Frequency hopping signal parameter estimation method and system based on time-frequency analysis and block clustering

Abstract

The application discloses a frequency hopping signal parameter estimation method and system based on time-frequency analysis and block clustering, which comprises the following steps: reading frequency hopping signal IQ data, performing time-frequency analysis to obtain a corresponding time-frequency matrix; performing binaryzation processing on the time-frequency matrix according to a dynamic threshold to obtain a binaryzation time-frequency matrix, and preliminarily detecting the position of the frequency hopping signal through the binaryzation time-frequency matrix; performing block processing on the binaryzation time-frequency matrix, clustering the frequency hopping signal detected in each block, calculating the frequency hopping signal parameters according to the clustering results, comparing the frequency hopping signal information obtained by the current block clustering with the frequency hopping signal information obtained by the last block clustering after the clustering of each block is completed, judging whether the frequency hopping signal information belongs to the same hopping data, and updating the frequency hopping signal parameters, and obtaining the final frequency hopping signal parameter estimation result according to all the frequency hopping signal parameters. The application can achieve a balance between time and performance, and realize accurate parameter estimation.

Description

Technical Field

[0001] This invention relates to the field of frequency hopping communication technology, and specifically to a method and system for estimating frequency hopping signal parameters based on time-frequency analysis and block clustering. Background Technology

[0002] Frequency-hopping signal IQ data is a sequence of in-phase (I) and quadrature (Q) components obtained at the receiver after down-converting and quadrature demodulating the original radio frequency signal. It reflects the instantaneous amplitude and phase information of the signal and is used for baseband representation in the carrier modulation and demodulation process. Frequency-hopping signal IQ data retains the characteristics of carrier frequency and frequency-hopping mode changes and is usually used as the basis for time-frequency analysis and frequency-hopping parameter estimation.

[0003] Current technologies typically employ time-frequency analysis methods such as Short-Time Fourier Transform (SFT) for frequency-hopping signal detection. This involves performing a SFT on the signal to obtain its time-frequency distribution, and then determining a suitable fixed threshold to identify the presence of the frequency-hopping signal. While the SFT method is simple and easy to implement, it suffers from poor time-frequency focusing. Furthermore, directly using SFT with a fixed threshold only achieves coarse detection with low accuracy and cannot adapt to changes in signal energy over time and frequency. This can lead to the loss of true signal components in low-energy or noisy regions, while generating numerous false detections in high-noise environments. Additionally, the fixed threshold is sensitive to amplitude differences in different frequency-hopping segments, and the same threshold cannot adequately detect multiple frequency-hopping signals. Moreover, the fixed threshold method lacks adaptability to sudden interference and environmental changes, requiring complex pre-processing or post-processing to obtain reliable results, increasing system complexity and computational overhead. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for estimating frequency hopping signal parameters based on time-frequency analysis and block clustering, which can balance estimation efficiency and accuracy, achieve a balance in time and performance, and realize accurate estimation of frequency hopping signal parameters.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for estimating parameters of frequency-hopping signals based on time-frequency analysis and block clustering includes: Step S1: Read the frequency hopping signal IQ data and perform time-frequency analysis on the frequency hopping signal IQ data to obtain the corresponding time-frequency matrix; Step S2: The time-frequency matrix is ​​binarized according to the dynamic threshold to distinguish the location of noise and the location of the spectrum to obtain a binarized time-frequency matrix. The location of the frequency hopping signal is initially detected by the binarized time-frequency matrix. The dynamic threshold is obtained by dividing the time-frequency matrix into multiple sub-matrices and determining the distribution trend of the values ​​of each sub-matrix. Step S3: Divide the binarized time-frequency matrix into blocks, and cluster the frequency hopping signals detected in each block. During the clustering process, the start and end times and frequency parameters of each cluster are continuously updated by gradually increasing the number of points in the same cluster. The frequency hopping signal parameters are calculated based on the clustering results. After each block clustering is completed, the frequency hopping signal information obtained from the current block clustering is compared with the frequency hopping signal information obtained from the previous block clustering to determine whether they belong to the same hop data and update the frequency hopping signal parameters. The final frequency hopping signal parameter estimation result is obtained based on all the frequency hopping signal parameters.

[0006] Further, in step S1, the frequency hopping signal IQ data is analyzed using spectral transform to obtain the corresponding time-frequency matrix, wherein the number of rows of the time-frequency matrix is ​​the window length used when performing short-time Fourier transform, and the number of columns is determined by the length of the IQ data and the window length used when performing short-time Fourier transform.

[0007] Furthermore, the calculation expression for the time-frequency matrix is ​​as follows:

[0008] In the above formula, t represents the current time point of the analysis, and f represents the frequency component of the current analysis. Let be the integral variable obtained by summing over all time signals. This is a short-time Fourier transform; The expression for calculating the number of columns in the time-frequency matrix is ​​as follows:

[0009] In the above formula, floor is the floor function. The length of the IQ data, This refers to the window length used when performing a short-time Fourier transform on the IQ data.

[0010] Further, step S2 includes: The time-frequency matrix is ​​divided into multiple sub-time-frequency matrices according to the hopping rate range and time resolution of the frequency hopping signal; The number of elements in the time-frequency matrix that are greater than and less than the average value of each sub-time-frequency matrix is ​​counted to obtain multiple sets of corresponding first and second counts; Calculate the difference between the first and second quantities in each group, and take the maximum value among the differences in each group as the dynamic threshold; Elements in the time-frequency matrix that are greater than the dynamic threshold are marked as first values, and elements that are less than the dynamic threshold are marked as second values, to obtain a binarized time-frequency matrix, wherein the first value indicates the presence of a frequency hopping signal, and the second value indicates the absence of a frequency hopping signal.

[0011] Furthermore, the step of dividing the time-frequency matrix into multiple sub-time-frequency matrices based on the hopping rate range and time resolution of the frequency-hopping signal specifically includes: The minimum number of data points per hop of the frequency hopping signal is calculated based on the hopping speed range and time resolution of the frequency hopping signal. The number of sub-time-frequency matrix divisions is determined based on the minimum values ​​of the number of rows, columns, and points of the time-frequency matrix. The number of rows and columns of each sub-time-frequency matrix is ​​determined based on the minimum value of the points.

[0012] Furthermore, the expression for calculating the minimum number of points is:

[0013]

[0014] In the above formula, DT represents the maximum hopping rate range of the frequency hopping signal, and DT represents the time resolution. The length of the IQ data, This refers to the window length used when performing a short-time Fourier transform on the IQ data. is the sampling rate of the frequency hopping signal, and floor is the floor function; The formulas for calculating the number of rows and columns of each sub-time-frequency matrix are as follows:

[0015] In the above formula, Points_min is the minimum number of points.

[0016] Furthermore, in step S3, the block-based processing of the binarized time-frequency matrix specifically includes: The minimum number of data points per hop of the frequency hopping signal is calculated based on the hopping speed range and time resolution of the frequency hopping signal. The number of blocks is determined based on the maximum number of columns and points in the time-frequency matrix, so that each block contains at least one hop of the frequency-hopping signal data.

[0017] Further, step S3 includes: The binarized time-frequency matrix B is divided into several blocks, each block containing at least one complete hop of the frequency-hopping signal. The points containing the frequency-hopping signal in each block are recorded to form a point set. Set the neighborhood radius and minimum number of points for clustering, and cluster all frequency hopping signals within the point set to obtain multiple cluster categories. During the clustering process, the start and end times and frequency parameters of each cluster are continuously updated by gradually increasing the number of points in the same category. The frequency parameters include the start frequency SF, the end frequency EF, and the center frequency FC. For each cluster category, the corresponding frequency hopping signal parameters are calculated based on the start and end times and frequency parameters of each cluster. After the current block clustering is completed, the information of the frequency hopping signals represented by all clusters in the previous block is compared with the information of the frequency hopping signals represented by all clusters in the current data block to determine whether the two clusters at the connection point belong to the same hop data. If the distance between the two clusters is less than the set neighborhood radius eps, they are determined to belong to the same class and correspond to the same frequency hopping signal, and the corresponding frequency hopping signal parameters are updated.

[0018] Furthermore, the frequency hopping signal parameters include one or more of the following: dwell time, signal bandwidth, frequency hopping frequency set, frequency hopping bandwidth, and frequency hopping frequency point interval. The dwell time of the current clustered frequency hopping signal is calculated based on the clustering end time and clustering start time. The signal bandwidth of the current clustered frequency hopping signal is calculated based on the clustering end frequency point and the clustering start frequency point. The center frequency point of the current clustered frequency hopping signal is calculated based on the clustering end frequency point and the current frequency hopping signal bandwidth.

[0019] A frequency hopping signal parameter measurement system based on time-frequency analysis and block clustering includes a microprocessor and a memory interconnected, wherein the microprocessor is programmed or configured to execute a frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering.

[0020] Compared with the prior art, the advantages of the present invention are as follows: This invention first obtains a time-frequency matrix using time-frequency analysis, then binarizes the matrix using a dynamic threshold. The location of the frequency-hopping signal is initially detected using this binarized matrix. The dynamic threshold adapts to changes in local noise levels and signal energy, preserving weak frequency-hopping components, suppressing noise interference, and improving the detection rate under low signal-to-noise ratio conditions. Compared to a fixed threshold, the dynamic threshold eliminates the need for manual parameter tuning between frequency bands, taking into account amplitude differences in different frequency-hopping bands, reducing false alarm rates, and enhancing algorithm robustness and automation. The binarized time-frequency matrix is ​​then further divided into blocks, and the detected frequency-hopping signals in each block are clustered. This block-based clustering method better captures local frequency-hopping features and effectively calculates the parameters of the initially detected frequency-hopping signals within each block. Finally, the results of the combined block clustering are used to obtain the frequency-hopping signal parameter estimation result. This effectively improves the accuracy of frequency-hopping signal parameter estimation while increasing processing speed, achieving a balance between time and performance. It enables fast and accurate parameter estimation that is less susceptible to environmental changes and interference. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering in this embodiment.

[0022] Figure 2 This is a time-frequency diagram of the time-frequency matrix obtained after spectral transformation of the frequency-hopping signal IQ data in a specific application embodiment.

[0023] Figure 3 This is a binarized time-frequency matrix diagram after binarizing the time-frequency matrix using a dynamic threshold in a specific application embodiment.

[0024] Figure 4 This is a diagram of frequency hopping signals accurately detected by using the DBSCAN clustering algorithm to perform block clustering on the binarized time-frequency matrix in a specific application embodiment. Detailed Implementation

[0025] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0026] like Figure 1 As shown, the frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering in this embodiment includes: S1, Read the frequency hopping signal IQ data, and perform time-frequency analysis on the frequency hopping signal IQ data to obtain the corresponding time-frequency matrix; S2, the time-frequency matrix is ​​binarized according to the dynamic threshold to distinguish the location of noise and the location of the spectrum to obtain the binarized time-frequency matrix. The location of the frequency hopping signal is initially detected by the binarized time-frequency matrix. The dynamic threshold is obtained by dividing the time-frequency matrix into multiple sub-matrices and determining it according to the distribution trend of the values ​​of each sub-matrix. S3. The binarized time-frequency matrix is ​​divided into blocks, and the frequency hopping signals detected in each block are clustered. During the clustering process, the start and end times and frequency parameters of each cluster are continuously updated by gradually increasing the number of points in the same cluster. The frequency hopping signal parameters are calculated based on the clustering results. After the clustering of each block is completed, the frequency hopping signal information obtained from the current block clustering is compared with the frequency hopping signal information obtained from the previous block clustering to determine whether they belong to the same hop data and update the frequency hopping signal parameters. The final frequency hopping signal parameter estimation result is obtained based on all the frequency hopping signal parameters.

[0027] It is understood that in this embodiment, the time-frequency matrix is ​​binarized based on a dynamic threshold, and the position of the frequency hopping signal is initially detected through the binarized time-frequency matrix. The dynamic threshold can improve the detection rate under low signal-to-noise ratio, take into account the amplitude differences of different frequency hopping bands, reduce the false alarm rate, enhance the robustness and automation level of the algorithm, reduce the impact of non-stationary noise on the binarization results, and facilitate more accurate parameter estimation in the future. In this embodiment, the binarized time-frequency matrix is ​​divided into blocks, and the frequency hopping signals detected in each block are clustered, which can reduce the amount of data for a single clustering and improve the processing speed. By comparing the information between blocks, the overall detection integrity and the accuracy of the final frequency hopping parameter estimation can be improved.

[0028] In step S1 of this embodiment, the frequency-hopping signal IQ data is analyzed using spectral transform to obtain the corresponding time-frequency matrix. The number of rows in the time-frequency matrix is ​​the window length used when performing short-time Fourier transform, and the number of columns is determined by the length of the IQ data and the window length used when performing short-time Fourier transform.

[0029] Time-frequency analysis methods include Short-Time Fourier Transform (STFT), Wegener's Distribution (WVD) correlation transform, and Spectral Transform (SP). Among these, STFT is simple and easy to implement, but has poor time-frequency focusing. WVD has excellent time-frequency resolution, but suffers from significant cross-term interference. SP, on the other hand, has very low information entropy, excellent focusing performance, and no cross-term interference; its algorithm is simple and easy to implement in hardware. Specifically, Spectral Transform (SP) is used to process the read frequency-hopping signal IQ data to obtain the time-frequency matrix. Let the acquired IQ data be... Since the square of the modulus of the spectral transform (SP), i.e., the short-time Fourier transform (STFT), also belongs to the quadratic distribution, the corresponding expression for calculating the time-frequency matrix is: (1) In the above formula, t represents the current time point of the analysis, and f represents the frequency component of the current analysis. Let be the integral variable obtained by summing over all time signals. This is a short-time Fourier transform; The expression for calculating the number of columns in the time-frequency matrix is: (2) In the above formula, floor is the floor function. The length of the IQ data. The window length used when performing a short-time Fourier transform on IQ data.

[0030] The time-frequency matrix of the IQ data can be calculated according to equation (1). The size of the matrix is: mf rows, nt columns, where mf = N1, and nt is calculated according to equation (2). The STFT transform can be understood as a windowed Fourier transform (FFT). In step S1, let the IQ data length be Len, the window length (i.e., step size) selected by the STFT transform be N1, the overlap length of the two windows be N1 / 2, and the number of points of the Fourier transform be Nfft.

[0031] In this embodiment, step S2 includes: S21. Divide the time-frequency matrix into multiple sub-time-frequency matrices according to the hopping speed range and time resolution of the frequency hopping signal; S22. Count the number of elements in the time-frequency matrix that are greater than and less than the average value of each sub-time-frequency matrix to obtain multiple sets of corresponding first and second counts; S23. Calculate the difference between the first and second quantities in each group, and use the maximum value of the difference in each group as the dynamic threshold. S24. Mark the elements in the time-frequency matrix that are greater than the dynamic threshold as the first value (e.g., 1) and the elements that are less than the dynamic threshold as the second value (e.g., 0) to obtain a binarized time-frequency matrix, where the first value indicates that the frequency hopping signal exists and the second value indicates that the frequency hopping signal does not exist.

[0032] In this embodiment, step S21, which divides the time-frequency matrix into multiple sub-time-frequency matrices according to the hopping speed range and time resolution of the frequency-hopping signal, specifically includes: The minimum number of data points per hop of the frequency hopping signal is calculated based on the hopping speed range and time resolution of the frequency hopping signal. The number of sub-time-frequency matrix divisions is determined based on the minimum values ​​of the number of rows, columns, and points in the time-frequency matrix; The number of rows and columns of each sub-time-frequency matrix is ​​determined based on the minimum number of points.

[0033] The following is an example of using the method of the present invention to perform preliminary detection of the position of a frequency hopping signal in a specific application embodiment. The detailed process is as follows: S21, the time-frequency matrix Divide the data into m small r×r matrices. Assuming the hopping rate range of the detected frequency hopping data is HopRate: 100~1000 hop / s, then the dwell time (DwellTime) of one hop is in the range of 0.01~0.001s. Therefore, the number of points (Points) contained in each hop is 0.001 / DT~0.01 / DT. The expression for calculating the minimum number of points (Points_min) in each hop of the frequency hopping signal is as follows: (3)

[0034] In the above formula, DT represents the maximum hopping rate range of the frequency hopping signal, and DT represents the time resolution. The length of the IQ data. The window length used when performing a short-time Fourier transform on IQ data. is the sampling rate of the frequency hopping signal, and floor is the floor function. In this embodiment, =1000 hops / s.

[0035] The number of sub-time-frequency matrix divisions m is calculated as follows: divide Points_min by mf to get m1, and divide Points_min by nt to get n1, thus obtaining the number of sub-time-frequency matrix divisions: m = m1 × n1.

[0036] The expressions for calculating the number of rows r and the number of columns r of each sub-time-frequency matrix are as follows: (4) In this way, the smallest possible average value within the matrix can be used as the dynamic threshold.

[0037] S22. Calculate the average value of each sub-time-frequency matrix. ,in ; S23. For the average value of each sub-time-frequency matrix, calculate the statistical time-frequency matrix. middle or The number of, if Then it is recorded as ,like Then it is recorded as ; S24, Calculate each Find the difference between the two, and find its maximum value. Then, find the value of the difference when it reaches its maximum value. This serves as a dynamic threshold T. At this point, noise points, blank areas, and locations containing the spectrum can be better distinguished. S25. After binarization of the time-frequency matrix, the resulting matrix is ​​a 0,1 matrix, denoted as... In this binary matrix, the positions with a value of 1 represent the locations of the frequency-hopping signals obtained from coarse detection. The horizontal axis of each element in the matrix represents time, and the vertical axis represents frequency. The expression is as follows: (5) In step S3 of this embodiment, the block-based processing of the binarized time-frequency matrix specifically includes: The minimum number of data points per hop of the frequency hopping signal is calculated based on the hopping speed range and time resolution of the frequency hopping signal. The number of blocks is determined based on the maximum number of columns and points in the time-frequency matrix, so that each block contains at least one hop of the frequency-hopping signal data.

[0038] In this embodiment, the frequency hopping signal parameters include: dwell time, signal bandwidth, frequency hopping frequency set, frequency hopping bandwidth, and frequency hopping frequency interval, etc. The dwell time of the current clustered frequency hopping signal can be calculated based on the clustering end time and clustering start time. The signal bandwidth of the current clustered frequency hopping signal can be calculated based on the clustering end frequency and clustering start frequency. The center frequency of the current clustered frequency hopping signal can be calculated based on the clustering end frequency and the current frequency hopping signal bandwidth.

[0039] Specifically, in step S3, the binarized time-frequency matrix B is first divided into several blocks, each block containing at least one complete hop of the frequency-hopping signal. Points containing the frequency-hopping signal within each block are recorded to form a point set. A neighborhood radius and a minimum number of points are set for clustering. All frequency-hopping signals within the point set are clustered to obtain multiple cluster categories. During the clustering process, the start and end times and frequency parameters of each cluster are continuously updated by gradually increasing the number of points within the same category. The frequency parameters include the start frequency SF, end frequency EF, and center frequency FC. For each cluster category, the corresponding frequency-hopping signal parameters are calculated based on the start and end times and frequency parameters of each cluster. After the current block is clustered, the information of the frequency-hopping signals represented by all clusters in the previous block is compared with the information of the frequency-hopping signals represented by all clusters in the current data block to determine whether the two clusters at the connection point belong to the same hop data. If the distance between the two clusters is less than the set neighborhood radius eps, they are determined to belong to the same category, corresponding to the same frequency-hopping signal, and the corresponding frequency-hopping signal parameters are updated.

[0040] K-Means clustering is a local optima-based clustering algorithm. It is easy to understand, produces good clustering results, and has low computational complexity. However, it is very sensitive to the initial cluster centers; different initial cluster center choices will affect the number of subsequent iterations and the clustering results. Furthermore, it is not suitable for handling overly discrete classifications, classifications with large differences in sample categories, or classifications with non-convex shapes. Mean-shift clustering is a center-based algorithm that updates the candidate center points to achieve the average value of points in a sliding window. It does not require selecting the number of clusters, but the choice of the sliding window radius r has no fixed standard. The DBSCAN algorithm is a typical density-based clustering method that can discover clusters of any shape and has good noise recognition performance. (Frequency analysis of frequency hopping signals is also discussed.) Figure 1 Generally exhibiting a linear shape, it is very suitable for using the DBSCAN clustering algorithm. Therefore, this embodiment uses the DBSCAN clustering method to cluster the binary time-frequency matrix. In a specific application embodiment, the specific steps of step S3, using DBSCAN clustering for block clustering, are as follows: S31. Divide the binarized time-frequency matrix B into several blocks, denoted as `partnum`. The number of blocks, `partnum`, is calculated by dividing the number of columns in the time-frequency matrix by the maximum number of points, `Points_max`, ensuring that each part contains at least one complete hop of data. Since `Points_max = 1 / HopRate_min / DT = 0.01 / DT`, the number of points in each part, `partpoints`, must be greater than `Points_max`. Both `partnum` and `partpoints` must be integers. This determines the values. Record the points containing the signal within each part, forming a point set: `sigpoint`. This point set contains x-values ​​and y-values; x-values ​​represent time information, and y-values ​​represent frequency information. Using matrix block division saves time for subsequent clustering of each small block of data.

[0041] S32. Perform DBSCAN clustering on all points within the point set sigpoint in the current small data block. Use the distance (x, y coordinates) between two points as the judgment value of the neighborhood radius eps for DBSCAN clustering. Points with a distance less than the set eps are considered to belong to the same class. If the number of points within the circle of a core point with radius eps is less than the set minnb value, it is considered a noise point; otherwise, it is considered a signal point that meets expectations. Since the x and y coordinates of each point in the point set differ too much, it is difficult to set the neighborhood radius by directly calculating the distance. Therefore, when calculating the distance between two points, the two coordinates can be normalized to the same order of magnitude before calculation. Set the minimum frequency hopping interval of the frequency hopping signal that can be monitored to F_hop. The initial values ​​of the neighborhood radius eps and minnb are set according to the following understanding: the duration for a class to contain at least one hop of data Points_max points on the x-axis, which is Points_max × DT, and the interval between two classes on the y-axis to be at least one frequency hopping interval F_hop. The duration Points_max×DT and the frequency hopping interval F_hop are normalized to the same order of magnitude, and their binary values ​​are set as t_normal and f_normal, respectively. Therefore, the neighborhood radius eps of the clustering parameter is set to max(t_normal, f_normal) + 1, ensuring that the distance in both dimensions is large enough to distinguish different hop data. Since the binarized time-frequency matrix has been divided into blocks, each sub-block often contains some fragmented cluster data, which may not contain all the points of a complete cluster. Therefore, the other clustering parameter minnb is set to a moderate value of 10 (based on experience). When the number of points in eps is less than minnb, the cluster is considered to be noise points, not frequency hopping signals.

[0042] S33. During the clustering process, the start time ST, end time ET, start frequency SF, end frequency EF, and center frequency FC of the cluster are continuously updated by gradually increasing the number of points within the same cluster. Here, the start time ST is the minimum x-coordinate value (x_min) of all points in the cluster, the end time ET is the maximum x-coordinate value (x_max) of all points in the cluster, the start frequency SF is the minimum y-coordinate value (y_min) of all points in the cluster, and the end frequency EF is the maximum y-coordinate value (y_max) of all points in the cluster. Therefore, as a clustering process ends, the corresponding parameters of the frequency-hopping signal represented by that cluster can be estimated, as shown in the following formula: Hop_time = ET - ST Current frequency hopping signal bandwidth Band_sig = EF-SF Current frequency hopping signal center frequency FC = EF + Band_sig / 2 S34. After clustering a data block is completed, the information of the frequency hopping signals represented by all clusters in the previous block is compared one by one with the information of the frequency hopping signals represented by all clusters in the current data block to determine whether the two clusters at the connection point belong to the same hop signal, and the relevant parameter values ​​of the frequency hopping signal are updated accordingly. The specific comparison method is as follows: Each cluster uses its end time ET and center frequency FC as the horizontal and vertical coordinate values ​​to calculate the distance between the two clusters of the two data blocks. When the distance between the two clusters is less than eps, they are considered to belong to the same class and are the same frequency hopping signal, and the relevant parameter values ​​of the frequency hopping signal are updated accordingly.

[0043] S35. Collect all known parameters of frequency hopping signals to obtain the final frequency hopping signal parameter estimate.

[0044] Specifically, the dwell time of all frequency-hopping signals is averaged to obtain the final dwell time Hop_time; the signal bandwidth of all frequency-hopping signals is averaged to obtain the signal bandwidth Band_sig; the set of center frequencies of all frequency-hopping signals is the frequency-hopping frequency set; the maximum value of the end frequency of all frequency-hopping signals minus the minimum value of the start frequency of all frequency-hopping signals is the frequency-hopping bandwidth; after sorting all center frequencies in the frequency-hopping frequency set, the average of the differences between each pair of frequency points is the frequency-hopping frequency interval.

[0045] Therefore, based on all the steps described above, frequency-hopping signal parameter estimation based on spectrogram transformation and DBSCAN clustering can be achieved, yielding a relatively accurate set of frequency hopping frequencies, dwell time (hopping speed), frequency hopping bandwidth, frequency hopping interval, and frequency hopping signal bandwidth. Spectrogram transformation provides good time-frequency focusing and eliminates cross-term interference. Simultaneously, the block-based DBSCAN clustering algorithm achieves excellent clustering results for linear time-frequency graphs of frequency-hopping signals with low processing time, achieving a balance between time and performance, and resulting in good parameter estimation performance.

[0046] The following explanation uses a selected time period in an exemplary real-world application scenario to illustrate the estimation of frequency-hopping signal parameters. The time precision of the data is in seconds, and the frequency precision is in Hz. The test signal IQ data is a frequency-hopping signal of approximately 21ms length generated under the following parameters: sampling rate Fs = 40MHz, code rate Rb = 20kHz, hopping rate HopRate = 250hop / s, carrier Fc = 10MHz, hopping frequency interval F_hop = 200kHz, and a total of 32 frequency points. The frequency point information is as follows: {6.8MHz,7.0MHz,7.2MHz,7.4MHz,7.6MHz,7.8MHz,8.0MHz,8.2MHz,8.4MHz,8.6MHz,8.8MHz,9.0MHz,9.2MHz,9.4MHz,9.6MHz,9.8MHz,10.0MHz, 10.2MHz, 10.4MHz, 10.6MHz, 10.8MHz, 11.0MHz, 11.2MHz, 11.4MHz, 11.6MHz, 11.8MHz, 12.0MHz, 12.2MHz, 12.4MHz, 12.6MHz, 12.8MHz, 13.0MHz}.

[0047] Step S1: The IQ data length Len = 8360000. Perform a spectroscopic transform (SP) on the previously generated test signal IQ data. The selected window size (step size) for the Short Time Fourier Transform (STFT) is N1 = 8192, the overlap length of the two windows is Overlap = 4096, and the number of points in the Fourier transform is Nfft = 8192. Therefore, the frequency resolution DF = Fs / Nfft ≈ 4882.8 Hz, and the time resolution is DT = floor(Len / N1) × N1 / Fs / (2 × floor(Len / N1)) = 1.024e-4s. The size of the time-frequency matrix is: mf = 8192, nt = 2040. The time-frequency plot of the time-frequency matrix obtained after spectroscopic transforming the IQ data is shown below. Figure 2 As shown.

[0048] Step S2: Dynamic threshold calculation and time-frequency matrix binarization: S21: Based on the test data's hop rate HopRate = 250 hop / s, the dwell time of one hop is DwellTime = 0.004 s. Therefore, the number of points in each hop is Points = 0.004 / DT = 39.0625. We can calculate r = 2^(floor(log2(39.0625))) = 32. Therefore, the time-frequency matrix is ​​divided into m equal 32×32 sub-matrices, m = mf / 32×nt / 32 = 16128.

[0049] S22. Calculate the average value of each small matrix. ,in ; S23. For the average value of each small matrix, calculate the statistical time-frequency matrix. middle or The number of, if Then it is recorded as ,like Then it is recorded as ; S24, Calculate each Find the difference between the two, and find its maximum value. Then, find the value of the difference when it reaches its maximum value. The threshold T is used as the threshold value. Specifically, T = 212.915; S25. After binarization of the time-frequency matrix, the resulting matrix is ​​a 0,1 matrix, denoted as... The size of matrix B remains 8192×2040. The binarized time-frequency matrix after binarization using dynamic thresholding is shown in the figure below. Figure 3 As shown.

[0050] Step S3: Frequency hopping signal clustering and parameter estimation: S31. Determine the number of sub-data blocks (partnum) to which the binarized time-frequency matrix B needs to be divided. Since each sub-data block must contain at least one complete hop, the number of points (partpoints) within each part must be greater than the number of points (39.0625) in each hop. Also, both partnum and partpoints must be integers, so let partpoints = 40 and partnum = 51. Iterate through each sub-data block sequentially, calculating the number of points within each sub-data block. The points (Formula 5) form a point set: sigpoint, which contains x-values ​​and y-values. The x-values ​​are time information, and the y-values ​​are frequency information.

[0051] S32. Perform DBSCAN clustering on all points within the sigpoint set in each sub-data block. Since one cluster ultimately represents one hop, the duration must contain at least 40 points on the x-axis, i.e., 40 × DT = 4.096e-3s. On the y-axis, to distinguish between two hops, there must be at least one hop frequency point interval (200 kHz) between the two clusters. For more accurate detection, this value is preferably set to 20 kHz, i.e., F_hop = 20 kHz. Normalize the duration 4.096e-3s and the hop frequency point interval F_hop to the same order of magnitude, resulting in values ​​of t_normal = 4.096 and f_normal = 2, respectively. Therefore, set the clustering parameter neighborhood radius eps to 5 to ensure sufficient spacing in both dimensions. Set the minnb value to 10, chosen empirically. When the number of points within eps is less than minnb, the cluster is considered noise rather than a frequency hopping signal.

[0052] S33. During the clustering process, the start time ST, end time ET, start frequency SF, end frequency EF, and center frequency FC of the cluster are continuously updated by gradually increasing the number of points within the same cluster. Here, the start time ST is the minimum x-coordinate value (x_min) of all points in the cluster, the end time ET is the maximum x-coordinate value (x_max) of all points in the cluster, the start frequency SF is the minimum y-coordinate value (y_min) of all points in the cluster, and the end frequency EF is the maximum y-coordinate value (y_max) of all points in the cluster. As a clustering process ends, the corresponding parameters of the frequency-hopping signal represented by that cluster can be obtained, calculated as follows: Hop_time = ET - ST Current frequency hopping signal bandwidth Band_sig = EF-SF Current frequency hopping signal center frequency FC = EF + Band_sig / 2 S34. After clustering a data block, the information of the frequency hopping signals represented by all clusters in the previous block is compared one by one with the information of the frequency hopping signals represented by all clusters in the current data block to determine whether the two clusters at the connection point belong to the same hop signal, and the relevant parameter values ​​of the frequency hopping signal are updated accordingly. The specific comparison method is as follows: Each cluster uses its end time ET and center frequency FC as the horizontal and vertical coordinates to calculate the distance between the two clusters in the two data blocks. When the distance between the two clusters is less than eps, they are considered to belong to the same cluster and are the same frequency hopping signal, and the relevant parameter values ​​of the frequency hopping signal are updated accordingly. Finally, the frequency hopping signals accurately detected by using the DBSCAN clustering algorithm to perform block clustering on the binarized time-frequency matrix are as follows: Figure 4 As shown.

[0053] S35. Collect all known parameters of frequency hopping signals to obtain the final frequency hopping signal parameter estimate.

[0054] 1) Averaging the dwell times of all frequency-hopping signals, the final dwell time of the frequency-hopping signal is obtained as Hop_time = 3.899 e-3s; 2) The average signal bandwidth of all frequency hopping signals is calculated to obtain the signal bandwidth of the frequency hopping signal Band_sig = 77.4676 e3Hz; 3) The set of all center frequencies of frequency hopping signals is the frequency hopping frequency set, which contains 32 frequencies, as detailed below: {6.8018MHz,7.0044MHz,7.1997MHz,7.3987MHz,7.605MHz,7.8052MHz,8.0054MHz,8.2056M Hz,8.4021MHz,8.6047MHz,8.8037MHz,9.0063MHz,9.2017MHz,9.4043MHz,9.6008MHz,9.80 47MHz,9.9963MHz,10.2051MHz,10.4053MHz,10.6055MHz,10.8057MHz,11.0046MHz,11.201 2MHz,11.4063MHz,11.604MHz,11.8042MHz,12.0044MHz,12.1997MHz,12.4072MHz,12.6038 MHz,12.8101MHz,12.9993MHz}; 4) After sorting all the center frequencies of the frequency hopping frequency set, the average of the differences between each pair of frequency points is the frequency hopping interval of 198.698KHz.

[0055] 5) The maximum value of the end frequency of all frequency hopping signals minus the minimum value of the start frequency of all frequency hopping signals, plus the frequency hopping interval, gives the frequency hopping bandwidth of 6.402MHz. 6) The minimum value of all frequency hopping points plus half of the frequency hopping bandwidth is the carrier frequency of the current frequency hopping signal, which is 10.002MHz.

[0056] The present invention further provides a frequency hopping signal parameter measurement system based on time-frequency analysis and block clustering, including a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to execute a frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering.

[0057] The system of the present invention corresponds to the method described above and has the same advantages as the method described above.

[0058] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. Computer-readable media include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0059] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for estimating parameters of frequency-hopping signals based on time-frequency analysis and block clustering, characterized in that, include: Step S1: Read the frequency hopping signal IQ data and perform time-frequency analysis on the frequency hopping signal IQ data to obtain the corresponding time-frequency matrix; Step S2: The time-frequency matrix is ​​binarized according to the dynamic threshold to distinguish the location of noise and the location of the spectrum to obtain a binarized time-frequency matrix. The location of the frequency hopping signal is initially detected by the binarized time-frequency matrix. The dynamic threshold is obtained by dividing the time-frequency matrix into multiple sub-matrices and determining the distribution trend of the values ​​of each sub-matrix. Step S3: Divide the binarized time-frequency matrix into blocks, cluster the frequency hopping signals detected in each block, and continuously update the start and end times and frequency parameters of each cluster by gradually increasing the number of points in the same cluster. Calculate the frequency hopping signal parameters based on the clustering results, and compare the frequency hopping signal information obtained from the current block clustering with the frequency hopping signal information obtained from the previous block clustering after each block clustering is completed. Determine whether they belong to the same hop data and update the frequency hopping signal parameters accordingly. Obtain the final frequency hopping signal parameter estimation result based on all frequency hopping signal parameters. Step S2 includes: The time-frequency matrix is ​​divided into multiple sub-time-frequency matrices according to the hopping rate range and time resolution of the frequency hopping signal; The number of elements in the time-frequency matrix that are greater than and less than the average value of each sub-time-frequency matrix is ​​counted to obtain multiple sets of corresponding first and second counts; Calculate the difference between the first and second quantities in each group, and take the maximum value among the differences in each group as the dynamic threshold; Elements in the time-frequency matrix that are greater than the dynamic threshold are marked as first values, and elements that are less than the dynamic threshold are marked as second values, to obtain a binarized time-frequency matrix, wherein the first value indicates the presence of a frequency hopping signal, and the second value indicates the absence of a frequency hopping signal; In step S3, the frequency hopping signal parameters include one or more of the following: dwell time, signal bandwidth, frequency hopping frequency set, frequency hopping bandwidth, and frequency hopping frequency point interval. Specifically, the dwell time of the current clustered frequency hopping signal is calculated based on the clustering end time and clustering start time; the signal bandwidth of the current clustered frequency hopping signal is calculated based on the clustering end frequency point and clustering start frequency point; the set of all center frequencies of the frequency hopping signals is used as the frequency hopping frequency set; the maximum value of the end frequency points of all frequency hopping signals minus the minimum value of the start frequency points of all frequency hopping signals is used as the frequency hopping bandwidth; after sorting all center frequencies in the frequency hopping frequency set, the difference between each pair of frequency points is calculated, and then the average of the differences is taken to obtain the frequency hopping frequency point interval.

2. The frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering according to claim 1, characterized in that, In step S1, the frequency hopping signal IQ data is analyzed using spectral transform to obtain the corresponding time-frequency matrix. The number of rows in the time-frequency matrix is ​​the window length used when performing short-time Fourier transform, and the number of columns is determined by the length of the IQ data and the window length used when performing short-time Fourier transform.

3. The frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering according to claim 2, characterized in that, The expression for calculating the time-frequency matrix is ​​as follows: In the above formula, t represents the current time point of the analysis, and f represents the frequency component of the current analysis. Let be the integral variable obtained by summing over all time signals. This is a short-time Fourier transform; The expression for calculating the number of columns in the time-frequency matrix is ​​as follows: In the above formula, floor is the floor function. The length of the IQ data, This refers to the window length used when performing a short-time Fourier transform on the IQ data.

4. The frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering according to claim 1, characterized in that, The step of dividing the time-frequency matrix into multiple sub-time-frequency matrices based on the hopping speed range and time resolution of the frequency-hopping signal specifically includes: The minimum number of data points per hop of the frequency hopping signal is calculated based on the hopping speed range and time resolution of the frequency hopping signal. The number of sub-time-frequency matrix divisions is determined based on the minimum values ​​of the number of rows, columns, and points of the time-frequency matrix. The number of rows and columns of each sub-time-frequency matrix is ​​determined based on the minimum value of the points.

5. The frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering according to claim 4, characterized in that, The expression for calculating the minimum number of points is: In the above formula, DT represents the maximum hopping rate range of the frequency hopping signal, and DT represents the time resolution. The length of the IQ data, This refers to the window length used when performing a short-time Fourier transform on the IQ data. is the sampling rate of the frequency hopping signal, and floor is the floor function; The formulas for calculating the number of rows and columns of each sub-time-frequency matrix are as follows: In the above formula, Points_min is the minimum number of points.

6. The frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering according to any one of claims 1 to 5, characterized in that, In step S3, the block-based processing of the binarized time-frequency matrix specifically includes: The minimum number of data points per hop of the frequency hopping signal is calculated based on the hopping speed range and time resolution of the frequency hopping signal. The number of blocks is determined based on the maximum number of columns and points in the time-frequency matrix, so that each block contains at least one hop of the frequency-hopping signal data.

7. The frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering according to claim 1, characterized in that, Step S3 includes: The binarized time-frequency matrix B is divided into several blocks, each block containing at least one complete hop of the frequency-hopping signal. The points containing the frequency-hopping signal in each block are recorded to form a point set. Set the neighborhood radius and minimum number of points for clustering, and cluster all frequency hopping signals within the point set to obtain multiple cluster categories. During the clustering process, the start and end times and frequency parameters of each cluster are continuously updated by gradually increasing the number of points in the same category. The frequency parameters include the start frequency SF, the end frequency EF, and the center frequency FC. For each cluster category, the corresponding frequency hopping signal parameters are calculated based on the start and end times and frequency parameters of each cluster. After the current block clustering is completed, the information of the frequency hopping signals represented by all clusters in the previous block is compared with the information of the frequency hopping signals represented by all clusters in the current data block to determine whether the two clusters at the connection point belong to the same hop data. If the distance between the two clusters is less than the set neighborhood radius eps, they are determined to belong to the same class and correspond to the same frequency hopping signal, and the corresponding frequency hopping signal parameters are updated.

8. A frequency-hopping signal parameter measurement system based on time-frequency analysis and block clustering, comprising a microprocessor and a memory interconnected, characterized in that, The microprocessor is programmed or configured to execute the frequency hopping signal parameter estimation method based on time-frequency analysis and block clustering as described in any one of claims 1 to 7.

Images