A method for blind estimation of frequency hopping signal parameters based on atomic features

Abstract

This invention discloses a blind estimation method for frequency-hopping signal parameters based on atomic features. The method involves: first, windowing, compressed sampling, and atomic feature extraction of the frequency-hopping signal; then, roughly estimating the hopping time based on the distribution of atomic features within different windows; next, precisely estimating the hopping time, hopping period, and hopping speed using a block-diagonalized Fourier orthogonal matrix; and finally, estimating the frequency of each hop signal using an orthogonal matching pursuit algorithm, and obtaining the frequency-hopping pattern of the entire sample by concatenating each hop signal. This invention can adjust the search step size and compression ratio to change the computational complexity and estimation accuracy, adapting to different signal-to-noise ratio (SNR) environments. In high SNR environments, increasing the search step size and compression ratio effectively reduces computational complexity, while in low SNR environments, decreasing the search step size and compression ratio effectively improves estimation accuracy.

Description

Technical Field

[0001] This invention belongs to the field of communication countermeasures technology, and in particular, it is a blind estimation method for frequency hopping signal parameters based on atomic features. Background Technology

[0002] Frequency-hopping communication (FHH) has been widely used in the military field due to its excellent anti-interception and anti-jamming capabilities. With the continuous upgrading of communication countermeasures, reconnaissance of non-friendly FHH communication has become an important aspect of the informationized battlefield. Blind parameter estimation, as a key link in the entire reconnaissance process, plays a crucial role, and the estimated parameters can be used for subsequent decryption or jamming.

[0003] Because frequency-hopping signals exhibit frequency domain sparsity, fully utilizing this characteristic can significantly reduce the signal sampling amount. Therefore, in recent years, compressed sensing theory has been applied to blind estimation of frequency-hopping signal parameters. While there has been considerable research on blind estimation of frequency-hopping signal parameters based on compressed sensing, the estimation accuracy of these algorithms needs improvement when the signal-to-noise ratio (SNR) is below 0 dB, and their computational complexity is high when the SNR is above 0 dB. Summary of the Invention

[0004] The purpose of this invention is to provide a blind estimation method for frequency hopping signal parameters based on atomic features, thereby obtaining the frequency hopping pattern of non-self frequency hopping signals, and having high estimation accuracy under low signal-to-noise ratio conditions and low computational complexity under high signal-to-noise ratio conditions.

[0005] The technical solution to achieve the objective of this invention is as follows: a blind estimation method for frequency hopping signal parameters based on atomic features, comprising the following steps:

[0006] S1. Window and sliding window processing are applied to the frequency hopping signal to obtain signals within different windows;

[0007] S2. Compress and sample the signals within different windows to obtain the inner product distribution of the signals within different windows;

[0008] S3. Extract atomic features from the inner product distribution and arrange the atomic features in chronological order to obtain the distribution pattern. Based on the distribution pattern, estimate the time range of the jump moment.

[0009] S4. Construct a block-diagonalized Fourier orthogonal matrix. The boundary points of this matrix traverse the time range according to the set search step size. Estimate the jump time by the change of sparse coefficients, and then obtain the frequency hopping period and hopping speed.

[0010] S5. Use the orthogonal matching pursuit algorithm to estimate the frequency of each hop signal, and combine the hopping time to obtain the frequency hopping pattern.

[0011] Furthermore, in S1, the frequency-hopping signal is windowed and slide-windowed to obtain signals within different windows, as follows:

[0012] Step S11: According to the Nyquist sampling theorem, sample the intercepted frequency hopping signal at a sampling frequency that is twice the highest frequency in the signal.

[0013] Step S12: Window and slide window processing is performed on the signal sampling data to extract data within different windows. The window length is less than the data volume of any single hop signal, and the sliding step size is set to one-quarter of the window length.

[0014] Furthermore, in S2, compressed sampling is performed on the signals within different windows to obtain the inner product distribution of the signals within different windows, as follows:

[0015] Step S21: Compress and sample the captured data within the window, i.e., multiply it by a Gaussian random measurement matrix on the left to obtain the compressed measurement value of the data within the window, as shown in the following formula:

[0016] y = Φx

[0017] Where y is the compressed measurement value, x is the data within the window, and Φ is the Gaussian random measurement matrix;

[0018] Step S22: Perform an inner product between the compressed measurement value y and the recovery matrix Θ to obtain the inner product value and its distribution, as shown below:

[0019] Θ=ΦΨ

[0020] Ω = |<Θ,y>|

[0021] Where Ψ is the discrete Fourier transform basis, Θ is the recovery matrix, and Ω is the inner product value.

[0022] Furthermore, the time range within which the transition moment is estimated in S3 is as follows:

[0023] Step S31: Extract the maximum value val1 and the second largest value val2 from the inner product value distribution, and obtain the atomic features according to the following formula:

[0024]

[0025] Step S32: Arrange the atomic features in chronological order to obtain the distribution;

[0026] Step S33: Lock the trough position in the atomic feature distribution. The data within the window corresponding to the trough position is the time range of the transition moment.

[0027] Furthermore, in S4, a block-diagonalized Fourier orthogonal matrix is ​​constructed. The boundary points of this matrix traverse the time range according to the set search step size. The transition time is estimated by the change of sparse coefficients, thereby obtaining the frequency hopping period and hopping speed, as detailed below:

[0028] Step S41: Construct a block-diagonalized Fourier orthogonal matrix. Since the window contains at most two-hop signals, this matrix is ​​represented as a bidiagonal Fourier orthogonal matrix. The size of this matrix is ​​(window length * window length), as shown in the following formula:

[0029]

[0030] Step S42: Based on the characteristic that the transition point is located in the middle part of the data within the window at the trough, the specific search range is obtained, as shown below:

[0031]

[0032] Where N is the window length and δ is the sliding step size;

[0033] Step S43: Move the boundary point K in the matrix according to the set search step size, and traverse the specific search range. The sparse coefficients for each movement are calculated and recorded using the orthogonal matching pursuit algorithm;

[0034] Step S44: Locate the largest sparse coefficient; the boundary point corresponding to this sparse coefficient is the transition time t. hop ;

[0035] Step S45: The difference between two adjacent transition times is the frequency hopping period, and the reciprocal of the frequency hopping period is the hopping speed, as shown in the following formula:

[0036]

[0037]

[0038] Among them, T hop For the frequency hopping period, V hop This refers to the jumping speed.

[0039] Furthermore, in S5, the frequency of each hop signal is estimated using the orthogonal matching pursuit algorithm, and the frequency hopping pattern is obtained by combining the hopping moments, as follows:

[0040] Step S51: Calculate the frequency of each hop signal using the orthogonal matching pursuit algorithm. The calculation formula is as follows:

[0041] pos = argmax(|<Θ, y>|)

[0042]

[0043] Among them, f s Where is the sampling rate, and pos is the position corresponding to the maximum value in the inner product distribution;

[0044] Step S52: Concatenate each hop signal in chronological order to obtain the frequency hopping pattern.

[0045] A blind estimation device for frequency hopping signal parameters based on atomic features, the device comprising:

[0046] The windowing and sliding window processing module is used to perform windowing and sliding window processing on the frequency hopping signal to obtain signals within different windows;

[0047] The compression sampling module is used to compress and sample signals within different windows to obtain the inner product distribution of signals within different windows;

[0048] The module for determining the time range of the jump moment is used to extract atomic features from the inner product value distribution, arrange the atomic features in chronological order to obtain the distribution pattern, and estimate the time range of the jump moment based on the distribution pattern.

[0049] The jump time determination module is used to construct a block-diagonalized Fourier orthogonal matrix. The boundary points of this matrix traverse the time range according to the set search step size. The jump time is estimated by the change of sparse coefficients, and then the frequency hopping period and hopping speed are obtained.

[0050] The frequency hopping pattern determination module uses an orthogonal matching pursuit algorithm to estimate the frequency of each hop signal and combines the hopping moments to obtain the frequency hopping pattern.

[0051] A mobile terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the blind estimation method for frequency hopping signal parameters based on atomic features when executing the program.

[0052] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the blind estimation method for frequency-hopping signal parameters based on atomic features.

[0053] Compared with the prior art, the present invention has the following significant advantages: (1) It combines the idea of ​​compressed sensing, which significantly reduces the amount of signal sampling and does not require signal reconstruction, thus having strong anti-noise performance; (2) Without filtering out noise or reconstructing the signal, it combines the constructed block diagonalized Fourier orthogonal matrix and orthogonal matching pursuit algorithm to obtain the frequency hopping pattern of the signal; (3) The method is flexible and adaptable. By adjusting the search step size and compression ratio, it can adapt to different signal-to-noise ratio environments. It has high estimation accuracy under low signal-to-noise ratio conditions and low computational complexity under high signal-to-noise ratio conditions. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the principle of the present invention.

[0055] Figure 2 This is a schematic diagram showing the distribution of inner product values.

[0056] Figure 3 This is a schematic diagram showing the distribution of atomic characteristics.

[0057] Figure 4 This is a schematic diagram of two diagonal Fourier orthogonal matrices.

[0058] Figure 5 This is a flowchart for frequency estimation.

[0059] Figure 6 The simulation results show the comparison under different search step sizes.

[0060] Figure 7 The figures show a comparison of simulations under different compression ratios.

[0061] Figure 8 The simulation diagrams show the comparison of different algorithms. Detailed Implementation

[0062] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0063] like Figures 1-5 As shown, the blind estimation method for frequency hopping signal parameters based on atomic features of this invention has the following specific implementation steps:

[0064] S1. Window and sliding window processing are applied to the frequency hopping signal to obtain signals within different windows;

[0065] S2. Compress and sample the signals within different windows to obtain the inner product distribution of the signals within different windows;

[0066] S3. Extract atomic features from the inner product distribution and arrange the atomic features in chronological order to obtain the distribution pattern. Based on the distribution pattern, estimate the time range of the jump moment.

[0067] S4. Construct a block-diagonalized Fourier orthogonal matrix. The boundary points of this matrix traverse the time range according to the set search step size. Estimate the jump time by the change of sparse coefficients, and then obtain the frequency hopping period and hopping speed.

[0068] S5. Use the orthogonal matching pursuit algorithm to estimate the frequency of each hop signal, and combine the hopping time to obtain the frequency hopping pattern.

[0069] As a specific example, S1 performs windowing and sliding windowing on the frequency hopping signal to obtain signals within different windows, as follows:

[0070] Step S11: According to the Nyquist sampling theorem, sample the intercepted frequency hopping signal at a sampling frequency that is twice the highest frequency in the signal.

[0071] Step S12: Window and slide window processing is performed on the signal sampling data to extract data within different windows. The window length is less than the data volume of any single hop signal, and the sliding step size is set to one-quarter of the window length.

[0072] As a specific example, S2 performs compressed sampling on signals within different windows to obtain the inner product distribution of signals within different windows, as follows:

[0073] Step S21: Compress and sample the captured data within the window, i.e., multiply it by a Gaussian random measurement matrix on the left to obtain the compressed measurement value of the data within the window, as shown in the following formula:

[0074] y = Φx

[0075] Where y is the compressed measurement value, x is the data within the window, and Φ is the Gaussian random measurement matrix;

[0076] Step S22: Perform an inner product between the compressed measurement value y and the recovery matrix Θ to obtain the inner product value and its distribution, as shown below:

[0077] Θ=ΦΨ

[0078] Ω = |<Θ,y>|

[0079] Where Ψ is the discrete Fourier transform basis, Θ is the recovery matrix, and Ω is the inner product value.

[0080] As a specific example, the time range within which the transition moment is estimated in S3 is as follows:

[0081] Step S31: Extract the maximum value val1 and the second largest value val2 from the inner product value distribution, and obtain the atomic features according to the following formula:

[0082]

[0083] Step S32: Arrange the atomic features in chronological order to obtain the distribution;

[0084] Step S33: Lock the trough position in the atomic feature distribution. The data within the window corresponding to the trough position is the time range of the transition moment.

[0085] As a specific example, S4 constructs a block-diagonalized Fourier orthogonal matrix. The boundary points of this matrix traverse the time range according to the set search step size. The transition time is estimated by the change of sparse coefficients, thereby obtaining the frequency hopping period and hopping speed, as detailed below:

[0086] Step S41: Construct a block-diagonalized Fourier orthogonal matrix. Since the window contains at most two-hop signals, this matrix is ​​represented as a bidiagonal Fourier orthogonal matrix. The size of this matrix is ​​(window length * window length), as shown in the following formula:

[0087]

[0088] Step S42: Based on the characteristic that the transition point is located in the middle part of the data within the window at the trough, the specific search range is obtained, as shown below:

[0089]

[0090] Where N is the window length and δ is the sliding step size;

[0091] Step S43: Move the boundary point K in the matrix according to the set search step size, and traverse the specific search range. The sparse coefficients for each movement are calculated and recorded using the orthogonal matching pursuit algorithm;

[0092] Step S44: Locate the largest sparse coefficient; the boundary point corresponding to this sparse coefficient is the transition time t. hop ;

[0093] Step S45: The difference between two adjacent transition times is the frequency hopping period, and the reciprocal of the frequency hopping period is the hopping speed, as shown in the following formula:

[0094]

[0095]

[0096] Among them, T hop For the frequency hopping period, V hop This refers to the jumping speed.

[0097] As a specific example, S5 uses the orthogonal matching pursuit algorithm to estimate the frequency of each hop signal, and combines the hopping moments to obtain the frequency hopping pattern, as follows:

[0098] Step S51: Calculate the frequency of each hop signal using the orthogonal matching pursuit algorithm. The calculation formula is as follows:

[0099] pos = argmax(|<Θ, y>|)

[0100]

[0101] Among them, f s Where is the sampling rate, and pos is the position corresponding to the maximum value in the inner product distribution;

[0102] Step S52: Concatenate each hop signal in chronological order to obtain the frequency hopping pattern.

[0103] The present invention also provides a blind estimation device for frequency hopping signal parameters based on atomic features, the device comprising:

[0104] The windowing and sliding window processing module is used to perform windowing and sliding window processing on the frequency hopping signal to obtain signals within different windows;

[0105] The compression sampling module is used to compress and sample signals within different windows to obtain the inner product distribution of signals within different windows;

[0106] The module for determining the time range of the jump moment is used to extract atomic features from the inner product value distribution, arrange the atomic features in chronological order to obtain the distribution pattern, and estimate the time range of the jump moment based on the distribution pattern.

[0107] The jump time determination module is used to construct a block-diagonalized Fourier orthogonal matrix. The boundary points of this matrix traverse the time range according to the set search step size. The jump time is estimated by the change of sparse coefficients, and then the frequency hopping period and hopping speed are obtained.

[0108] The frequency hopping pattern determination module uses an orthogonal matching pursuit algorithm to estimate the frequency of each hop signal and combines the hopping moments to obtain the frequency hopping pattern.

[0109] The present invention also provides a mobile terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the aforementioned blind estimation method for frequency hopping signal parameters based on atomic features when executing the program.

[0110] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the blind estimation method for frequency hopping signal parameters based on atomic features.

[0111] The advantages of the present invention are demonstrated below with reference to the experiments in the embodiments.

[0112] Example

[0113] In this embodiment, the experimental platform is MATLAB, and the experimental data is the frequency hopping signal generated by simulation. The simulation parameters of the frequency hopping signal are shown in Table 1.

[0114] Table 1 Simulation Parameters

[0115]

[0116] Experiment 1: Comparative Experiment under Different Search Step Sizes

[0117] The ambient noise is additive white Gaussian noise, the window length is 5120, the sliding step size is 1280, and the observation length is 160. The compression ratio is the ratio of the sliding window length before compressed sampling to the observation length after compressed sampling; therefore, the compression ratio is 32. In this experiment, the search step size was set to 128 and 32 respectively, and the simulation results are as follows. Figure 6 As shown.

[0118] Simulation results show that reducing the search step size can improve estimation performance and reduce estimation error. When the signal-to-noise ratio (SNR) is greater than 0 dB, the improvement in estimation performance is not significant; when the SNR is less than 0 dB, the improvement is more pronounced. This is because reducing the search step size allows for more points to be searched, thus reducing the gap between adjacent searches. However, reducing the search step size increases the running time, so a balance needs to be struck between computational complexity and parameter estimation accuracy.

[0119] Experiment 2: Comparative Experiments Under Different Compression Ratios

[0120] The ambient noise was additive white Gaussian noise, the window length was 5120, the sliding step size was 1280, and the search step size was 128. In this experiment, the compression ratios were set to 160, 64, 32, and 16, respectively. The simulation results are as follows: Figure 7 As shown.

[0121] Simulation results show that reducing the compression ratio decreases the estimation error of the takeoff time. When the signal-to-noise ratio (SNR) is less than 0 dB, the improvement in estimation performance due to the reduction in compression ratio is more significant; when the SNR is greater than 0 dB, the improvement is less significant. A reduction in compression ratio implies an increase in computational complexity; therefore, a trade-off must be made between computational complexity and parameter estimation accuracy.

[0122] Experiment 3: Comparison Experiment of Different Algorithms

[0123] The ambient noise is additive white Gaussian noise, the window length is 5120, the sliding step size is 1280, the search step size is 128, and the compression ratio is 32. This experiment compares the invention with the smoothed pseudo-Wigner-Will distribution (SPWVD), and the simulation results are as follows: Figure 8 As shown.

[0124] Simulation results show that, under conditions where the signal-to-noise ratio (SNR) is less than 2 dB, the estimation error of this invention is lower than that of SPWVD; however, under conditions where the SNR is greater than 2 dB, the estimation error is higher than that of SPWVD. Therefore, the estimation performance of this invention is better under low SNR conditions, and it has strong noise resistance.

[0125] Experiment 4: Algorithm Complexity Analysis and Comparison

[0126] This experiment will analyze and compare the computational complexity of the present invention and the blind estimation method for frequency hopping signal parameters based on compressed sensing. The computational complexity is characterized by the number of floating-point multiplication operations, and the analysis results are shown in Table 2.

[0127] Table 2 Comparison of computational complexity of different methods

[0128]

[0129] As shown in Table 2, both methods have the same computational complexity in frequency estimation, but differ in computational complexity in accurate estimation of transition times. By setting the compression ratio and search step size, the computational complexity of this invention can be made much smaller than that of the blind estimation method for frequency hopping signal parameters based on compressed sensing.

[0130] In summary, this invention optimizes noise immunity at low signal-to-noise ratios and improves computational speed at high signal-to-noise ratios. Furthermore, this invention eliminates the need for signal reconstruction, significantly reducing the amount of signal sampling data.

Claims

1. A blind estimation method for frequency-hopping signal parameters based on atomic features, characterized in that, The steps are as follows: S1. Window and sliding window processing are applied to the frequency hopping signal to obtain signals within different windows; S2. Compress and sample the signals within different windows to obtain the inner product distribution of the signals within different windows; S3. Extract atomic features from the inner product distribution and arrange the atomic features in chronological order to obtain the distribution pattern. Based on the distribution pattern, estimate the time range of the jump moment. S4. Construct a block-diagonalized Fourier orthogonal matrix. The boundary points of this matrix traverse the time range according to the set search step size. Estimate the jump time by the change of sparse coefficients, and then obtain the frequency hopping period and hopping speed. S5. Use the orthogonal matching pursuit algorithm to estimate the frequency of each hop signal, and combine the hopping time to obtain the frequency hopping pattern.

2. The blind estimation method for frequency hopping signal parameters based on atomic features according to claim 1, characterized in that, In S1, the frequency-hopping signal is windowed and slide-windowed to obtain signals within different windows, as detailed below: Step S11: According to the Nyquist sampling theorem, sample the intercepted frequency hopping signal at a sampling frequency that is twice the highest frequency in the signal. Step S12: Window and slide window processing is performed on the signal sampling data to extract data within different windows. The window length is less than the data volume of any single hop signal, and the sliding step size is set to one-quarter of the window length.

3. The blind estimation method for frequency hopping signal parameters based on atomic features according to claim 1, characterized in that, In S2, compressed sampling is performed on signals within different windows to obtain the inner product distribution of signals within different windows, as detailed below: Step S21: Compress and sample the captured data within the window, i.e., multiply it by a Gaussian random measurement matrix on the left to obtain the compressed measurement value of the data within the window, as shown in the following formula: y = Φx Where y is the compressed measurement value, x is the data within the window, and Φ is the Gaussian random measurement matrix; Step S22: Perform an inner product between the compressed measurement value y and the recovery matrix Θ to obtain the inner product value and its distribution, as shown below: Θ=ΦΨ Ω = |<Θ,y>| Where Ψ is the discrete Fourier transform basis, Θ is the recovery matrix, and Ω is the inner product value.

4. The blind estimation method for frequency hopping signal parameters based on atomic features according to claim 1, characterized in that, S3 estimates the time range of the transition moment as follows: Step S31: Extract the maximum value val1 and the second largest value val2 from the inner product value distribution, and obtain the atomic features according to the following formula: Step S32: Arrange the atomic features in chronological order to obtain the distribution; Step S33: Lock the trough position in the atomic feature distribution. The data within the window corresponding to the trough position is the time range of the transition moment.

5. The blind estimation method for frequency hopping signal parameters based on atomic features according to claim 1, characterized in that, In S4, a block-diagonalized Fourier orthogonal matrix is ​​constructed. The boundary points of this matrix traverse the time range according to the set search step size. The transition time is estimated by the change of sparse coefficients, and then the frequency hopping period and hopping speed are obtained, as follows: Step S41: Construct a block-diagonalized Fourier orthogonal matrix. Since the window contains at most two-hop signals, this matrix is ​​represented as a bidiagonal Fourier orthogonal matrix. The size of this matrix is ​​(window length * window length), as shown in the following formula: Step S42: Based on the characteristic that the transition point is located in the middle part of the data within the window at the trough, the specific search range is obtained, as shown below: Where N is the window length and δ is the sliding step size; Step S43: Move the boundary point K in the matrix according to the set search step size, and traverse the specific search range. The sparse coefficients for each movement are calculated and recorded using the orthogonal matching pursuit algorithm; Step S44: Locate the largest sparse coefficient; the boundary point corresponding to this sparse coefficient is the transition time t. hop ; Step S45: The difference between two adjacent transition times is the frequency hopping period, and the reciprocal of the frequency hopping period is the hopping speed, as shown in the following formula: Among them, T hop For the frequency hopping period, V hop This refers to the jumping speed.

6. The blind estimation method for frequency hopping signal parameters based on atomic features according to claim 1, characterized in that, In S5, the frequency of each hop signal is estimated using the orthogonal matching pursuit algorithm, and the frequency hopping pattern is obtained by combining the hopping moments, as follows: Step S51: Calculate the frequency of each hop signal using the orthogonal matching pursuit algorithm. The calculation formula is as follows: pos = argmax(|<Θ, y>|) Among them, f s Where is the sampling rate, and pos is the position corresponding to the maximum value in the inner product distribution; Step S52: Concatenate each hop signal in chronological order to obtain the frequency hopping pattern.

7. A blind estimation device for frequency hopping signal parameters based on atomic features, characterized in that, The device includes: The windowing and sliding window processing module is used to perform windowing and sliding window processing on the frequency hopping signal to obtain signals within different windows; The compression sampling module is used to compress and sample signals within different windows to obtain the inner product distribution of signals within different windows; The module for determining the time range of the jump moment is used to extract atomic features from the inner product value distribution, arrange the atomic features in chronological order to obtain the distribution pattern, and estimate the time range of the jump moment based on the distribution pattern. The jump time determination module is used to construct a block-diagonalized Fourier orthogonal matrix. The boundary points of this matrix traverse the time range according to the set search step size. The jump time is estimated by the change of sparse coefficients, and then the frequency hopping period and hopping speed are obtained. The frequency hopping pattern determination module uses an orthogonal matching pursuit algorithm to estimate the frequency of each hop signal and combines the hopping moments to obtain the frequency hopping pattern.

8. A mobile terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the blind estimation method for frequency hopping signal parameters based on atomic features as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the blind estimation method for frequency hopping signal parameters based on atomic features as described in any one of claims 1 to 6.

Images