A frequency hopping signal detection method and device based on dynamic clustering

Abstract

This invention discloses a method and apparatus for detecting frequency-hopping signals based on dynamic clustering. The method includes the following steps: continuously acquiring data of the signal under test, acquiring a segment of data each time to obtain the current segment data; transforming the current segment data to obtain time-frequency data; detecting frequency-hopping signals in the time-frequency data of the current segment data to obtain multiple frequency-hopping signal data points; clustering the currently detected frequency-hopping signal data points to obtain the current segment data clustering result; calculating the required frequency-hopping signal parameters based on the information of each cluster in the current clustering result; and after each clustering, connecting the current segment data clustering result with the previous segment data clustering result to obtain the final clustering result. This invention can simultaneously measure the parameters of multiple frequency-hopping signals without prior information and has advantages such as simple implementation, high detection efficiency and accuracy, and strong noise resistance.

Description

Technical Field

[0001] This invention relates to the field of frequency hopping signal detection technology, and in particular to a method and apparatus for frequency hopping signal detection based on dynamic clustering. Background Technology

[0002] Frequency-hopping communication is widely used in shortwave and VHF radios and data link technologies, as well as in Bluetooth, GSM, and satellite communications. It boasts advantages such as strong anti-interference capabilities, low spectrum utilization, low interception probability, flexible networking options, and slow channel attenuation. Communication can be categorized into cooperative and non-cooperative communication based on whether the communicating parties have mutual knowledge. Non-cooperative communication is an unauthorized access method where the receiving end needs to estimate and identify the signal without knowing the other party's signal parameters, and then perform data analysis and processing. Research on frequency-hopping communication in non-cooperative environments mainly includes three aspects: frequency-hopping signal detection, frequency-hopping signal parameter estimation, and frequency-hopping network station selection. Frequency-hopping signals possess rich characteristic parameters, including hopping period, hopping rate, hopping pattern, hopping bandwidth, hopping time, and frequency-hopping modulation method. Therefore, in non-cooperative environments, if these characteristic parameters of the frequency-hopping signal can be detected from the intercepted signal, de-hopping and demodulation processes can be used to obtain intelligence information, facilitating interference.

[0003] Frequency hopping signal parameter estimation involves detecting and calculating the characteristic data of the frequency hopping signal, such as... Figure 1 As shown, the main parameters include center frequency, start time, end time, dwell time, signal power, and instantaneous bandwidth. Frequency-hopping signal parameter estimation is a prerequisite for frequency-hopping network station selection. Frequency-hopping network station selection refers to the ability to separate all or specific station-specific hopping signals from the mixed signals of multiple frequency-hopping stations, which is the most important prerequisite for effective detection. However, with the rapid increase in the number of wireless devices in the current environment, the electromagnetic environment is becoming increasingly complex. Different communication signals and various noise signals coexist in space, making it extremely difficult to quickly detect the target frequency-hopping signal from various interference signals. Furthermore, in complex electromagnetic environments, for multi-component frequency-hopping signals mixed from multiple networks, the frequency-hopping sequence of each signal is often unknown. These signals are superimposed, making subsequent processing quite difficult.

[0004] For parameter estimation in frequency-hopping signal detection, existing technologies typically employ two main approaches: autocorrelation-based and time-frequency analysis-based methods, to estimate parameters such as hopping period, hopping rate, hopping time, and power. Autocorrelation-based estimation algorithms estimate the hopping rate by sampling a sufficient number of hops and performing autocorrelation on the signal sequence. However, this algorithm is not fully blind and requires certain prior information regarding the hopping rate and power. Furthermore, autocorrelation-based estimation algorithms are usually based on a single frequency-hopping signal; when multiple frequency-hopping signals occur within the observation period, the autocorrelation method becomes inapplicable. Time-frequency analysis-based methods, based on the Short-Time Fourier Transform (STFT), estimate the parameters of non-stationary signals, including frequency-hopping signals. This allows for efficient estimation without changing the hardware design or requiring any prior information. However, traditional time-frequency analysis methods suffer from long analysis times, slow convergence of clustering algorithms used in blind time-hopping, and poor overall performance, indicating significant room for improvement in their performance.

[0005] To address the above issues, some practitioners have proposed applying image processing methods to time-frequency images to extract parameters of frequency-hopping signals. For example, Chinese patent application 202210675321.6 discloses a real-time blind detection method for frequency-hopping signals. This method involves performing dual-threshold thresholding on the time-frequency image of the frequency-hopping signal, and using thresholding to extract parameters from the time-frequency image. Figure Two Value-based methods employ connected component labeling to obtain the state information of all connected components, resulting in features of the connected component image. Coarse clustering is then performed on the connected components based on their length features, yielding coarse clustering results with different hopping rates. However, this type of method has the following problems:

[0006] 1. By performing dual-threshold thresholding on the time-frequency image and employing a thresholding method to segment the time-frequency image... Figure Two Value-added processing can introduce unremovable interference into the uneven edges of the frequency-hopping signal and the power fluctuations of different segments of the same line. This is especially problematic when there are signals with low signal-to-noise ratios, as it can severely affect the extraction quality of the frequency-hopping signal.

[0007] 2. Use a threshold segmentation method to perform time-frequency analysis on the time-frequency graph. Figure Two Binarization: When two frequency hopping signals are close in frequency, binarizing the time-frequency diagram may connect the two frequency hopping signals into a single frequency hopping signal, which may also cause errors in the calculation of frequency hopping parameters.

[0008] 3. Since the frequency hopping clustering is achieved by coarsely clustering the connected components based on the length characteristics of the connected component image, it relies on a set threshold to distinguish and cluster frequency hopping signals. This is difficult to adapt to the complex electromagnetic environment composed of multi-component frequency hopping signals from multiple networks. Furthermore, it can only achieve static clustering analysis of a segment of data and cannot achieve correlation between preceding and subsequent data, resulting in low detection accuracy. Summary of the Invention

[0009] The technical problem to be solved by the present invention is: in view of the technical problems existing in the prior art, the present invention provides a method and device for detecting frequency hopping signals based on dynamic clustering that is simple to implement, has high detection efficiency and accuracy and strong noise resistance, and can simultaneously achieve accurate measurement of multiple frequency hopping signal parameters without prior information.

[0010] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0011] A frequency hopping signal detection method based on dynamic clustering, comprising the following steps:

[0012] Continuously acquire data from the signal under test, acquiring a segment of data each time to obtain the current segment of data;

[0013] Transform the current segment of data to obtain the time-frequency data of the current segment;

[0014] Frequency hopping signal detection is performed on the time-frequency data of the current segment to obtain multiple frequency hopping signal data points;

[0015] Cluster the currently detected frequency hopping signal data points to obtain the frequency hopping signal clustering result for the current segment of data;

[0016] Based on the information of each class in the current final clustering results, the required frequency hopping signal parameters are calculated and output.

[0017] After each clustering is completed, the clustering results of the current segment of data frequency hopping signals are connected with the clustering results of the previous segment of data frequency hopping signals to classify the same frequency hopping signal data points into the same class and obtain the final clustering result.

[0018] Furthermore, the frequency hopping signal detection of the time-frequency data of the current segment includes: performing frequency hopping signal detection on each snapshot of the time-frequency data of the current segment to obtain each frequency hopping signal data point. The frequency hopping signal data point includes the time when the frequency hopping signal occurs and the corresponding center frequency point position. A time-frequency plot of the frequency hopping signal is obtained by plotting each frequency hopping signal data point.

[0019] Furthermore, the DBSCAN clustering algorithm is used to cluster each of the frequency hopping signal data points, and different weight values ​​are configured for the horizontal and vertical axes when calculating the distance between two data points in the DBSCAN clustering algorithm.

[0020] Furthermore, the DBSCAN clustering algorithm uses the following formula to calculate the Euclidean distance between two data points:

[0021]

[0022] Where α and β represent the weight coefficients of the horizontal and vertical axes, respectively, with α less than 1 and β greater than 1. i ,y i ), (x j ,y j ) represent the x and y coordinates of two different data points.

[0023] Furthermore, the frequency hopping signal parameters include any one or more of the following: the start time, end time, hopping period, hopping time, center frequency, signal bandwidth, and signal power. The frequency hopping signal parameters are calculated based on each time point of the frequency hopping signal point set corresponding to the clustering category. The start time is obtained from the minimum time value in the frequency hopping signal point set; the end time is obtained from the maximum time value in the frequency hopping signal point set; the hopping period is obtained from the difference between the minimum and maximum time values; the hopping time is calculated from the start time and the hopping period; and the center frequency, signal bandwidth, and signal power are obtained by calculating the statistical values ​​of the center frequency, signal bandwidth, and signal power of each data point in the frequency hopping signal point set.

[0024] Furthermore, the step of connecting the current segment's data frequency hopping signal clustering result with the previous segment's data frequency hopping signal clustering result includes:

[0025] Find the first frequency hopping signal category that meets the backward concatenation clustering condition from the previous data frequency hopping signal clustering results, and find the second frequency hopping signal category that meets the forward concatenation clustering condition from the current data frequency hopping signal clustering results;

[0026] If the distance between the first frequency hopping signal category and the second frequency hopping signal category is less than a preset threshold, the first frequency hopping signal and the second frequency hopping signal category are connected into the same category;

[0027] Once all frequency hopping signals that meet the connection conditions have been clustered, the final frequency hopping signal clustering result is obtained.

[0028] Furthermore, the step of finding the first frequency hopping signal category that satisfies the subsequent clustering conditions from the previous segment of frequency hopping signal clustering results includes:

[0029] Iterate through the frequency hopping signals in the previous data frequency hopping signal clustering results. If a first frequency hopping signal category HOP exists... i HOP that meets the conditions i .hop end ≥(sec_curr-sec_thr), where sec_curr represents the start time of the current time-frequency data segment, sec_thr represents the time threshold, and HOP i .hop end HOP i The first frequency hopping signal category (HOP) is determined by the end time of the hop. i The conditions for backward-joining clustering are met.

[0030] Furthermore, the step of searching for a second frequency hopping signal category that satisfies the forward-connection clustering condition in the current segment data frequency hopping signal clustering results includes:

[0031] Iterate through the frequency hopping signals in the current data segment's frequency hopping signal clustering results. If a second frequency hopping signal category HOP exists... k HOP that meets the conditions k .hop start ≤(sec_curr+sec_thr), HOP k .hop start HOP i The start time of the category determines the second frequency hopping signal category HOP. k It meets the conditions for forward-connected clustering.

[0032] A frequency hopping signal detection device based on dynamic clustering, comprising:

[0033] The data acquisition module is used to continuously acquire data from the signal under test, acquiring a segment of data each time to obtain the current segment of data;

[0034] The time-frequency data acquisition module is used to transform the current segment data and obtain the time-frequency data of the current segment data;

[0035] The frequency hopping signal detection module is used to detect frequency hopping signals in the time-frequency data of the current segment of data and obtain multiple frequency hopping signal data points.

[0036] The dynamic clustering module is used to cluster the currently detected frequency hopping signal data points to obtain the frequency hopping signal clustering result of the current segment of data.

[0037] The frequency hopping parameter calculation module is used to calculate the required frequency hopping signal parameters based on the information of each class in the current clustering results.

[0038] After each clustering is completed, the dynamic clustering module also includes connecting the clustering results of the current segment of data frequency hopping signals with the clustering results of the previous segment of data frequency hopping signals, so as to classify the same frequency hopping signal data points into the same class and obtain the final clustering result.

[0039] A frequency hopping signal detection device based on dynamic clustering includes a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to perform the method described above.

[0040] Compared with the prior art, the advantages of the present invention are as follows: The present invention continuously collects the signal to be detected for time-frequency data extraction, and then detects the frequency hopping signal from the time-frequency data. Then, the detected frequency hopping signal data points are clustered to classify the frequency hopping signals within the same segment into the same category. At the same time, dynamic clustering is achieved by connecting the clustering results between the data segments before and after. Based on the dynamic clustering method, the accuracy of frequency hopping signal parameter extraction and anti-noise performance can be effectively improved. Thus, it can be applied to the detection of frequency hopping signals in complex environments where multiple frequency hopping networks are working simultaneously, and achieve accurate and rapid detection of multiple frequency hopping signals in complex environments. Attached Figure Description

[0041] Figure 1 This is a schematic diagram illustrating the principle of frequency hopping signal parameters.

[0042] Figure 2 This is a schematic diagram illustrating the implementation process of the frequency hopping signal detection method based on dynamic clustering in this embodiment.

[0043] Figure 3 This is a schematic diagram of the time-frequency data of the current segment and the initial segment obtained in a specific application embodiment of the present invention.

[0044] Figure 4 This is a schematic diagram of a snapshot signal at a certain moment obtained in a specific application embodiment of the present invention.

[0045] Figure 5 This is a time-frequency plot of the initial segment and the current segment signals obtained in a specific application embodiment of the present invention.

[0046] Figure 6 This is a schematic diagram of the frequency hopping signal clustering results of the initial segment and the current segment signals obtained in a specific application embodiment of the present invention.

[0047] Figure 7 This is a schematic diagram of the new clustering result formed by clustering connections in a specific application embodiment of the present invention.

[0048] Figure 8 This is a schematic diagram illustrating the specific process of dynamic clustering in frequency hopping signal detection based on dynamic clustering in this embodiment. Detailed Implementation

[0049] The present invention will be further described below with reference to the accompanying drawings and specific preferred embodiments, but this does not limit the scope of protection of the present invention.

[0050] like Figure 2 As shown, the steps of the frequency hopping signal detection method based on dynamic clustering in this embodiment include:

[0051] Step S01. Data Acquisition: Continuously acquire data from the signal under test, acquiring a segment of data each time to obtain the current segment of data.

[0052] In this embodiment, data is continuously collected from the signal under test. Each time, a segment of data of a specified length is collected. After a segment of data is collected, the process proceeds to step S02 to perform subsequent processing.

[0053] Step S02. Time-frequency data acquisition: Transform the current segment data to obtain the time-frequency data of the current segment data.

[0054] This embodiment specifically performs a short-time Fourier transform on the acquired current segment data to obtain the time-frequency data of the signal. In a specific application embodiment, such as... Figure 3 The figure shows the time-frequency diagrams of 10 frequency-hopping signals with different parameters distributed within a 180MHz bandwidth. Figure 3 (a) corresponds to the time-frequency diagram of the initial segment. Figure 3 (b) corresponds to the time-frequency diagram of the current segment.

[0055] Step S03. Frequency hopping signal detection: Detect frequency hopping signals on the time-frequency data of the current segment to obtain multiple frequency hopping signal data points.

[0056] In this embodiment, frequency hopping signal detection is performed on each snapshot of the time-frequency data of the current segment to obtain each frequency hopping signal data point. The frequency hopping signal data point includes the time when the frequency hopping signal occurs and the corresponding center frequency point position. The time-frequency plot of the frequency hopping signal is obtained by plotting each frequency hopping signal data point.

[0057] In a specific application embodiment, a snapshot image obtained by signal detection of a snapshot of time-frequency data is shown below. Figure 4 As shown, the horizontal axis represents frequency, and the vertical axis represents signal amplitude. Using a dual-threshold signal detection method, the signal parameters at this moment can be obtained, including the signal's duration (sec), center frequency (fc), signal bandwidth (bw), and signal power (pw), etc., which can be represented by a set po. i{sec,fc,bw,pw} represents the data point set, where pw{} represents the data point index. Using sec and fc as point values, with the horizontal axis representing time (sec) and the vertical axis representing frequency (fc), a time-frequency plot is created. Each point represents the position of the signal in the frequency spectrum at that moment, resulting in the time-frequency plot shown below. Figure 5 As shown, where Figure 5 (a) corresponds to the initial segment time-frequency plot. Figure 5 (b) corresponds to the time-frequency plot of the current segment.

[0058] Step S04. Dynamic clustering: Cluster the currently detected frequency hopping signal data points to obtain the frequency hopping signal clustering result for the current segment of data.

[0059] From such Figure 5 The time-frequency plot shown demonstrates that the time and frequency points of the frequency-hopping signal are accurately identified, but some noise causing false alarms is also introduced. To remove this false alarm noise and cluster similar frequency-hopping signals into a single group, this embodiment employs the DBSCAN clustering algorithm to cluster the time-frequency plot data. This process removes false alarm noise and groups similar frequency-hopping signals together for accurate extraction of frequency-hopping parameters. The frequency-hopping clustering results obtained in this specific application embodiment are as follows: Figure 6 As shown, where Figure 6 (a) corresponds to the initial segment frequency hopping clustering result. Figure 6 (b) corresponds to the frequency hopping clustering result of the current segment.

[0060] The DBSCAN clustering algorithm can cluster dense datasets of arbitrary shapes without prior knowledge of the number of clusters. This embodiment utilizes DBSCAN to cluster various frequency-hopping signal data points, accurately grouping the same frequency-hopping signal segment into one class. This effectively achieves a fully blind detection process for frequency-hopping signals. Furthermore, it can detect outliers during clustering, is insensitive to outliers in the dataset, and effectively removes false alarm noise. The clustering results are unbiased, and the selection of initial values ​​for the clustering algorithm does not affect the results, ensuring accurate calculation of frequency-hopping signal parameters. Figure 6 As can be seen, using the DBSCAN algorithm to cluster the same frequency hopping signal can effectively eliminate false alarm noise and accurately cluster the signal points of the same frequency hopping signal.

[0061] The DBSCAN clustering algorithm requires calculating the Euclidean distance between each point to determine whether they can be grouped into a single cluster. This embodiment leverages the characteristics of frequency-hopping signals, utilizing the continuity of the same frequency-hopping signal on the horizontal axis (time axis) and the discreteness of different frequency-hopping signals on the vertical axis (frequency axis). The DBSCAN algorithm optimizes the calculation method by using different weighted distance values ​​on the horizontal and vertical axes, effectively separating two frequency-hopping signals that are close in frequency. This allows for the effective separation of individual frequency-hopping signals, facilitating the subsequent accurate extraction of frequency-hopping signal parameters.

[0062] The traditional DBSCAN clustering algorithm uses the following formula to calculate the Euclidean distance:

[0063]

[0064] Where (x) i ,y i ), (x j ,y j () indicates two different points.

[0065] In this specific embodiment, when calculating the Euclidean distance between two data points in the DBSCAN clustering algorithm, different weight values ​​are assigned to the horizontal and vertical axes, and the calculation expression is as follows:

[0066]

[0067] Where α and β represent the weight coefficients of the horizontal and vertical axes, respectively, with α less than 1 and β greater than 1. i ,y i ), (x j ,y j ) represent the x and y coordinates of two different data points.

[0068] This embodiment combines the characteristics of frequency-hopping signals, reducing the α value to less than 1. This allows the points of the same frequency-hopping signal at different times to be brought closer together, thus achieving better classification. The β value is then increased to greater than 1, which can widen the distance between frequency-hopping signals at different frequencies to avoid clustering them into one class. This effectively avoids aliasing of frequency-hopping signals and the resulting incorrect clustering results.

[0069] Step S05. Frequency hopping parameter calculation: Based on the information of each class in the final clustering result obtained, calculate the required frequency hopping signal parameters for output.

[0070] After completing the frequency hopping signal clustering in step S04, it is further necessary to calculate and extract the frequency hopping signal parameters. These parameters include the start time, end time, hopping period, hopping duration, center frequency, signal bandwidth, and signal power. Specifically, the frequency hopping signal parameters can be obtained from each time point of the corresponding category of frequency hopping signal points obtained through clustering. The start time is obtained from the minimum time value in the frequency hopping signal point set; the end time is obtained from the maximum time value in the set; the hopping period is obtained from the difference between the minimum and maximum time values; the hopping duration is calculated from the start time and hopping period; and the center frequency, signal bandwidth, and signal power are obtained by calculating the statistical values ​​of the center frequency, signal bandwidth, and signal power of each data point in the frequency hopping signal point set.

[0071] Specifically, the set of points for a frequency hopping signal cluster is represented by C. i (...) N Let i represent the category number, and N represent the number of points contained in this category. Then C i (...) N It can be represented as:

[0072] C i (po n {sec,fc,bw,pw},...,po m {sec,fc,bw,pw},...) N (3)

[0073] Where n and m represent the sequence numbers of the different points in the corresponding categories.

[0074] Find C i (...) N The maximum value of time in seconds is po.sec max and minimum value po.sec min The start time of the frequency hopping signal can be obtained through calculation. start , Hop at the end of the jump end Hop cycle T Calculation of hop time time The specific calculation formula is as follows:

[0075]

[0076] This embodiment specifically calculates C. i (...) N By taking the average center frequency fc, signal bandwidth bw, and signal power pw of all points, the center frequency, signal bandwidth, and signal power corresponding to the frequency hopping signal are obtained, thus determining more accurate parameters. The specific calculation expression is as follows:

[0077]

[0078] Where k = 1, ..., N.

[0079] It is understandable that the above parameters can be calculated using other methods to meet different needs.

[0080] The final parameters of the current frequency hopping signal are:

[0081] HOP i ={hop start ,hop end ,hop T ,hop time ,hop fc ,hop bw ,hop pw} (6)

[0082] Complete the parameter detection and extraction of the frequency hopping signal.

[0083] Step S06. Clustering Connection: During each clustering, the clustering result of the current segment of data frequency hopping signal is connected with the clustering result of the previous segment of data frequency hopping signal to classify the same frequency hopping signal data points into the same category, thus obtaining the final clustering result of the current segment of data frequency hopping signal.

[0084] After completing the clustering and parameter extraction of a segment of time-frequency data, in order to perform dynamic clustering and parameter extraction on frequency-hopping signals, it is necessary to connect the clustering results of the two segments. Cluster connection means connecting the categories of the two segments of frequency-hopping signals that are in different segments and merging them into one category to remove redundant information. That is, connecting the data of the same category of frequency-hopping signals to update the clustering results and obtain new frequency-hopping information.

[0085] Considering the current segment's frequency hopping signal time-frequency diagram, if the frequency hopping signal to be connected is to be a signal whose hopping ends in the previous segment's time-frequency clustering result, and whose hopping starts are very close to those in the current segment's time-frequency clustering result, this embodiment utilizes these characteristics to connect the clustering results of the preceding and following segments of data frequency hopping signals. The specific steps for connecting the current segment's data frequency hopping signal clustering result with the previous segment's data frequency hopping signal clustering result in this embodiment include:

[0086] S601. Find the first frequency hopping signal category that meets the backward concatenation clustering condition from the previous data frequency hopping signal clustering results, and find the second frequency hopping signal category that meets the forward concatenation clustering condition from the current data frequency hopping signal clustering results;

[0087] S602. If the distance between the first frequency hopping signal category and the second frequency hopping signal category is less than a preset threshold, the first frequency hopping signal category and the second frequency hopping signal category are connected into the same category;

[0088] S603. After all frequency hopping signals that meet the connection conditions are clustered, the final frequency hopping signal clustering result is obtained.

[0089] In this embodiment, step S601, which involves finding the first frequency hopping signal category that satisfies the subsequent clustering condition from the previous data frequency hopping signal clustering results, specifically includes: traversing the frequency hopping signals in the previous data frequency hopping signal clustering results; if a first frequency hopping signal category HOP exists... i HOP that meets the conditions i .hop end ≥(sec_curr-sec_thr), where sec_curr represents the start time of the current time-frequency data segment, sec_thr represents the time threshold (specifically, sec_thr can be configured as the time value of 4 short-time Fourier transform steps), HOP i .hop end HOP i The first frequency hopping signal category (HOP) is determined by the end time of the hop. i The conditions for backward-joining clustering are met;

[0090] In this embodiment, step S602, which involves searching for a second frequency hopping signal category that satisfies the forward-connecting clustering condition in the current segment data frequency hopping signal clustering results, specifically includes: traversing the frequency hopping signals in the current segment data frequency hopping signal clustering results; if a second frequency hopping signal category HOP exists... k HOP that meets the conditions k .hop start ≤(sec_curr+sec_thr), HOP k .hop start HOP i The start time of the category determines the second frequency hopping signal category HOP. k It meets the conditions for forward-connected clustering.

[0091] This embodiment borrows from the DBSCAN clustering algorithm, which calculates Euclidean distance to determine neighboring points and then groups them into the same class. When calculating the distance between the first frequency hopping signal category and the second frequency hopping signal category, it utilizes the characteristic that the same frequency hopping signal is continuous on the horizontal axis (time axis) and different frequency hopping signals are discrete on the vertical axis (frequency axis). It optimizes the Euclidean distance calculation formula in a way similar to equation (2) for the classification of points of the same class. The specific calculation expression of the Euclidean distance is as follows:

[0092]

[0093] Similar to equation (2), the value of α is reduced to less than 1, which brings the same frequency-hopping signal closer together at different times, thus achieving better classification. The value of β is increased to greater than 1, which can expand the distance between frequency-hopping signals at different frequencies to avoid clustering into one class, effectively avoiding aliasing of frequency-hopping signals and obtaining incorrect clustering results.

[0094] Taking the above method for connecting the clustering results of frequency hopping signals between the preceding and following segments in a specific application embodiment as an example, the detailed implementation steps are as follows:

[0095] Assume the frequency hopping signal set extracted from the previous clustering results is HOP_H{...} N The frequency hopping signal class extracted from the current segment clustering results is HOP_L{...} M , where N and M represent the number of clusters for frequency hopping signals.

[0096]

[0097] First, design a time interval to identify the frequency hopping signal categories that satisfy the condition of the previous segment connecting backwards and the current segment connecting forwards:

[0098] Iterate through HOP_H{...,HOP i ,...} N The frequency hopping signal in the middle, if there is HOP i .hop end If ≥(sec_curr-sec_thr), then in HOP_H{...,HOP i ,...} N HOP i The conditions for backward-joining clustering are met.

[0099] Iterate through HOP_L{...,HOP k ,...} M The frequency hopping signal in the middle, if there is HOP k .hop start If ≤(sec_curr+sec_thr), then in HOP_L{...,HOP k ,...} M HOP k It meets the conditions for forward-connected clustering.

[0100] Calculate HOP_H{...,HOP according to equation (7) i ,...} N HOP_L{...,HOP k ,...} MThe Euclidean distance between interconnected points, i.e., HOP_H{...,HOP i ,...} N HOP i .hop end HOP i .hop fc ),HOP_L{...,HOP k ,...} M HOP k .hop start HOP k .hop fc ); if dist ik If ≤eps, it indicates that the frequency hopping signal in the previous paragraph belongs to the HOP category. i Category HOP of the current frequency hopping signal k They can be grouped into the same category, where eps is the neighborhood radius parameter set in the DBSCAN algorithm.

[0101] Update HOP_H{...,HOP i ,...} N The previous section of the frequency hopping signal HOP i Signal parameters:

[0102]

[0103] Clear HOP_L{...,HOP k ,...} M HOP in the current segment k Parameter data.

[0104] Once all frequency hopping signals that satisfy the connection criteria have been clustered, a new frequency hopping signal clustering result is obtained, which is:

[0105] HOP_New{...,HOP i ,...} N+M' =HOP_H{...,HOP i ,...} N &HOP_L{...,HOP k ,...} M' (10)

[0106] Where HOP_New={...,HOP i ,...} N+M' The new clustering result set is N+M', where N+M' is the number of new clustering frequency hopping signals, M' is the number of current segment clusters M minus the number of frequency hopping signals that meet the forward clustering conditions, and & represents the union.

[0107] After the above steps, a new frequency hopping signal cluster set HOP_New = {...,HOP} can be obtained. i ,...} N+M' This data is used to connect with the previous clustering data in the next frequency hopping clustering iteration, and so on. This allows for continuous dynamic clustering of frequency hopping signals, enabling accurate extraction of frequency hopping parameters, which are then provided to frequency hopping network stations for sorting. In a specific application example, the clustering connection results in a new clustering outcome, such as... Figure 7 As shown.

[0108] like Figure 8 As shown, this embodiment obtains time-frequency data by continuously performing short-time Fourier transform on the currently acquired data segment. Signal detection is performed on each snapshot of the time-frequency data. Based on the signal detection results, the time of signal occurrence and the corresponding center frequency point position are plotted to obtain a time-frequency plot of the signal. The DBSCAN clustering algorithm is used to cluster the signal points to obtain the current frequency hopping signal clustering result. After each clustering, the current clustering result is connected with the previous clustering result to group the points of the same frequency hopping signal into the same class, resulting in a new frequency hopping signal clustering result. This process is repeated until the clustering result of the next data segment is obtained, and so on, ultimately achieving dynamic clustering of the frequency hopping signal. Based on the clustering result of the frequency hopping signal obtained from each clustering, the information of each class can be calculated to obtain various accurate parameters of the frequency hopping signal, thereby completing the detection of the frequency hopping signal.

[0109] This invention acquires time-frequency data by continuously collecting the signal to be detected, performs frequency hopping signal detection on the time-frequency data, and then clusters the detected frequency hopping signal data points to classify frequency hopping signals within the same segment into the same category. At the same time, dynamic clustering is achieved by connecting the clustering results between the preceding and following segments. Based on the dynamic clustering method, the accuracy of frequency hopping signal parameter extraction and noise resistance can be effectively improved, thus making it applicable to the detection of frequency hopping signals in complex environments where multiple frequency hopping networks are operating simultaneously, achieving accurate and rapid detection of multiple frequency hopping signals in complex environments.

[0110] The frequency hopping signal detection device based on dynamic clustering in this embodiment includes:

[0111] The data acquisition module is used to acquire a segment of data from the signal under test in real time to obtain the current segment of data;

[0112] The time-frequency data acquisition module is used to transform the current segment data and obtain the time-frequency data of the current segment data;

[0113] The frequency hopping signal detection module is used to detect frequency hopping signals in the time-frequency data of the current segment of data and obtain multiple frequency hopping signal data points.

[0114] The dynamic clustering module is used to cluster the currently detected frequency hopping signal data points to obtain the current segment of data frequency hopping signal clustering result and connect it with the previous segment of data frequency hopping signal clustering result, so as to classify the same frequency hopping signal data points into the same category and obtain the final clustering result of the current segment of data frequency hopping signal.

[0115] The frequency hopping parameter calculation module is used to obtain the required frequency hopping signal parameters output based on the information of each class in the final clustering result.

[0116] The frequency hopping signal detection device based on dynamic clustering in this embodiment corresponds one-to-one with the frequency hopping signal detection method based on dynamic clustering described above, and will not be described in detail here.

[0117] In other embodiments, the frequency hopping signal detection device based on dynamic clustering of the present invention may also include a processor and a memory, wherein the memory is used to store a computer program and the processor is used to execute the computer program to perform the method described above.

[0118] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention should fall within the protection scope of the present invention.

Claims

1. A frequency hopping signal detection method based on dynamic clustering, characterized in that the steps include... include: Continuously acquire data from the signal under test, acquiring a segment of data each time to obtain the current segment of data; Transform the current segment of data to obtain the time-frequency data of the current segment; Frequency hopping signal detection is performed on the time-frequency data of the current segment to obtain multiple frequency hopping signal data points; Cluster the currently detected frequency hopping signal data points to obtain the frequency hopping signal clustering result for the current segment of data; Based on the information of each class in the current final clustering results, the required frequency hopping signal parameters are calculated and output. After each clustering is completed, the process also includes connecting the clustering results of the current segment of data frequency hopping signals with the clustering results of the previous segment of data frequency hopping signals, so as to classify the same frequency hopping signal data points into the same category and obtain the final clustering result. The step of connecting the current segment's frequency hopping signal clustering result with the previous segment's frequency hopping signal clustering result includes: Find the first frequency hopping signal category that meets the backward concatenation clustering condition from the previous data frequency hopping signal clustering results, and find the second frequency hopping signal category that meets the forward concatenation clustering condition from the current data frequency hopping signal clustering results; If the distance between the first frequency hopping signal category and the second frequency hopping signal category is less than a preset threshold, the first frequency hopping signal and the second frequency hopping signal category will be connected into the same category, and a new frequency hopping signal cluster set will be obtained for the next frequency hopping cluster connection of the previous clustering data; Once all frequency hopping signals that meet the connection conditions have been clustered, the final frequency hopping signal clustering result is obtained. The first frequency hopping signal category that satisfies the subsequent concatenation clustering condition from the previous segment of frequency hopping signal clustering results includes: Iterate through the frequency hopping signals in the previous data frequency hopping signal clustering results, and if a first frequency hopping signal category exists... Meet the conditions ,in Indicates the start time of the current time-frequency data segment. Indicates the time threshold. express The first frequency hopping signal category is determined by the end time of the hop. The conditions for backward-joining clustering are met; The step of finding the second frequency hopping signal category that satisfies the forward continuation clustering condition in the current segment data frequency hopping signal clustering results includes: Iterate through the frequency hopping signals in the current data segment's frequency hopping signal clustering results. If a second frequency hopping signal category exists... Meet the conditions , HOP k The start time of the category determines the second frequency hopping signal category HOP. k The conditions for forward-joining clustering are met; The distance between the first frequency hopping signal category and the second frequency hopping signal category is calculated using the following formula: in, These represent the weight coefficients for the horizontal and vertical axes, respectively. Less than 1, Greater than 1; if Then the category of frequency hopping signal in the previous paragraph will be... Category of the current frequency hopping signal Connect them into the same category, among which The neighborhood radius parameter set in the DBSCAN algorithm.

2. The frequency hopping signal detection method based on dynamic clustering according to claim 1, characterized in that, The step of detecting frequency hopping signals in the time-frequency data of the current segment includes: detecting frequency hopping signals in each snapshot of the time-frequency data of the current segment to obtain each frequency hopping signal data point. The frequency hopping signal data point includes the time when the frequency hopping signal occurs and the corresponding center frequency point position. A time-frequency plot of the frequency hopping signal is obtained by plotting each frequency hopping signal data point.

3. The frequency hopping signal detection method based on dynamic clustering according to claim 1, characterized in that, The DBSCAN clustering algorithm is used to cluster the frequency hopping signal data points, and different weight values ​​are configured for the horizontal and vertical axes when calculating the distance between two data points in the DBSCAN clustering algorithm.

4. The frequency hopping signal detection method based on dynamic clustering according to claim 3, characterized in that, The DBSCAN clustering algorithm uses the following formula to calculate the Euclidean distance between two data points: in, These represent the weight coefficients for the horizontal and vertical axes, respectively. Value less than 1 Value greater than 1 , These represent the x and y coordinates of two different data points.

5. The frequency hopping signal detection method based on dynamic clustering according to claim 1, characterized in that, The frequency hopping signal parameters include any one or more of the following: hopping start time, hopping end time, hopping period, hopping time, center frequency, signal bandwidth, and signal power. The frequency hopping signal parameters are calculated based on each time point of the frequency hopping signal point set corresponding to the clustering category. Specifically, the hopping start time is obtained from the minimum time value in the frequency hopping signal point set; the hopping end time is obtained from the maximum time value in the frequency hopping signal point set; the hopping period is obtained from the difference between the minimum and maximum time values; the hopping time is calculated based on the hopping start time and the hopping period; and the center frequency, signal bandwidth, and signal power are obtained by calculating the statistical values ​​of the center frequency, signal bandwidth, and signal power of each data point in the frequency hopping signal point set.

6. A frequency hopping signal detection device based on dynamic clustering, wherein the device is applied to the method described in any one of claims 1-5, characterized in that, The device includes: The data acquisition module is used to continuously acquire data from the signal under test, acquiring a segment of data each time to obtain the current segment of data; The time-frequency data acquisition module is used to transform the current segment data and obtain the time-frequency data of the current segment data; The frequency hopping signal detection module is used to detect frequency hopping signals in the time-frequency data of the current segment of data and obtain multiple frequency hopping signal data points. The dynamic clustering module is used to cluster the currently detected frequency hopping signal data points to obtain the frequency hopping signal clustering result of the current segment of data. The frequency hopping parameter calculation module is used to calculate the required frequency hopping signal parameters based on the information of each class in the current clustering results. After each clustering is completed, the dynamic clustering module also includes connecting the clustering results of the current segment of data frequency hopping signals with the clustering results of the previous segment of data frequency hopping signals, so as to classify the same frequency hopping signal data points into the same class and obtain the final clustering result.

7. A frequency hopping signal detection device based on dynamic clustering, comprising a processor and a memory, wherein the memory is used to store a computer program, characterized in that, The processor is used to execute the computer program to perform the method as described in any one of claims 1 to 5.

Images