of each frequency hopping signal is estimated through step S4, and each frequency hopping signal is separated according to the direction information; the source signal data blocks with the same incoming direction separated from all the segmented signals are spliced ​​together, that is, the complete source signal is restored; finally, the separated frequency hopping signal is subjected to time-frequency analysis through step S5 to estimate the frequency hopping pattern.

[0046] The present invention has the following advantages:

[0047] The present invention provides a variable hopping rate frequency hopping signal sorting method based on sparse reconstruction. Firstly, a sparse representation model of the variable hopping rate frequency hopping signal is constructed by utilizing the sparse characteristics of the incoming wave direction of the frequency hopping signal in the spatial domain. The model takes into account the influence of frequency domain and spatial domain parameters on the model. According to the characteristics of the frequency hopping change of the frequency hopping signal, a spatial domain atomic basis is constructed at each frequency point within the duration of the signal. In order to reduce the size of the sparse representation model and reduce the computational complexity of the model solution, the received frequency hopping signal is processed in segments. After the sparse representation model is established, the model is solved based on a sparse reconstruction algorithm, and the incoming wave direction of each frequency hopping signal is estimated based on the solution result, and the frequency hopping signals of each channel are separated according to the direction information. The sampling data of the frequency hopping signals in the same direction are spliced ​​to obtain each complete frequency hopping signal, and the frequency hopping frequency set and the hopping speed of each frequency hopping signal are estimated by using the time-frequency analysis technology to obtain the frequency hopping pattern, and the frequency hopping signal sorting is completed. The present invention is suitable for sorting constant-hopping-rate and variable-hopping-rate frequency-hopping signals, overcoming the limitation that the existing methods can only sort constant-hopping-rate frequency-hopping signals. The present invention has a wider range of application scenarios and can obtain the time domain waveform, spatial domain incoming wave direction, and frequency domain frequency-hopping frequency set of each variable-hopping-rate frequency-hopping signal, thus enriching the communication countermeasure reconnaissance information. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of a variable-rate frequency-hopping signal sorting method based on sparse reconstruction described in Example 1 of the present invention.

[0049] Figure 2 This is a flow chart of a variable-rate frequency-hopping signal sorting method based on sparse reconstruction described in Example 2 of the present invention. DETAILED DESCRIPTION

[0050] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention.

[0051] Example 1

[0052] like Figure 1As shown, this embodiment is a variable hopping rate frequency hopping signal sorting method based on sparse reconstruction, comprising the following steps:

[0053] S1. Use the frequency measurement function of the channelized receiver to estimate the frequency of the frequency hopping signal:

[0054] The channelized receiver divides the receiving channel bandwidth into three levels: sub-band, sub-band and sub-channel;

[0055] The instantaneous frequency coverage range f A ~f B Divided into M parts, the bandwidth of each sub-band is B M =(f B -f A ) / M;

[0056] Sub-bands further subdivide sub-bands. If each sub-band is divided into N sub-bands, the bandwidth of each sub-band is B. N =(f B -f A ) / MN;

[0057] Sub-channels further divide the sub-bands into K sub-channels, and the bandwidth of each sub-channel is B. K =(f B -f A ) / MNK;

[0058] After the intercepted signal enters the channelized receiver, if there is a signal output in the mth sub-band, the nth sub-band, and the kth sub-channel, the center frequency f of the channel is used. s As an estimate of the output signal carrier frequency:

[0059]

[0060] Since the carrier frequency of the frequency hopping signal is constantly changing, the intercepted frequency hopping communication signal is output in multiple sub-channels. Therefore, the estimated value of each hopping carrier frequency can be obtained from the center frequency of each sub-channel.

[0061] S2. Construct a sparse representation model for frequency hopping signals:

[0062] The arrival direction of the frequency hopping signal is sparse in the spatial domain, so an array antenna can be used to establish a sparse representation model of the frequency hopping signal in the spatial domain.

[0063] The array antenna used in this embodiment is a uniform linear array. The angle coverage interval of the entire airspace is [-π / 2, π / 2]. The airspace is divided into M equal parts, and the interval between adjacent angles is Δθ=π / M. That is, a dictionary matrix A(θ) is established in the airspace:

[0064] A(θ)=[a(θ1 ),a(θ 2 ),…,a(θ M )] (2)

[0065] in, f k Indicates a frequency point of a frequency hopping signal. If the receiver receives an aliased frequency hopping signal, then the signal received within a period of time contains multiple frequency hopping frequencies, and the maximum frequency value f max Corresponding to the smallest wavelength λ min ,Right now c = 3 × 10 8 m / s, represents the speed of light, d is the array element spacing, and These frequency hopping frequencies can be measured by the frequency measurement function of the channelized receiver. Suppose these frequency hopping frequencies are [f 1 ,f 2 ,…,f K ], each frequency point can correspond to a spatial domain dictionary matrix, thereby constructing the dictionary matrix D(θ) in the complete sparse representation model:

[0066] D(θ)=[A 1 (θ),A 2 (θ),…,A K (θ)] (3)

[0067] Since the frequency hopping signal only exists at a certain frequency point and direction at a certain moment, the dictionary matrix D(θ) can be used to construct a sparse representation model of the frequency hopping signal, as shown in formula (4):

[0068] Y(t)=D(θ)S(t) (4)

[0069] Where Y(t) is the array antenna receiving signal matrix, S(t) = [s 1 (t),s 2 (t),…,s n (t)] T is the frequency hopping source signal matrix.

[0070] S3. Solve the sparse representation model:

[0071] The sparse representation model can be solved using a convex optimization algorithm, which transforms the solution of the sparse representation model into a convex optimization problem as shown in formula (5):

[0072]

[0073] Among them, ||Y(t)-D(θ)S(t)|| F To reconstruct the residual, It means that after taking the 2-norm of each row of the source signal matrix, the resulting vector is taken to the 1-norm. This term is used to measure the sparsity of the solution vector. λ is the regularization factor that balances the reconstruction residual and sparsity. The optimization problem shown in formula (5) can be solved by the interior point method in convex optimization.

[0074] S4. Estimate the direction of the frequency hopping signal and separate the frequency hopping signals:

[0075] The corresponding atomic vector basis is found from the location of the larger value of the sparse solution vector. The direction of the frequency hopping signal is estimated according to the direction information contained in the atomic vector basis. The steering matrix B(θ) can be constructed from the direction of the wave. The source signal can be restored by taking the pseudo-inverse of the steering matrix B(θ), that is:

[0076]

[0077] S5. Perform time-frequency analysis on the separated frequency hopping signal to estimate the frequency hopping pattern:

[0078] By performing time-frequency analysis on the separated frequency-hopping signals, the law of signal frequency variation over time can be observed. The time-frequency analysis technology to be adopted in this embodiment is short-time Fourier transform. Short-time Fourier transform is to add a time window on the basis of conventional Fourier transform and perform Fourier transform on a section of data. The mathematical expression of discrete short-time Fourier transform is shown in formula (7):

[0079]

[0080] In formula (7), g(kT-mT) is the window function.

[0081] The time-frequency diagram obtained by short-time Fourier transform shows the relationship between frequency hopping and time, and the hopping speed can be estimated from the duration of each frequency point.

[0082] By further statistically analyzing all frequency points of the time-frequency diagram within a period of time, the frequency hopping pattern of the frequency hopping signal can be estimated.

[0083] Example 2

[0084] The present embodiment is a method for sorting frequency hopping signals with variable hopping rates based on sparse reconstruction. In order to reduce the size of the sparse representation model and reduce the computational complexity of the model solution, the present embodiment processes the received source signal in segments. For each segment of the signal, a sparse representation model is constructed through step S2, and the model is solved through step S3. Then, the incoming directions of each frequency hopping signal are estimated through step S4, and each frequency hopping signal is separated according to the direction information. Then, the signal data blocks with the same incoming directions in all the segmented signals are spliced ​​together to restore the complete source signal. Finally, time-frequency analysis is performed on the separated frequency hopping signals through step S5 to estimate the frequency hopping pattern.

[0085] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A variable-speed frequency-hopping signal sorting method based on sparse reconstruction, characterized in that: The following steps are involved: S1. Estimating the frequency of the frequency hopping signal using the frequency measurement function of the channelized receiver; S2. Construct a sparse representation model of frequency hopping signals: The incoming wave direction of the frequency hopping signal is sparse in the spatial domain. The sparse representation model of the frequency hopping signal can be established in the spatial domain by using an array antenna; S3. Solve the sparse representation model; S4 estimates the frequency hopping signal wave direction, and separates each frequency hopping signal; S5. Perform time-frequency analysis on the separated frequency hopping signal to estimate the frequency hopping pattern.

2. The variable-speed frequency hopping signal sorting method based on sparse reconstruction as claimed in claim 1, characterized in that: The step S1 comprises: The channelized receiver divides the receiving channel bandwidth into three levels: sub-band, sub-band and sub-channel; The instantaneous frequency coverage range f A ~f B Divided into M parts, the bandwidth of each sub-band is B M =(f B -f A ) / M; Sub-bands further subdivide sub-bands. If each sub-band is divided into N sub-bands, the bandwidth of each sub-band is B. N =(f B -f A ) / MN; Sub-channels further divide the sub-bands into K sub-channels, and the bandwidth of each sub-channel is B. K =(f B -f A ) / MNK; After the received source signal enters the channelized receiver, if there is a signal output in the mth sub-band, the nth sub-band, and the kth sub-channel, the center frequency f of the channel is used. s As an estimate of the output signal carrier frequency: Since the carrier frequency of the frequency hopping signal is constantly changing, the received frequency hopping communication signal is output on multiple sub-channels, so the estimated value of each hopping carrier frequency is obtained from the center frequency of each sub-channel.

3. The variable-speed frequency hopping signal sorting method based on sparse reconstruction as claimed in claim 1, characterized in that: The step S2 comprises: The array antenna is a uniform linear array, and the angle coverage interval of the entire airspace is [-π / 2, π / 2]. The airspace is divided into M equal parts, and the interval between adjacent angles is Δθ=π / M, that is, a dictionary matrix A(θ) is established in the airspace: A(θ)=[a(θ1),a(θ2),…,a(θ M )] in, f k Indicates a frequency point of the frequency hopping signal; If the receiver receives an aliased frequency-hopping signal, then the signal received within a period of time contains multiple frequency-hopping frequencies, and the maximum frequency value f max Corresponding to the smallest wavelength λ min ,Right now c represents the speed of light, d is the array element spacing, and Assume that these frequency hopping frequencies are [f1,f2,…,f K ], each frequency point corresponds to a spatial domain dictionary matrix, thereby constructing the dictionary matrix D(θ) in the complete sparse representation model: D(θ)=[A1(θ),A2(θ),…,A K (i)] Since the frequency hopping signal only exists at a certain frequency point and direction at a certain moment, the dictionary matrix D(θ) is used to construct a sparse representation model of the frequency hopping signal: Y(t)=D(θ)S(t) Where Y(t) is the array antenna receiving signal matrix, S(t) = [s1(t), s2(t), …, s n (t)] T is the frequency hopping source signal matrix.

4. The variable-speed frequency hopping signal sorting method based on sparse reconstruction as claimed in claim 1, characterized in that: The step S3 comprises: The sparse representation model is solved using a convex optimization algorithm, which transforms the solution of the sparse representation model into a convex optimization problem: Among them, ||Y(t)-D(θ)S(t)|| F To reconstruct the residual, It means that after taking the 2-norm of each row of the source signal matrix, the resulting vector is taken to the 1-norm. This term is used to measure the sparsity of the solution vector. λ is the regularization factor that balances the reconstruction residual and sparsity. The optimization problem is solved using the interior point method in convex optimization.

5. The variable-speed frequency hopping signal sorting method based on sparse reconstruction as claimed in claim 1, characterized in that: The step S4 comprises: The corresponding atomic vector basis is found from the location of the larger value of the sparse solution vector. The direction of the frequency hopping signal is estimated according to the direction information contained in the atomic vector basis. The steering matrix B(θ) can be constructed from the direction of the wave. The pseudo-inverse of the steering matrix B(θ) is calculated to restore the source signal:

6. The variable-speed frequency hopping signal sorting method based on sparse reconstruction as claimed in claim 1, characterized in that: The step S5 comprises: Performing time-frequency analysis on each frequency hopping signal separated in step S4 by short-time Fourier transform; The time-frequency diagram obtained by short-time Fourier transform shows the relationship between frequency jump and time, and the jump speed can be estimated from the duration of each frequency point; Further statistical analysis is performed on all frequency points of the time-frequency diagram within a period of time to estimate the frequency hopping pattern of the frequency hopping signal.

7. A method for sorting variable-speed frequency-hopping signals based on sparse reconstruction as claimed in any one of claims 1 to 6, characterized in that: First, the received source signal is processed in segments; for each segment of the signal, a sparse representation model is constructed through step S2, and the model is solved through step S3, and then the incoming direction of each frequency hopping signal is estimated through step S4, and each frequency hopping signal is separated according to the direction information; the source signal data blocks with the same incoming direction separated from all the segmented signals are spliced ​​together, that is, the complete source signal is restored; finally, the separated frequency hopping signal is subjected to time-frequency analysis through step S5, and the frequency hopping pattern is estimated.

Images