Asynchronous network station sorting method based on take-off time of frequency hopping signal

Abstract

The invention belongs to the technical field of communication countermeasure, and discloses an asynchronous network station sorting method based on take-off time of a frequency hopping signal, which specifically comprises the following steps of: receiving the frequency hopping signal of a single channel, extracting an I-path signal and a Q-path signal, and generating a complex signal; performing short-time Fourier transform on the complex signal, and converting the complex signal into a time-frequency energy distribution matrix in a spectrogram form; estimating the number of network stations based on the time-frequency energy distribution matrix; performing signal enhancement processing on the time-frequency energy distribution matrix; performing focusing analysis on the enhanced time-frequency energy distribution matrix, extracting effective time-frequency data, extracting a frequency set and generating a frequency slice; extracting a multi-hop signal from the frequency slice, calculating residence time and determining a take-off moment; through single-channel receiving and time-frequency analysis technologies and in cooperation with an innovative take-off moment feature extraction method, efficient and reliable asynchronous network table sorting is achieved, and the problem that in the prior art, effective sorting can only be achieved by depending on a multi-channel receiving system is effectively solved.

Description

Technical Field

[0001] This invention belongs to the field of communication countermeasures technology, specifically relating to an asynchronous network station sorting method based on the start time of frequency hopping signals. Background Technology

[0002] Frequency-hopping communication technology, with its excellent anti-jamming performance, has become a core technology of modern military communication systems and is widely used in tactical data links (such as LINK16) and UAV communications. However, in the complex electromagnetic environment of the battlefield, the network sorting of frequency-hopping signals faces severe challenges; currently, network sorting mainly relies on blind source separation and frequency-hopping descriptor clustering methods.

[0003] While blind source separation methods can achieve blind sorting of multiple stations without prior parameters, they have significant shortcomings in practical applications. The separation effect drops significantly when processing four or more stations, and its adaptability to nonlinear mixed signals is poor. Blind source separation methods can be divided into independent component analysis under positive definite / super-positive definite conditions and a "two-step method" under underdetermined conditions, i.e., estimating the mixing matrix before separating the signal; in contrast, the "one-step method" is too computationally complex to be applied in practical systems. Frequency hopping descriptor clustering methods achieve synchronous and asynchronous station sorting by extracting parameters such as DOA, power, and hop time, but their performance is highly dependent on the accuracy of descriptor extraction. Synchronous sorting requires accurate DOA and power parameters, while asynchronous sorting is extremely sensitive to the accuracy of hop time parameters. This limits the reliability of the method in complex electromagnetic environments. Furthermore, existing methods generally suffer from algorithmic complexity, implementation difficulty, and reliance on multi-channel sensors; most research remains at the simulation verification stage, failing to meet practical needs. Therefore, existing technologies rely on multi-channel receiving systems for effective sorting. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide an asynchronous network station sorting method based on the start time of frequency hopping signals, thereby solving the problem that existing technologies rely on multi-channel receiving systems to achieve effective sorting.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] An asynchronous network station sorting method based on the start time of frequency hopping signals specifically includes the following steps:

[0007] It receives a single-channel frequency hopping signal, extracts the I and Q signals, and generates a complex signal;

[0008] Perform a short-time Fourier transform on the complex signal and convert it into a time-frequency energy distribution matrix in spectral form;

[0009] Estimate the number of network stations based on the time-frequency energy distribution matrix;

[0010] Perform signal enhancement processing on the time-frequency energy distribution matrix;

[0011] Focusing analysis is performed on the enhanced time-frequency energy distribution matrix to extract effective time-frequency data, extract frequency sets, and generate frequency slices.

[0012] Extract multi-hop signals from frequency slices, calculate dwell time, and determine take-off time;

[0013] Calculate the frequency hopping period based on the take-off time, and calculate the remainder matrix at the take-off time to generate the remainder feature sequence;

[0014] The network stations are classified using the remainder feature sequence, and the sorting results are output.

[0015] The process of receiving a single-channel frequency-hopping signal, extracting the I and Q signals, and generating a complex signal includes the following steps:

[0016] The frequency hopping signal is quadrature demodulated to separate the in-phase component I(m) and the quadrature component Q(m);

[0017] The complex signal X(m) is constructed based on the in-phase component I(m) and the quadrature component Q(m). The expression for X(m) is as follows:

[0018] x(m) = I(m) + jQ(m)

[0019] Where m = 1, 2, 3…M, M represents the number of data points obtained when the frequency hopping signal is acquired in a single channel.

[0020] Performing a short-time Fourier transform on the complex signal and converting it into a time-frequency energy distribution matrix in spectral form includes the following steps:

[0021] The short-time Fourier transform of a complex signal is expressed as follows:

[0022]

[0023] In the formula, x(m) represents the discrete form of the complex signal; w(i) represents the window function, N is the corresponding window length; n represents the time point, and k is the discrete frequency data.

[0024] STFT x (n,k) is a two-dimensional function, which represents the time-frequency amplitude value of the k-th frequency point at time n;

[0025] The spectrum is STFT x The square of the (n,k) modulus, the time-frequency energy distribution matrix STFT x The expression for (p,l) is:

[0026] STFT x (p,l)=|STFT x (n,k)| 2

[0027] Where p = 1, 2, 3…P represents the time axis; P represents the number of times the window slides in the short-time Fourier transform; l = 1, 2, 3,…L represents the frequency axis; and L is the number of points in the Fourier transform of the signal.

[0028] Based on the time-frequency energy distribution matrix, the number of network stations is estimated, which includes the following steps:

[0029] Iterate through the spectral peaks corresponding to each time point in the time-frequency energy distribution matrix;

[0030] For any given time point, the maximum spectral peak value within that time period is used. A multiple is used as a screening threshold, and valid spectral peaks are selected based on this threshold.

[0031] Count the number of valid spectral peaks at each time point;

[0032] The number of effective spectral peaks with the highest frequency, K, is taken as the number of network stations.

[0033] Signal enhancement processing of the time-frequency energy distribution matrix includes the following steps:

[0034] The spectral peaks in each time window of the short-time Fourier transform are arranged from largest to smallest.

[0035] Take the first K spectral peaks and increase them to the maximum spectral peak value in the current time window.

[0036] Focusing analysis is performed on the enhanced time-frequency energy distribution matrix to extract effective time-frequency data, extract frequency sets, and generate frequency slices. The specific steps include:

[0037] Set the extraction threshold θ p ,θ p The expression is as follows:

[0038] θ p =max(STFT) x (p,l))×a

[0039] In the formula, a is the set extraction coefficient;

[0040] The time-frequency energy distribution matrix contains elements smaller than the extraction threshold θ. p The time-frequency data is set to 0, and for values ​​greater than or equal to the extraction threshold θ, the threshold is set to 0. pThe time-frequency data is preserved to obtain an effective time-frequency data matrix Y(p,l), where the expression for Y(p,l) is as follows:

[0041]

[0042] For each frequency point in Y(p,l), calculate the relationship vector between the frequency and the amplitude of the Y(p,l) signal. The relationship vector y corresponding to any frequency point l is... l The expression is as follows:

[0043]

[0044] For any frequency point l, a pole condition is set, and a local pole is selected. The expression for the pole condition is:

[0045]

[0046] y l-1 and y l+1 The relationship vectors between two adjacent frequency points of frequency point l;

[0047] After obtaining all local poles, the overall frequency set f is calculated as follows:

[0048]

[0049] In the formula, fs is the sampling frequency of the single-channel frequency-hopping signal data; L is the number of points in the Fourier transform; l max It is a sequence of local poles;

[0050] Based on the frequency set, the signal at each frequency point is extracted to generate a frequency slice corresponding to each frequency point. Each frequency slice contains at least one signal.

[0051] Extracting multi-hop signals from frequency slices, calculating dwell time, and determining the start time specifically includes the following steps:

[0052] For each signal in the frequency slice, iterate through the zero-crossing points and record the zero-crossing sequence;

[0053] The start and end positions of each signal are determined based on the zero-crossing sequence, and each signal segment is extracted as an independent frequency-hopping signal, i.e., a multi-hop signal.

[0054] Calculate the time difference between the start and end positions of each signal segment to obtain the dwell time Ttime;

[0055] Thresholds b1 and b2 are set as follows:

[0056]

[0057] Interference signals smaller than b1 or larger than b2 are eliminated, and valid frequency-hopping signal segments between b1 and b2 are retained;

[0058] An effective frequency hopping signal segment is represented by a narrow pulse signal, and the zero-crossing point of the rising edge of each pulse is the start-up time.

[0059] The frequency hopping period is calculated based on the takeoff time, and the remainder matrix at the takeoff time is calculated to generate the remainder feature sequence. The specific steps include:

[0060] The extracted jump times are arranged in chronological order to form a jump time sequence t(h), where h = 1, 2, 3, ..., H, and H is the number of multi-hop signals;

[0061] Perform a K-order difference operation on the jump time series t(h), as shown in the following formula:

[0062] T K (h)=t(h)-t(hK)

[0063] Statistical difference operation result T K (h) is the frequency hopping period T;

[0064] Using the first takeoff time as the baseline, calculate the difference between the first takeoff time and each other takeoff time, and then perform a remainder operation over the period T. The formula is as follows:

[0065] tr(h) = mod([t(h) - t(hi)], T)

[0066] In the formula, mod represents the modulo operation; i = 1, 2, 3, ... M, which is the number of multi-hop signals, and i ≠ h;

[0067] Using the second take-off time as the baseline, calculate the difference between the second take-off time and each other take-off time;

[0068] This process continues until the Mth take-off time is used as the baseline, and the difference between the Mth take-off time and each other take-off time is calculated.

[0069] Finally, tr(h) is obtained, which is an M×M remainder matrix;

[0070] Summing the remainder matrix tr(h) column by column yields a 1×M array. The remainder corresponding to the minimum value in the array is then used to generate the remainder feature sequence.

[0071] The classification of network stations is carried out using the remainder feature sequence, and the sorting results are output. The specific steps include:

[0072] Using the estimated number of network stations K as the cluster number, a single observation is randomly and uniformly selected from the remainder feature sequence as the first centroid C1;

[0073] Calculate the distance from each observation in the remainder feature sequence to the selected centroid. When selecting the next centroid Cj, the selection probability P is proportional to the square of the distance from the observation to the nearest selected centroid.

[0074] Repeat the above steps, selecting centroids C2, C3, ..., CK in sequence, until K centroids are selected, forming K clusters;

[0075] The observations within each cluster are analyzed to extract the cluster's characteristic parameters, including the frequency hopping period, dwell time, and frequency set.

[0076] Based on the cluster division and extracted feature parameters, the network station category to which each frequency hopping signal belongs is determined.

[0077] The beneficial effects of this invention are:

[0078] This invention achieves efficient and reliable asynchronous network station sorting through single-channel reception and time-frequency analysis technology, combined with an innovative method for extracting features at the take-off moment. This method utilizes signal enhancement processing and remainder matrix operations to significantly reduce system complexity while maintaining sorting accuracy, effectively overcoming the dependence of traditional methods on multi-channel systems. Through the coordinated processing of dwell time calculation and feature classification, it can accurately identify and suppress various interference signals, enabling the system to maintain stable performance even in complex electromagnetic environments. It boasts significant advantages such as flexible deployment, low cost, and strong practicality. Attached Figure Description

[0079] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0080] Figure 1 This is a schematic diagram of the asynchronous network station sorting method of the present invention. Detailed Implementation

[0081] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0082] like Figure 1 As shown, an asynchronous network station sorting method based on the start time of frequency hopping signals specifically includes the following steps:

[0083] It receives a single-channel frequency hopping signal, extracts the I and Q signals, and generates a complex signal;

[0084] Perform a short-time Fourier transform on the complex signal and convert it into a time-frequency energy distribution matrix in spectral form;

[0085] Estimate the number of network stations based on the time-frequency energy distribution matrix;

[0086] Perform signal enhancement processing on the time-frequency energy distribution matrix;

[0087] Focusing analysis is performed on the enhanced time-frequency energy distribution matrix to extract effective time-frequency data, extract frequency sets, and generate frequency slices.

[0088] Extract multi-hop signals from frequency slices, calculate dwell time, and determine take-off time;

[0089] Calculate the frequency hopping period based on the take-off time, and calculate the remainder matrix at the take-off time to generate the remainder feature sequence;

[0090] The network stations are classified using the remainder feature sequence, and the sorting results are output.

[0091] The process of receiving a single-channel frequency-hopping signal, extracting the I and Q signals, and generating a complex signal includes the following steps:

[0092] The frequency hopping signal is quadrature demodulated to separate the in-phase component I(m) and the quadrature component Q(m);

[0093] The complex signal X(m) is constructed based on the in-phase component I(m) and the quadrature component Q(m). The expression for X(m) is as follows:

[0094] x(m) = I(m) + jQ(m)

[0095] Where m = 1, 2, 3…M, M represents the number of data points obtained when the frequency hopping signal is acquired in a single channel.

[0096] Performing a short-time Fourier transform on the complex signal and converting it into a time-frequency energy distribution matrix in spectral form includes the following steps:

[0097] The short-time Fourier transform of a complex signal is expressed as follows:

[0098]

[0099] In the formula, x(m) represents the discrete form of the complex signal; w(i) represents the window function, N is the corresponding window length; n represents the time point, and k is the discrete frequency data.

[0100] STFT x(n,k) is a two-dimensional function, which represents the time-frequency amplitude value of the k-th frequency point at time n;

[0101] The spectrum is STFT x The square of the (n,k) modulus, the time-frequency energy distribution matrix STFT x The expression for (p,l) is:

[0102] STFT x (p,l)=|STFT x (n,k)| 2

[0103] Where p = 1, 2, 3…P represents the time axis; P represents the number of times the window slides in the short-time Fourier transform; l = 1, 2, 3,…L represents the frequency axis; and L is the number of points in the Fourier transform of the signal.

[0104] Based on the time-frequency energy distribution matrix, the number of network stations is estimated, which includes the following steps:

[0105] Iterate through the spectral peaks corresponding to each time point in the time-frequency energy distribution matrix;

[0106] For any given time point, the maximum spectral peak value within that time period is used. A multiple is used as a screening threshold, and valid spectral peaks are selected based on this threshold.

[0107] Count the number of valid spectral peaks at each time point;

[0108] The number of effective spectral peaks with the highest frequency, K, is taken as the number of network stations.

[0109] Signal enhancement processing of the time-frequency energy distribution matrix includes the following steps:

[0110] The spectral peaks in each time window of the short-time Fourier transform are arranged from largest to smallest.

[0111] Take the first K spectral peaks and increase them to the maximum spectral peak value in the current time window.

[0112] Focusing analysis is performed on the enhanced time-frequency energy distribution matrix to extract effective time-frequency data, extract frequency sets, and generate frequency slices. The specific steps include:

[0113] Set the extraction threshold θ p ,θ p The expression is as follows:

[0114] θ p =max(STFT) x (p,l))×a

[0115] In the formula, a is the set extraction coefficient;

[0116] The time-frequency energy distribution matrix contains elements smaller than the extraction threshold θ. p The time-frequency data is set to 0, and for values ​​greater than or equal to the extraction threshold θ, the threshold is set to 0. p The time-frequency data is preserved to obtain an effective time-frequency data matrix Y(p,l), where the expression for Y(p,l) is as follows:

[0117]

[0118] For each frequency point in Y(p,l), calculate the relationship vector between the frequency and the amplitude of the Y(p,l) signal. The relationship vector y corresponding to any frequency point l is... l The expression is as follows:

[0119]

[0120] For any frequency point l, a pole condition is set, and a local pole is selected. The expression for the pole condition is:

[0121]

[0122] y l-1 and y l+1 The relationship vectors between two adjacent frequency points of frequency point l;

[0123] After obtaining all local poles, the overall frequency set f is calculated as follows:

[0124]

[0125] In the formula, fs is the sampling frequency of the single-channel frequency-hopping signal data; L is the number of points in the Fourier transform; l max It is a sequence of local poles;

[0126] Based on the frequency set, the signal at each frequency point is extracted to generate a frequency slice corresponding to each frequency point. Each frequency slice contains at least one signal.

[0127] Extracting multi-hop signals from frequency slices, calculating dwell time, and determining the start time specifically includes the following steps:

[0128] For each signal in the frequency slice, iterate through the zero-crossing points and record the zero-crossing sequence;

[0129] The start and end positions of each signal are determined based on the zero-crossing sequence, and each signal segment is extracted as an independent frequency-hopping signal, i.e., a multi-hop signal.

[0130] Calculate the time difference between the start and end positions of each signal segment to obtain the dwell time Ttime;

[0131] Thresholds b1 and b2 are set as follows:

[0132]

[0133] Interference signals smaller than b1 or larger than b2 are eliminated, and valid frequency-hopping signal segments between b1 and b2 are retained;

[0134] An effective frequency hopping signal segment is represented by a narrow pulse signal, and the zero-crossing point of the rising edge of each pulse is the start-up time.

[0135] The frequency hopping period is calculated based on the takeoff time, and the remainder matrix at the takeoff time is calculated to generate the remainder feature sequence. The specific steps include:

[0136] The extracted jump times are arranged in chronological order to form a jump time sequence t(h), where h = 1, 2, 3, ..., H, and H is the number of multi-hop signals;

[0137] Perform a K-order difference operation on the jump time series t(h), as shown in the following formula:

[0138] T K (h)=t(h)-t(hK)

[0139] Statistical difference operation result T K (h) is the frequency hopping period T;

[0140] Using the first takeoff time as the baseline, calculate the difference between the first takeoff time and each other takeoff time, and then perform a remainder operation over the period T. The formula is as follows:

[0141] tr(h) = mod([t(h) - t(hi)], T)

[0142] In the formula, mod represents the modulo operation; i = 1, 2, 3, ... M, which is the number of multi-hop signals, and i ≠ h;

[0143] Using the second take-off time as the baseline, calculate the difference between the second take-off time and each other take-off time;

[0144] This process continues until the Mth take-off time is used as the baseline, and the difference between the Mth take-off time and each other take-off time is calculated.

[0145] Finally, tr(h) is obtained, which is an M×M remainder matrix;

[0146] Summing the remainder matrix tr(h) column by column yields a 1×M array. The remainder corresponding to the minimum value in the array is then used to generate the remainder feature sequence.

[0147] The classification of network stations is carried out using the remainder feature sequence, and the sorting results are output. The specific steps include:

[0148] Using the estimated number of network stations K as the cluster number, a single observation is randomly and uniformly selected from the remainder feature sequence as the first centroid C1;

[0149] Calculate the distance from each observation in the remainder feature sequence to the selected centroid. When selecting the next centroid Cj, the selection probability P is proportional to the square of the distance from the observation to the nearest selected centroid.

[0150] Repeat the above steps, selecting centroids C2, C3, ..., CK in sequence, until K centroids are selected, forming K clusters;

[0151] Taking the observed value xm as an example, the expression for the probability P(xm) that xm is selected is as follows:

[0152]

[0153] In the formula, d(xm, C) nearest ) 2 This indicates the distance from the observed value xm to the nearest selected centroid C. nearest The square of the distance;

[0154] ∑ x∈X d(x, C) nearest ) 2 This indicates that all observations up to the nearest selected centroid C nearest The sum of the squares of the distances;

[0155] The farther an observation is from the selected centroid, the greater the probability that it will be selected as the next centroid; this ensures that the distance between the new centroid and the selected centroid is as far as possible, thereby improving the efficiency and quality of clustering.

[0156] The observations within each cluster are analyzed to extract the cluster's characteristic parameters, including the frequency hopping period, dwell time, and frequency set.

[0157] Based on the cluster division and extracted feature parameters, the network station category to which each frequency hopping signal belongs is determined.

[0158] This invention achieves efficient and reliable asynchronous network station sorting through single-channel reception and time-frequency analysis technology, combined with an innovative method for extracting features at the take-off moment. This method utilizes signal enhancement processing and remainder matrix operations to significantly reduce system complexity while maintaining sorting accuracy, effectively overcoming the dependence of traditional methods on multi-channel systems. Through the coordinated processing of dwell time calculation and feature classification, it can accurately identify and suppress various interference signals, enabling the system to maintain stable performance even in complex electromagnetic environments. It boasts significant advantages such as flexible deployment, low cost, and strong practicality.

[0159] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0160] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. An asynchronous network station sorting method based on the start time of frequency hopping signals, characterized in that, Specifically, the following steps are included: It receives a single-channel frequency hopping signal, extracts the I and Q signals, and generates a complex signal; Perform a short-time Fourier transform on the complex signal and convert it into a time-frequency energy distribution matrix in spectral form; Estimate the number of network stations based on the time-frequency energy distribution matrix; Perform signal enhancement processing on the time-frequency energy distribution matrix; Focusing analysis is performed on the enhanced time-frequency energy distribution matrix to extract effective time-frequency data, extract frequency sets, and generate frequency slices. Extract multi-hop signals from frequency slices, calculate dwell time, and determine take-off time; Calculate the frequency hopping period based on the take-off time, and calculate the remainder matrix at the take-off time to generate the remainder feature sequence; The network stations are classified using the remainder feature sequence, and the sorting results are output.

2. The asynchronous network station sorting method according to claim 1, characterized in that, The process of receiving a single-channel frequency-hopping signal, extracting the I and Q signals, and generating a complex signal includes the following steps: The frequency hopping signal is quadrature demodulated to separate the in-phase component I(m) and the quadrature component Q(m); The complex signal X(m) is constructed based on the in-phase component I(m) and the quadrature component Q(m). The expression for X(m) is as follows: x(m) = I(m) + jQ(m) Where m = 1, 2, 3…M, M represents the number of data points obtained when the frequency hopping signal is acquired in a single channel.

3. The asynchronous network station sorting method according to claim 2, characterized in that, Performing a short-time Fourier transform on the complex signal and converting it into a time-frequency energy distribution matrix in spectral form includes the following steps: The short-time Fourier transform of a complex signal is expressed as follows: In the formula, x(m) represents the discrete form of the complex signal; w(i) represents the window function, N is the corresponding window length; n represents the time point, and k is the discrete frequency data. STFT x (n,k) is a two-dimensional function, which represents the time-frequency amplitude value of the k-th frequency point at time n; The spectrum is STFT x The square of the (n,k) modulus, the time-frequency energy distribution matrix STFT x The expression for (p,l) is: STFT x (p,l)=|STFT x (n,k)| 2 Where p = 1, 2, 3…P represents the time axis; P represents the number of times the window slides in the short-time Fourier transform; l = 1, 2, 3,…L represents the frequency axis; and L is the number of points in the Fourier transform of the signal.

4. The asynchronous network station sorting method according to claim 3, characterized in that, Based on the time-frequency energy distribution matrix, the number of network stations is estimated, which includes the following steps: Iterate through the spectral peaks corresponding to each time point in the time-frequency energy distribution matrix; For any given time point, the maximum spectral peak value within that time period is used. A multiple is used as a screening threshold, and valid spectral peaks are selected based on this threshold. Count the number of valid spectral peaks at each time point; The number of effective spectral peaks with the highest frequency, K, is taken as the number of network stations.

5. The asynchronous network station sorting method according to claim 4, characterized in that, Signal enhancement processing of the time-frequency energy distribution matrix includes the following steps: The spectral peaks in each time window of the short-time Fourier transform are arranged from largest to smallest. Take the first K spectral peaks and increase them to the maximum spectral peak value in the current time window.

6. The asynchronous network station sorting method according to claim 5, characterized in that, Focusing analysis is performed on the enhanced time-frequency energy distribution matrix to extract effective time-frequency data, extract frequency sets, and generate frequency slices. The specific steps include: Set the extraction threshold θ p ,θ p The expression is as follows: θ p =max(STFT x (p,l))×a In the formula, a is the set extraction coefficient; The time-frequency energy distribution matrix contains elements smaller than the extraction threshold θ. p The time-frequency data is set to 0, and for values ​​greater than or equal to the extraction threshold θ, the threshold is set to 0. p The time-frequency data is preserved to obtain an effective time-frequency data matrix Y(p,l), where the expression for Y(p,l) is as follows: For each frequency point in Y(p,l), calculate the relationship vector between the frequency and the amplitude of the Y(p,l) signal. The relationship vector y corresponding to any frequency point l is... l The expression is as follows: For any frequency point l, a pole condition is set, and a local pole is selected. The expression for the pole condition is: y l-1 and y l+1 The relationship vectors between two adjacent frequency points of frequency point l; After obtaining all local poles, the overall frequency set f is calculated as follows: In the formula, fs is the sampling frequency of the single-channel frequency-hopping signal data; L is the number of points in the Fourier transform; l max It is a sequence of local poles; Based on the frequency set, the signal at each frequency point is extracted to generate a frequency slice corresponding to each frequency point. Each frequency slice contains at least one signal.

7. The asynchronous network station sorting method according to claim 6, characterized in that, Extracting multi-hop signals from frequency slices, calculating dwell time, and determining the start time specifically includes the following steps: For each signal in the frequency slice, iterate through the zero-crossing points and record the zero-crossing sequence; The start and end positions of each signal are determined based on the zero-crossing sequence, and each signal segment is extracted as an independent frequency-hopping signal, i.e., a multi-hop signal. Calculate the time difference between the start and end positions of each signal segment to obtain the dwell time Ttime; Thresholds b1 and b2 are set as follows: Interference signals smaller than b1 or larger than b2 are eliminated, and valid frequency-hopping signal segments between b1 and b2 are retained; An effective frequency hopping signal segment is represented by a narrow pulse signal, and the zero-crossing point of the rising edge of each pulse is the start-up time.

8. The asynchronous network station sorting method according to claim 7, characterized in that, The frequency hopping period is calculated based on the takeoff time, and the remainder matrix at the takeoff time is calculated to generate the remainder feature sequence. The specific steps include: The extracted jump times are arranged in chronological order to form a jump time sequence t(h), where h = 1, 2, 3, ..., H, and H is the number of multi-hop signals; Perform a K-order difference operation on the jump time series t(h), as shown in the following formula: T K (h)=t(h)-t(h-K) Statistical difference operation result T K (h) is the frequency hopping period T; Using the first takeoff time as the baseline, calculate the difference between the first takeoff time and each other takeoff time, and then perform a remainder operation over the period T. The formula is as follows: tr(h) = mod([t(h) - t(hi)], T) In the formula, mod represents the modulo operation; i = 1, 2, 3, ... M, which is the number of multi-hop signals, and i ≠ h; Using the second take-off time as the baseline, calculate the difference between the second take-off time and each other take-off time; This process continues until the Mth take-off time is used as the baseline, and the difference between the Mth take-off time and each other take-off time is calculated. Finally, tr(h) is obtained, which is an M×M remainder matrix; Summing the remainder matrix tr(h) column by column yields a 1×M array. The remainder corresponding to the minimum value in the array is then used to generate the remainder feature sequence.

9. The asynchronous network station sorting method according to claim 8, characterized in that, The classification of network stations is carried out using the remainder feature sequence, and the sorting results are output. The specific steps include: Using the estimated number of network stations K as the cluster number, a single observation is randomly and uniformly selected from the remainder feature sequence as the first centroid C1; Calculate the distance from each observation in the remainder feature sequence to the selected centroid. When selecting the next centroid Cj, the selection probability P is proportional to the square of the distance from the observation to the nearest selected centroid. Repeat the above steps, selecting centroids C2, C3, ..., CK in sequence, until K centroids are selected, forming K clusters; The observations within each cluster are analyzed to extract the cluster's characteristic parameters, including the frequency hopping period, dwell time, and frequency set. Based on the cluster division and extracted feature parameters, the network station category to which each frequency hopping signal belongs is determined.

Images