A method for sorting frequency hopping signals from multiple networks in complex environments

Abstract

The present invention provides a method for sorting multi-network frequency hopping signals in a complex environment, including: establishing a mathematical model of multi-network frequency hopping signals in a frequency hopping communication system in a complex environment, performing short-time Fourier transform on the multi-network mixed frequency hopping signals to obtain a time-frequency energy matrix; performing adaptive threshold denoising based on an energy distribution histogram on the time-frequency matrix of the received signal; performing morphological filtering to remove sweep frequency interference and part of short burst interference; marking connected domains and counting the duration of each connected domain, removing short burst interference and fixed frequency interference according to the duration, and obtaining a pure signal containing only frequency hopping signals; extracting the average energy of each segment of the pure signal, and performing MeanShift clustering to sort each frequency hopping segment according to the average energy of each segment. Under low signal-to-noise ratio, compared with other frequency hopping network station sorting algorithms, the algorithm of the present invention has stronger anti-interference ability, higher sorting rate, wider applicability, and higher engineering practice value.

Description

Technical Field

[0001] The invention relates to the technical field of military communications, and in particular to a method for sorting frequency hopping signals of multiple stations under complex environments. Background Art

[0002] The rapid development of communication technology has led to the in-depth application of radio communication in both military and civilian fields. At the same time, radio communication networks have become more and more complex, and more and more interference has been introduced in the communication process, including intentional sabotage and interference by non-cooperative parties, resulting in an increasingly complex wireless communication environment. The usual fixed-frequency communication has been difficult to meet the requirements of safe and reliable communication. In the battlefield environment, the signals that need to be sent have the characteristics of anti-interference, anti-interception, and difficult to decipher, so as to gain the initiative in the confrontation. Spread spectrum communication came into being. It has excellent anti-interference and anti-interception characteristics and has developed rapidly in the military and civilian fields. Frequency hopping communication is one of the ways of spread spectrum communication. Due to its good confidentiality, anti-interference ability, anti-interception ability and flexible networking characteristics, it has been widely used in non-civilian fields such as shortwave communication and ultra-short wave communication. Frequency hopping communication has been widely used in military communications since the 1980s. The US military's LINK16 data link technology has played a vital role in gaining the initiative in various modern wars. In addition, frequency hopping communication also plays an important role in civilian fields such as Bluetooth and GSM.

[0003] As the name implies, frequency hopping communication refers to a communication method in which the carrier frequency of the transmitted signal changes continuously over time according to a certain rule. The advantage of this communication method is that if the non-cooperative party wants to interfere with the enemy's frequency hopping communication without accurately tracking the enemy's frequency hopping set, it can only perform blocking interference in the entire frequency band or most of the frequency band. When the enemy's frequency hopping bandwidth is large, the non-cooperative jammer must use a lot of power to work, which not only wastes energy, but may also affect the operation of its own radio station and even expose its own information. Therefore, in order to conduct high-efficiency, high-precision, and high-real-time tracking interference on the enemy radio station, it is imperative to carry out research on reconnaissance technology related to frequency hopping communication.

[0004] If you want to take the lead in the trend of electronic warfare, frequency hopping communication reconnaissance technology is an indispensable part. Frequency hopping signal existence detection, frequency hopping signal feature parameter extraction and frequency hopping signal sorting are three important links in the reconnaissance process. Among them, frequency hopping signal sorting is the most challenging part in the field of frequency hopping signal reconnaissance. The aliased signal received by the receiver may contain frequency hopping signals generated by multiple frequency hopping network stations. How to accurately sort out the frequency hopping signals of each network station without prior knowledge and implement targeted interference is a research with great research value and strategic significance.

[0005] The communication environment of modern battlefield is becoming increasingly complex. With the application of frequency hopping communication in shortwave channels, the channel environment is full of various sudden interferences, frequency sweeping interferences and fixed frequency interferences in addition to background noise, which poses a challenge to the sorting of frequency hopping network stations. Therefore, the research on multi-frequency hopping network station sorting technology in complex environments has great strategic significance and theoretical value for enhancing the development of electronic countermeasure technology and promoting the construction of informationized troops. Summary of the invention

[0006] In view of the defects in the prior art, the present invention considers the whole process from receiving to sorting of mixed frequency hopping signals of multiple networks, and provides a method for sorting frequency hopping signals of multiple networks in a complex environment. The frequency hopping signal duration and average energy extracted from the short-time Fourier transform of the signal are used, combined with MeanShift clustering to sort the frequency hopping signals of multiple networks, solve the problem of reconnaissance and reception of frequency hopping signals under wartime communication confrontation, and achieve the purpose of improving the accuracy of signal sorting.

[0007] In a first aspect, the present invention provides a method for sorting frequency hopping signals of multiple networks in a complex environment, the method comprising:

[0008] S1: Establish a mathematical model of multi-network frequency hopping signals in a frequency hopping communication system under complex environments, and perform short-time Fourier transform on the multi-network mixed frequency hopping signals to obtain the time-frequency energy matrix;

[0009] S2: Adaptive threshold denoising based on energy distribution histogram is performed on the received signal according to its time-frequency energy matrix;

[0010] S3: Perform morphological filtering to remove sweep frequency interference and some short burst interference;

[0011] S4: Mark the connected domains and count the duration of each connected domain, remove short burst interference and fixed frequency interference according to the duration, and obtain a pure signal containing only the frequency hopping signal;

[0012] S5: taking the average of the time-frequency points in the top 80% of the energy in each pure signal, and performing MeanShift clustering to sort each frequency hopping segment according to the energy mean of each segment.

[0013] Preferably, the step S1 specifically includes:

[0014] A mathematical model of multi-frequency hopping signals in a frequency hopping communication system under complex environments is established. The aliased signal received by the receiver should contain not only multiple frequency hopping network signals, but also multiple fixed-frequency interference signals, swept-frequency interference signals, burst interference signals and background noise. The received aliased signal is short-time Fourier transformed using time-frequency analysis technology to obtain its time-frequency energy matrix.

[0015] Preferably, the step S2 specifically includes:

[0016] The logarithmic results of the energy size of each time-frequency point in the time-frequency energy matrix are statistically analyzed to draw a time-frequency energy distribution histogram. The time-frequency energy size corresponding to the horizontal coordinate of the trough on the left between the two peaks is taken as the adaptive denoising threshold. The time-frequency points in the time-frequency energy matrix whose energy is less than the denoising threshold are assigned zero, thereby completing denoising and obtaining the denoised time-frequency energy matrix.

[0017] Preferably, the step S3 specifically includes:

[0018] The denoised time-frequency energy matrix is ​​subjected to morphological closing operation to fill the gaps in the frequency hopping signal HOP that may be caused by the threshold denoising process. Then, the entire matrix is ​​subjected to morphological erosion operation using shorter structural elements to remove sweep frequency interference and some short burst interference.

[0019] Preferably, the step S4 specifically includes:

[0020] The connected domain marking algorithm based on the stack idea is used to mark the connected domain of the time-frequency energy matrix, and the duration of each time-frequency signal segment is determined. The fixed-frequency interference and burst interference are removed according to the duration, and a pure signal containing only the frequency hopping signal is obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 It is a flow chart of a method for sorting frequency hopping signals of multiple networks in a complex environment provided by the present invention;

[0023] Figure 2 It is the grayscale time-frequency diagram of the synchronous networking aliasing signal collected within the observation time;

[0024] Figure 3 The time-frequency energy distribution histogram is drawn from the time-frequency energy matrix of the aliased signal;

[0025] Figure 4 It is the time-frequency diagram of the synchronous networking aliasing signal after being denoised by adaptive threshold based on energy distribution histogram;

[0026] Figure 5 for Figure 4 The marking result of each frequency hopping signal segment obtained after the synchronous networking signal is performed in step S4;

[0027] Figure 6for Figure 5 The clustering result diagram after MeanShift energy clustering of the synchronous networking signal;

[0028] Figure 7 To sort out the time-frequency diagram of frequency hopping signal 1;

[0029] Figure 8 To sort out the time-frequency diagram of the frequency hopping signal 2;

[0030] Fig. 9 It is the grayscale time-frequency diagram of the asynchronous networking aliasing signal collected during the observation time;

[0031] Fig.10 To sort out the time-frequency diagram of the frequency hopping signal 3;

[0032] Fig.11 To sort out the time-frequency diagram of the frequency hopping signal 4;

[0033] Fig.12 This is a graph showing the variation of the accuracy of the present invention in sorting the synchronous networking frequency hopping signal with the SNR;

[0034] Fig.13 This is a graph showing the variation of the accuracy of sorting asynchronous networking frequency hopping signals with SNR in the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] like Figure 1 As shown, a method for sorting frequency hopping signals of multiple networks in a complex environment is provided by an embodiment of the present invention. The method comprises the following steps:

[0037] S1: Establish a mathematical model of multi-network frequency hopping signals in a frequency hopping communication system under complex environments, and perform short-time Fourier transform on the multi-network mixed frequency hopping signals to obtain the time-frequency energy matrix;

[0038] S2: Adaptive threshold denoising based on energy distribution histogram is performed on the received signal according to its time-frequency energy matrix;

[0039] S3: Perform morphological filtering to remove sweep frequency interference and some short burst interference;

[0040] S4: Mark the connected domains and count the duration of each connected domain, remove short burst interference and fixed frequency interference according to the duration, and obtain a pure signal containing only the frequency hopping signal;

[0041] S5: taking the average of the time-frequency points in the top 80% of the energy in each pure signal, and performing MeanShift clustering to sort each frequency hopping segment according to the energy mean of each segment.

[0042] In this embodiment, the specific process of step S1 is as follows:

[0043] In the frequency hopping communication system, the mathematical model of the single frequency hopping signal at the transmitter can be defined as follows:

[0044]

[0045] In the above formula, 0<t≤T, S is the signal power, The width is T h Rectangular window, T h is the frequency hopping signal period, k is the number of hops, and f k is the frequency hopping frequency, θ is the timing deviation (θ = αT h ).

[0046] In a complex channel environment, there are noises, burst interference, sweep interference and fixed frequency interference in the transmission channel. For mixed signals of multiple networks, the actual received signal model in a period of observation time should be:

[0047]

[0048] Where n≥1, l≥1, i≥1, j≥1, r(t) is the received shortwave frequency hopping signal, s n (t) is the frequency hopping signal, c l (t) is the frequency sweep interference, d i (t) is fixed frequency interference, e j (t) is the burst interference, and v(t) is the additive Gaussian white noise.

[0049] The shortwave signal parameters of this implementation are as follows:

[0050] Signal-to-Noise Ratio (SNR) -5~10dB Signal sampling rate 10Mhz Observation time 4s Frequency Hopping Signal 1 Frequency Set [0.5,2,2.5,1.5,3.5,3,4,1]Mhz Frequency Hopping Signal 2 Frequency Set [2.5,1.5,3.5,3,4,1,0.5,2]Mhz Frequency Hopping Signal 3 Frequency Set [0.25,1,1.25,0.75,1.75,1.5,2,0.5]Mhz Frequency Hopping Signal 4 Frequency Set [3.5,2.5,4,3,4.5]Mhz Amplitude ratio of frequency hopping signals 1 to 4 1：1.2：1：1.2 Sweep signal frequency range 2.75Mhz~3.75Mhz Burst signal frequency set [0.6,1.4,1.8,2.2,3.8,4.2]Mhz Fixed frequency signal frequency set [2.6,4.6,3.8]Mhz

[0051] For the sorted aliased signal r(t), the formula for STFT transformation is:

[0052]

[0053] Where h(t) is the window function. r(t,f) is the time-frequency point matrix obtained after r(t) is transformed by STFT. Where t=1,2,…,T represents the time axis (T is the number of times the STFT window moves), and f=1,2,…,F represents the frequency axis (F is the number of points of the signal FFT).

[0054] Figure 2 This is a grayscale time-frequency diagram of the synchronous networking aliasing signal collected within the observation time of the present invention. It can be seen that the diagram contains the frequency hopping signals 1, 2 and various bursts and background noise.

[0055] In this embodiment, the specific process of step S2 is as follows:

[0056] STFT of the time-frequency energy matrix r The logarithmic results of the energy size of each time-frequency point in (t,f) are statistically analyzed to draw a time-frequency energy distribution histogram, such as Figure 3 The time-frequency energy corresponding to the horizontal coordinate of the trough on the left between the two peaks is taken as the adaptive denoising threshold T, and the time-frequency points with energy less than T in the time-frequency energy matrix are assigned zero, thereby completing denoising and obtaining the denoised time-frequency energy matrix STFT. r _2(t,f).

[0057] Figure 4 for Figure 2 The grayscale time-frequency diagram after denoising is Figure 4 Knowable Figure 3 The white fog noise in the image has been removed and the signal is well preserved.

[0058] In this embodiment, the specific process of step S3 is as follows:

[0059] STFT of the denoised time-frequency energy matrix r _2(t,f) selects a linear structure element b1 of appropriate length to perform morphological closing operation to fill the gap in the middle of the frequency hopping signal HOP that may be caused in the threshold denoising process:

[0060]

[0061] Then, a shorter structural element b2 is used to perform morphological corrosion operation on the entire matrix to remove sweep frequency interference and some short burst interference:

[0062] TF2=TF1θ g b2 (5)

[0063] Among them, TF1 and TF2 are two-dimensional time-frequency energy matrices, b1 and b2 represent structural elements, Θ g represents the morphological erosion operation, represents the morphological dilation operation, g Represents the morphological closing operation.

[0064] In this embodiment, the specific process of step S4 is as follows:

[0065] After removing the frequency sweep interference and some short burst interference, in order to distinguish the frequency hopping HOP from the fixed frequency interference and eliminate the fixed frequency interference without additional short burst interference, it is necessary to mark the connected domain of the time-frequency energy matrix. In order to reduce the complexity and perform fast marking, the present invention adopts an 8-connected connected domain marking algorithm based on the stack idea. The specific process of marking is as follows:

[0066] 1) Create a connected domain labeling matrix L1 with the same size as the time-frequency energy matrix TF, and initialize each element to 0. The initial coordinate for detection is (1,1).

[0067] 2) Check the coordinate TF(x,y). If TF(x,y)≠0 and has not been marked, mark it and push it into the stack, and then proceed to step 3; if TF(x,y)=0 or has been marked, continue to check the next coordinate point.

[0068] 3) Perform 8-connectivity check on the top coordinate point of the stack. If an unlabeled adjacent point (m, n) is found, assign the same label as (m, n) and push it into the stack; otherwise, pop the top coordinate point of the stack. Repeat step 3 until the stack is empty.

[0069] 4) Perform the second step detection on the time-frequency energy matrix in order from left to right and from top to bottom until the last coordinate point of the time-frequency energy matrix is ​​detected.

[0070] After the above steps, the marked time-frequency energy matrix TF3 can be obtained. The maximum row number and the minimum row number marked by each marked connected domain are calculated. The difference between the two is the duration of this connected domain, which is stored in the array Holdtime in the order of the numbers. The horizontal coordinates of the centroids of each connected domain are calculated and stored in the array X. Because the same fixed-frequency interference signal is segmented into multiple connected domains, the frequencies of these connected domains should be almost the same. Therefore, according to the array Holdtime and the array X, the duration points of several connected domains with similar centroids are added. If the addition result is almost the same as the entire signal observation duration, the corresponding time-frequency point in the time-frequency energy matrix TF3 is determined to be fixed-frequency interference and set to 0; because the duration points of short burst interference are significantly smaller than frequency hopping HOP and fixed-frequency interference, the time-frequency points in the time-frequency energy matrix TF3 corresponding to the connected domain with a duration point less than a certain threshold are determined to be short burst interference and set to 0. At this time, the time-frequency energy matrix TF3 is a pure signal containing only the hopping segments of each frequency hopping signal.

[0071] Figure 5 for Figure 4The marking result of each frequency hopping signal segment is obtained after the synchronous networking signal is processed in step S4. Figure 5 Available, Figure 4 All fixed frequency interference, swept frequency interference and short burst interference are removed. Figure 5 Each frequency hopping band is correctly marked.

[0072] In this embodiment, the specific process of step S5 is as follows:

[0073] Calculate the time-frequency energy mean in each frequency hopping signal segment, that is, find the energy mean of each time-frequency point in the connected domain and store it in the array Energy in the order of the label. Since the energy of the edge points of each connected domain is distorted after the morphological processing in step S3, only the energy mean of the time-frequency points with energy in the top 80% is taken at this time.

[0074] MeanShift clustering is performed based on the energy mean of each segment, and the frequency hopping signals belonging to each frequency hopping signal segment are sorted out based on the clustering results. The MeanShift clustering algorithm, also known as the mean shift clustering algorithm, is an iterative algorithm based on a sliding window. Its basic idea is to add a sliding window to the samples, find the direction in which the sample points in the window are most densely distributed, and drift the sliding window to the point where the sample density is maximum, and finally converge to the point where the local density is maximum. It is believed that the sample points that converge to the same maximum density belong to the same cluster.

[0075] The specific steps of the algorithm are as follows:

[0076] 1) Determine the sliding window radius r and randomly select an unlabeled data point as the center of the sphere C i .

[0077] 2) Find all points within radius r from the center of the sphere, and record them as set S r , calculate the mean shift vector within the sliding window:

[0078]

[0079] In the formula, x i , x are the coordinates of sample point i and the center of the sphere; k represents the point that falls into S r The number of points in the area; S r is a high-dimensional spherical region with radius r.

[0080] 3) Move the center of the sphere along the direction of the mean shift vector:

[0081]

[0082] The above formula is actually the sum of the values ​​falling into the area S r Find the average vector of the k sample points relative to the point x offset vector.

[0083] 4) Repeat steps 2) and 3) until M r (x) is less than a given threshold.

[0084] 5) If the distance between the current cluster center and other existing cluster centers is less than the threshold, the two classes are merged; otherwise, the current cluster is added as a new class.

[0085] 6) Repeat steps 1) to 5) until all sample points are marked.

[0086] 7) Classification: Count the frequency of each class visiting the point set, and take the class with the highest access frequency as the class to which the current point set belongs.

[0087] Different from the commonly used K-means clustering algorithm, MeanShift clustering does not require a preset number of clusters, and can adaptively obtain the optimal number of clusters, with less interference from outliers. As can be seen from formula (6), all points in the center of the sphere are r The weights of (x) are consistent, but for clustering requirements, the farther the points in the class are from the center point, the worse the clustering effect will be. Therefore, the farther away from the center point, the smaller the weight should be. Therefore, this paper adopts the improved MeanShift clustering algorithm and introduces the Gaussian kernel function when calculating the mean shift vector:

[0088]

[0089] Substituting (8) into (6) we get:

[0090]

[0091] Where G(x) is the unit Gaussian kernel function.

[0092] Figure 6 for Figure 5 The clustering result diagram after MeanShift energy clustering of the synchronous networking signal. Numbers 1, 4, 6, 7, 10, 12, 13, and 16 are frequency hopping signals 1, and numbers 2, 3, 5, 8, 9, 11, 14, and 15 are frequency hopping signals 2. The clustering results are stable.

[0093] According to the MeanShift energy clustering results, the frequency hopping signals are sorted, TF3 is assigned to TF3_temp, and starting from the first cluster, the time-frequency points corresponding to the connected domain in the first cluster are retained in TF3_temp, and the remaining points are set to 0, and the first frequency hopping signal is sorted out. Then, the remaining frequency hopping signals are sorted out according to the above steps.

[0094] Figure 7 To sort out the time-frequency diagram of frequency hopping signal 1, Figure 8This is a time-frequency diagram for sorting out frequency hopping signal 2. As can be seen from the figure, under low signal-to-noise ratio, the algorithm of the present invention can successfully sort out the frequency hopping signals of synchronous networking.

[0095] This paper adopts the MeanShift energy clustering algorithm, which does not rely on the duration of frequency hopping for sorting, and can also complete the sorting of frequency hopping signals in asynchronous networking.

[0096] Fig. 9 This is a grayscale time-frequency diagram of an asynchronous networking aliasing signal collected within the observation time of the present invention. It can be seen that the diagram contains the frequency hopping signals 3 and 4 and various bursts and background noises.

[0097] Fig.10 To sort out the time-frequency diagram of the frequency hopping signal 3, Fig.11 This is a time-frequency diagram for sorting out the frequency hopping signal 4. As can be seen from the figure, under low signal-to-noise ratio, the algorithm of the present invention can successfully sort the frequency hopping signal of asynchronous networking.

[0098] In order to analyze the sorting accuracy of frequency hopping signals under different SNRs in different networking modes, the anti-interference performance of the algorithm is evaluated. The SNR range is selected as -5dB~10dB, and 1000 Monte Carlo simulation experiments are performed at each signal-to-noise ratio. The sorting is successful when the time-frequency distribution of each frequency hopping signal is consistent with the original frequency hopping signal. The number of successful sorting in 1000 Monte Carlo simulation experiments divided by the total number is the sorting accuracy under the current SNR.

[0099] Fig.12 This is a graph showing the accuracy of the synchronous networking frequency hopping signal sorting implemented by the present invention as a function of SNR. The signal sorting accuracy increases as the SNR increases. When SNR≥0dB, the accuracy of the synchronous networking signal sorting implemented by the present algorithm can reach more than 99%.

[0100] Fig.13 This is a graph showing the accuracy of the asynchronous networking frequency hopping signal sorting implemented by the present invention as a function of SNR. The signal sorting accuracy increases as the SNR increases. When SNR ≥ 0dB, the sorting accuracy of the algorithm in this paper for asynchronous networking signals can reach more than 99%. The algorithm in this paper adopts MeanShift energy clustering, which does not require estimation of the number of signals and can automatically cluster the energy of multiple frequency hopping signals. It has a wider applicability and has a higher sorting accuracy at low SNRs than other clustering sorting algorithms.

Claims

1. A method for sorting frequency hopping signals of multiple networks in a complex environment, characterized in that: The method comprises: S1: Establish a mathematical model of multi-network frequency hopping signals in a frequency hopping communication system under complex environments, and perform short-time Fourier transform on the multi-network mixed frequency hopping signals to obtain the time-frequency energy matrix; S2: Adaptive threshold denoising based on energy distribution histogram is performed on the received signal according to its time-frequency energy matrix; S3: Perform morphological filtering to remove frequency sweep interference and some short burst interference; S4: Mark the connected domains and count the duration of each connected domain, remove short burst interference and fixed frequency interference according to the duration, and obtain a pure signal containing only frequency hopping signals; S5: extract the average energy of each segment of the pure signal, and perform MeanShift clustering according to the average energy of each segment to sort each frequency hopping segment; The step S5 specifically includes: Calculate the energy mean of each time-frequency point in each frequency-hopping signal segment. Because the energy of the edge points of each connected domain is distorted after the morphological processing in the third step, only the energy mean of the time-frequency points with energy in the top 80% is taken at this time, and MeanShift clustering is performed based on the energy mean of each segment. The frequency-hopping signals belonging to each frequency-hopping signal segment are selected based on the clustering results.

2. The method for sorting multi-network frequency hopping signals under complex environment according to claim 1, characterized in that: The step S1 specifically includes: A mathematical model of multi-station frequency hopping signals in a frequency hopping communication system under complex environment is established. The aliased signal received by the receiving detector includes not only multiple station frequency hopping signals, but also multiple fixed-frequency interference signals, swept-frequency interference signals, burst interference signals and background noise. The received aliased signal is short-time Fourier transformed using time-frequency analysis technology to obtain its time-frequency energy matrix.

3. The method for sorting frequency hopping signals of multiple networks in a complex environment according to claim 1, characterized in that: The step S2 specifically includes: The logarithmic results of the energy size of each time-frequency point in the time-frequency energy matrix are statistically analyzed to draw a time-frequency energy distribution histogram. The time-frequency energy size corresponding to the horizontal coordinate of the trough on the left between the two peaks is taken as the adaptive denoising threshold. The time-frequency points in the time-frequency energy matrix whose energy is less than the denoising threshold are assigned zero, thereby completing denoising and obtaining the denoised time-frequency energy matrix.

4. The method for sorting multi-network frequency hopping signals in a complex environment according to claim 1, characterized in that: The step S3 specifically includes: The denoised time-frequency energy matrix is ​​subjected to morphological closing operation to fill the gaps in the frequency hopping signal HOP that may be caused by the threshold denoising process. Then, the entire matrix is ​​subjected to morphological erosion operation using shorter structural elements to remove sweep frequency interference and some short burst interference.

5. The method for sorting frequency hopping signals of multiple networks in a complex environment according to claim 1, characterized in that: The step S4 specifically includes: The connected domain marking algorithm based on the stack idea is used to mark the connected domain of the time-frequency energy matrix, and the duration of each time-frequency signal segment is determined. The fixed-frequency interference and burst interference are removed according to the duration, and a pure signal containing only the frequency hopping signal is obtained.

Images