Unmanned aerial vehicle frequency hopping signal sorting and identification method based on frequency spectrum characteristics

Abstract

The invention discloses an unmanned aerial vehicle frequency hopping signal sorting and identification method based on spectrum characteristics, and the method comprises the following steps: carrying out the short-time Fourier transform of a signal received by a receiver, and obtaining a time-frequency matrix; eliminating fixed-frequency signal interference by adopting an energy cancellation method to obtain a time-frequency matrix; partial background noise is processed and eliminated through a threshold value method, and a binary matrix is obtained; removing interference by adopting a morphological filtering method to obtain a binary matrix; estimating parameters of the unmanned aerial vehicle by adopting a time-frequency analysis method to obtain a frequency hopping signal parameter set; matching the frequency hopping signal parameter set with features in an unmanned aerial vehicle signal feature library to obtain a pre-classification cluster; performing frequency hopping period estimation on the frequency hopping moment sequence in the pre-classification clusters by adopting a repetition period transformation method to obtain the frequency hopping period of each pre-classification cluster; adopting a sequence signal retrieval algorithm to screen out a frequency hopping sequence conforming to a frequency hopping period estimation value in the pre-classification cluster to obtain a main classification cluster; and matching the parameter set of the main classification cluster with the features in the unmanned aerial vehicle feature library and a frequency hopping signal number threshold, and if matching succeeds, identifying a specific unmanned aerial vehicle model. The method can solve the problems that interference cannot be effectively eliminated and multi-unmanned aerial vehicle frequency hopping signals cannot be sorted in a limited technology.

Description

Technical Field

[0001] The present invention relates to the technical field of UAV frequency-hopping signal sorting and recognition, and particularly to a method for sorting and recognizing UAV frequency-hopping signals based on spectral features. Background Art

[0002] Frequency-hopping communication is one of the commonly used spread-spectrum communication methods, which has high anti-interception, anti-jamming, and anti-multipath capabilities and is widely used in the military field. In the process of communication countermeasures, accurately identifying frequency-hopping signals is a prerequisite for implementing effective communication jamming. However, in a complex electromagnetic environment, frequency-hopping signals are easily interfered by fixed-frequency signals, burst signals, and noise, which makes the sorting and recognition of frequency-hopping signals difficult.

[0003] The current technology cannot effectively eliminate fixed-frequency signals, burst signals, swept-frequency signals, and noise, and cannot sort and recognize multi-UAV frequency-hopping signals. Summary of the Invention

[0004] Object of the Invention: The present invention provides a method for sorting and recognizing UAV frequency-hopping signals based on spectral features to solve the problems in the existing technology that interference cannot be effectively eliminated and multi-UAV frequency-hopping signals cannot be sorted.

[0005] Technical Solution: A method for sorting and recognizing UAV frequency-hopping signals based on spectral features according to the present invention includes the following steps:

[0006] Step 1: Perform short-time Fourier transform on the signals received by the receiver to obtain a time-frequency matrix;

[0007] Step 2: Use the method of energy cancellation to eliminate the interference of fixed-frequency signals and obtain a time-frequency matrix;

[0008] Step 3: Process and eliminate part of the background noise through the threshold method to obtain a binary matrix;

[0009] Step 4: Use the method of morphological filtering to remove swept-frequency interference, burst interference, and noise interference and obtain a binary matrix;

[0010] Step 5: Use the method of time-frequency analysis to estimate the time width, bandwidth, hopping time, and hopping frequency of UAV frequency-hopping signals to obtain a set of hopping signal parameters;

[0011] Step 6: Match the set of hopping signal parameters with the time width, bandwidth, and hopping frequency in the UAV signal feature library to obtain a pre-classification cluster;

[0012] Step 7: Use the repeated period transformation method to estimate the hopping period of the hopping time sequence in the pre-classification cluster to obtain the hopping period of each pre-classification cluster;

[0013] Step 8: Use the serial signal retrieval algorithm to screen out the frequency hopping sequences in the pre-classified clusters that meet the estimated frequency hopping period value, and obtain the main classified clusters;

[0014] Step 9: Match the parameter set of the main classified cluster with the time width, bandwidth, hopping frequency, hopping period, and number threshold of hopping signals in the UAV feature library. If the match is successful, the specific UAV model is identified.

[0015] Further, in Step 1, the short-time Fourier transform is The UAV hopping signal is x(t), and the time-frequency matrix is TFR(f,t) = |STFT(f,t)| 2 .

[0016] Further, in Step 2, the energy cancellation method is to subtract the mean value of each frequency component from the time-frequency matrix to obtain the time-frequency matrix after cancellation K is the number of points in the frequency dimension of the time-frequency matrix, and N is the number of points in the time dimension of the time-frequency matrix.

[0017] Further, in Step 3, the threshold method is to judge by setting the threshold of each frequency component. When the value of the time-frequency matrix is greater than the threshold, it is judged as useful data, and if it is less, it is filtered out. Specifically,

[0018]

[0019] Thresold(k) = (max(TFR1(k,:)) + min(TFR1(k,:))) / 2

[0020] Further, in Step 4, the morphological filtering method is to perform opening and closing operations on each frequency component of the time-frequency matrix TFR2(k,n). Perform the opening operation on each frequency component of the time-frequency matrix TFR2(k,n) b f is the structuring element; perform the closing operation on each frequency component of the time-frequency matrix TFR3(k,n), TFR4(k,n) = TFR3(k,n)·b f , k = 1, …, K, b f is the structuring element.

[0021] Further, in Step 5, use the time-frequency analysis method to estimate the time width, bandwidth, hopping time, and hopping frequency of the UAV hopping signal, and obtain the hopping signal parameter set. The specific steps are as follows:

[0022] Step 51: Extract the time-frequency ridge line V f (k) of TFR4(k,n), that is, the maximum time corresponding to each frequency point. The extraction method is

[0023] Step 52: For Vf (k) Perform differential operation to obtain the difference sequence d f (k). Locate the positions of 1 and -1 in the difference sequence to obtain the start positions FreqStartPos(h), h = 1, … H and end positions FreqEndPos(h), h = 1, … H of the frequency dimension of the frequency-hopping signal. Calculate the bandwidth B(h) = (FreqEndPos(h) - FreqStartPos(h)) × Δf, h = 1, … H, and the hopping frequency Freq(h) = (FreqEndPos(h) + FreqStartPos(h)) / 2 × Δf, h = 1, … H, where Δf is the frequency resolution and H is the number of frequency-hopping signals in the frequency dimension;

[0024] Step 53: Extract the time-frequency ridge line V t (n) of TFR4(FreqStartPos(h):FreqEndPos(h), n), that is, the maximum frequency corresponding to each time point. The extraction method is

[0025] where h = 1, … H and H is the number of frequency-hopping signals in the frequency dimension.

[0026] Step 54: Perform differential operation on V t (n) to obtain the difference sequence d t (n). Locate the positions of 1 and -1 in the difference sequence to obtain the start positions TimeStartPos(l), l = 1, … L and end positions TimeStartPos(l), l =, … L of the time dimension of the frequency-hopping signal. Calculate the time width TP(l) = (TimeEndPos(l) - TimeStartPos(l)) × Δt, l = 1, … L, and the start time of frequency hopping TimeStart(l) = TimeStartPos(l) × Δt, l = 1, … L, where L is the number of frequency-hopping signals in the time dimension and Δt is the time resolution;

[0027] Step 55: Combine the bandwidth, time width, hopping frequency, and start time of frequency hopping to form the frequency-hopping signal parameter set Feature = [B(i), Freq(i), TP(i), TimeStart(i)], i = 1, … I, where I is the number of frequency-hopping signals.

[0028] Furthermore, in Step 6, the frequency-hopping signal parameter set Feature = [B(i), Freq(i), TP(i), TimeStart(i)], and the UAV feature library Feature uav =[B min ,B max,Freq min ,Freq max ,TP min ,TP max ,P min ,P max , the pre-classification cluster of the frequency-hopping signal Feature = [B(i'), Freq(i'), TP(i'), t(i'), Label], i = 1, … I, i' = 1, … I1, I is the number of frequency-hopping signals, I1 is the number of frequency-hopping signals successfully matched in the database, and Label is the pre-classified UAV label.

[0029] Further, in step 7, the repeated period transformation method is used to estimate the frequency-hopping period of the frequency-hopping time series in the pre-classification cluster, and the specific steps for obtaining the frequency-hopping period of each pre-classification cluster are as follows:

[0030] Step 71: Set [τ min , τ max as the frequency-hopping period range to be studied, and divide this range into K uniform small intervals. Then the k-th interval is The width b of each interval = (τ max - τ min ) / K;

[0031] Step 72: Calculate N is the number of frequency-hopping time series;

[0032] Step 73: Obtain the index k corresponding to the maximum value that satisfies , and τ k is the estimated repeated period.

[0033] Further, in step 8, the sequence signal retrieval algorithm is used to screen out the frequency-hopping sequences that meet the frequency-hopping period estimated value in the pre-classification cluster to obtain the main classification cluster, and the specific steps are as follows:

[0034] Step 81: Set the frequency-hopping time series to be searched {t1, t2, t3, … t n}, the estimated frequency-hopping period is P, initialize its actual frequency-hopping time subscript start = 1, the maximum jitter rate a, and the minimum number of retrieved frequency-hopping times pmin;

[0035] Step 82: Select the frequency-hopping time closest to its frequency-hopping time value t among all the frequency-hopping times within the search jitter range of the start-th frequency-hopping time as the reference frequency-hopping time m;

[0036] Step 83: Search for the hopping time at a distance of P from m with P as the metric. Select the hopping time closest to the P value within the 2a jitter range as the reference hopping time m. Then continue to search to the right for the next interval of P. As long as a hopping time is found, change the reference hopping time m, and so on. If there is no suitable hopping time at P, execute Step 84;

[0037] Step 84: Record that the number of consecutive missing hopping times missing is incremented by 1, and search to the right at an interval of 2 times P. If no hopping time is selected after continuously searching for P at an interval of missing_max + 1, it is determined that the sequence search for the start-th hopping time from sitting up is completed;

[0038] Step 85: Judge the selected hopping times. If the length is less than pmin, return the searched hopping time values and the hopping time sequence to t and num_t respectively, increment start by 1, and execute Step S802. If the length is sufficient for pmin, continue to retrieve the remaining hopping time sequence until the number of hopping times to the right of start is less than pmin, then end the search and form the main classification cluster parameter set Feature = [B(n), Freq(n), TP(n), t(n), P], where n is the number of hopping sequences with P as the hopping period.

[0039] Further, in Step 9, the main classification cluster parameter set of the hopping signal is Feature = [B(n), Freq(n), TP(n), t(n), P], and the UAV feature library is Feature uav =[B min ,B max ,Freq min ,Freq max ,TP min ,TP max ,P min ,P max ,ThresoldNum].

[0040] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: The present invention can realize blind detection of UAV hopping signals in non-cooperative and low signal-to-noise ratio environments, effectively estimate the parameters of UAV hopping signals, and has good robustness, and is applicable to sorting and recognition of UAV hopping signals of multiple targets. Description of the Drawings

[0041] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0042] Figure 2 It is the time-frequency diagram after short-time Fourier transform of the present invention.

[0043] Figure 3This is the time-frequency diagram after energy cancellation of the present invention.

[0044] Figure 4 This is the time-frequency diagram after threshold processing of the present invention.

[0045] Figure 5 This is the time-frequency diagram after morphological filtering of the present invention.

[0046] Figure 6 This is the time-frequency diagram of the Frsky drone sorted and identified by the present invention.

[0047] Figure 7 This is the time-frequency diagram of the Fimi drone sorted and identified by the present invention. Detailed implementation manners

[0048] As Figure 1 shown, a method for sorting and identifying UAV hopping signals based on spectral features includes the following steps:

[0049] Step 1: Perform short-time Fourier transform on the signal received by the receiver to obtain a time-frequency matrix;

[0050] Step 2: Use the method of energy cancellation to eliminate the interference of fixed-frequency signals and obtain a time-frequency matrix;

[0051] Step 3: Process through the threshold method to eliminate part of the background noise and obtain a binary matrix;

[0052] Step 4: Use the method of morphological filtering to remove sweep interference, burst interference and noise interference and obtain a binary matrix;

[0053] Step 5: Use the method of time-frequency analysis to estimate the time width, bandwidth, hopping moment and hopping frequency of the UAV hopping signal to obtain a hopping signal parameter set;

[0054] Step 6: Match the hopping signal parameter set with the time width, bandwidth and hopping frequency in the UAV signal feature library to obtain a pre-classification cluster;

[0055] Step 7: Use the repeated period transformation method to estimate the hopping period of the hopping moment sequence in the pre-classification cluster to obtain the hopping period of each pre-classification cluster;

[0056] Step 8: Use the sequence signal retrieval algorithm to screen out the hopping sequences that meet the hopping period estimated value in the pre-classification cluster to obtain the main classification cluster;

[0057] Step 9: Match the parameter set of the main classification cluster with the time width, bandwidth, hopping frequency, hopping period and the threshold of the number of hopping signals in the UAV feature library. If the match is successful, the specific UAV model is identified.

[0058] The intermediate frequency point sampled by the receiver is 2.45 GHZ, and the sampling rate fs is 122.88 MHz; the hopping period of the Fimi drone is 14 ms, the time width of the hopping signal is 2.24 ms, the signal bandwidth is 1.1 MHz, and the hopping range is from 2400 to 2480 MHz. The hopping period of the Frsky racer is 9 ms, the time width of the hopping signal is 4.4 ms, the signal bandwidth is 0.2 MHz, and the hopping range is from 2400 to 2480 MHz.

[0059] Step S1: Turn on the Fimi drone and the Frsky racer, use the remote control to control the Fimi drone and the Elrs racer to fly respectively, and use the receiver to receive the air interface signal.

[0060] Step S2: Perform a short-time Fourier transform on the 200 ms data after down-conversion and perform modulus-squaring processing to obtain the time-frequency matrix TFR(k,n), where k = 1,…4096 and n = 1,…3000, as Figure 2 shown. There are hopping signals of Fimi and Frsky drones, as well as fixed-frequency interference, noise interference, and burst interference in the time-frequency diagram.

[0061] Step S3: Subtract the time-frequency matrix from the mean value of each frequency component to obtain the time-frequency matrix after cancellation As Figure 3 shown, it can be observed that the fixed-frequency interference in the time-frequency diagram has been filtered out.

[0062] Step S4: Set the threshold for each frequency component to make a judgment. If the value of the time-frequency matrix is greater than the threshold, it is judged as useful data; if it is less, it is filtered out. The specific method is as follows:

[0063]

[0064] Thresold(k) = (max(TFR1(k,:)) + min(TFR1(k,:))) / 2

[0065] After threshold processing, as Figure 4 shown, from Figure 4 it can be observed that some noise interference in the time-frequency diagram has been filtered out.

[0066] Step S5: Perform morphological opening and closing operations on the time-frequency matrix respectively. b f is a 1×7 linear structuring element, and the filtering method is as follows:

[0067]

[0068] TFR4(k,n) = TFR3(k,n)·b f , k = 1,…,K

[0069] After morphological filtering, as Figure 5As shown, it can be observed that most of the noise interference and burst interference in the time-frequency diagram have been filtered out, leaving only the UAV frequency-hopping signal.

[0070] Step S6: Use the time-frequency analysis method to estimate the time width, bandwidth, frequency-hopping moment, and frequency-hopping frequency of the UAV frequency-hopping signal to obtain the frequency-hopping signal parameter set, including the following sub-steps:

[0071] Step S601: Extract the time-frequency ridge line V f (k) of TFR4(k,n), that is, the maximum time corresponding to each frequency point. The extraction method is

[0072] Step S602: Perform a difference operation on V f (k) to obtain the difference sequence d f (k). Find the positions of 1 and -1 in the difference sequence to obtain the start position FreqStartPos(h) of the frequency dimension of the frequency-hopping signal, h = 1, … 33, and the end position FreqEndPos(h) of the frequency dimension, h = 1, … 33. Calculate the bandwidth B(h) = (FreqEndPos(h) - FreqStartPos(h)) × Δf, h = 1, … 33, and the frequency-hopping frequency Freq(h) = (FreqEndPos(h) + FreqStartPos(h)) / 2 × Δf, h = 1, … H, where Δf = 0.003 MHz.

[0073] Step S603: Extract the time-frequency ridge line V t (n) of TFR4(FreqStartPos(h):FreqEndPos(h),n), that is, the maximum frequency corresponding to each time point. The extraction method is

[0074]

[0075] Step S604: Perform a difference operation on V t (n) to obtain the difference sequence d t (n). Find the positions of 1 and -1 in the difference sequence to obtain the start position TimeStartPos(l) of the time dimension of the frequency-hopping signal, l = 1, … 83, and the end position TimeEndPos(l), n = 1, … 83. Calculate the time width TP(l) = (TimeEndPos(l) - TimeStartPos(l)) × Δt, l = 1, … 83, and the frequency-hopping start moment TimeStart(l) = TimeStartPos(l) × Δt, l = 1, … 83, where Δt = 6.67×10 -5 s is the time resolution.

[0076] Step S605: Compose a frequency hopping signal parameter set Feature = [B(h), Freq(h), TP(h), TimeStart(h)], where h = 1, …, 83, with the bandwidth, time width, frequency hopping frequency, and frequency hopping start time.

[0077] Step S7: Match the frequency hopping parameter set with the minimum and maximum time widths, minimum and maximum bandwidths, and minimum and maximum frequency hopping frequencies in the UAV signal feature library. After successful matching, obtain four pre-classified clusters of frequency hopping signals:

[0078] Feature1 = [B(h1), Freq(h1), TP(h1), t(h1), 'Frsky'], where h1 = 1, …, 12, Feature2 = [B(h2), Freq(h2), TP(h2), t(h2), 'Fimi'],

[0079] h1 = 1, …, 9, Feature3 = [B(h3), Freq(h3), TP(h3), t(h3), 'Dsmx'], where h3 = 1, …, 2,

[0080] Feature4 = [B(h4), Freq(h4), TP(h4), t(h4), 'Futaba'], where h4 = 1, …, 7

[0081] Step S8: Estimate the frequency hopping period for the frequency hopping time series of each pre-classified cluster using the frequency hopping period transformation method, including the following sub-steps:

[0082] Step S801: [0.0015s, 0.012s] is the frequency hopping period range to be studied. Divide this range into K = 500 uniform small intervals, then the k-th interval is The width b of each interval is 6.53×10 -7 .

[0083] Step S802: Calculate

[0084] Step S803: Obtain the index k corresponding to the maximum value that satisfies τ, and the maximum value is the estimated repetition period, obtaining P1 = 13.99ms, P2 = 9ms, where T = 10.5ms and α = 0.7. k That is, the estimated repetition period, obtaining P1 = 13.99ms, P2 = 9ms, where T = 10.5ms and α = 0.7.

[0085] Step S9: Using the sequence signal retrieval algorithm, screen out the hopping signal parameter sets that meet the hopping period estimated value among all the hopping signals, and obtain two main classification clusters of hopping signals Feature1 = [B(n1), Freq(n1), TP(n1), t(n1), 13.99], Feature2 = [B(n2), Freq(n2), TP(n2), t(n2), 9], where n1 = 1, … 9, n2 = 1, … 12. Compare the main classification clusters of hopping signals Feature1 = [B(n1), Freq(n1), TP(n1), t(n1), 13.99] and Feature2 = [B(n2), Freq(n2), TP(n2), t(n2), 9] with the UAV feature library:

[0086] Feature Frsky = [0.1, 0.3, 2400, 2480, 4.18, 4.62, 8.55, 9.45, 6] Feature Fimi = [0.9, 1.3, 2400, 2480, 2.128, 2.352, 13.3, 14.7, 3] for matching. If the matching is successful, Frsky and Fimi are identified. As Figure 6 , 7 shown, they are the time-frequency diagrams of the selected Frsky UAV and the selected Fimi UAV respectively. The colored parts in the figures indicate the positions of the hopping signals of the selected UAVs.

Claims

1. A method for sorting and identifying frequency-hopping signals of unmanned aerial vehicles based on spectral features, characterized in that, It includes the following steps: Step 1: Perform short-time Fourier transform on the signal received by the receiver to obtain a time-frequency matrix; Step 2: Use the method of energy cancellation to eliminate the interference of fixed-frequency signals and obtain a time-frequency matrix; Step 3: Eliminate part of the background noise through threshold processing to obtain a binary matrix; Step 4: Use the method of morphological filtering to remove swept-frequency interference, burst interference, and noise interference to obtain a binary matrix; Step 5: Use the method of time-frequency analysis to estimate the time width, bandwidth, hopping time, and hopping frequency of the UAV hopping signal to obtain a hopping signal parameter set; Step 6: Match the hopping signal parameter set with the time width, bandwidth, and hopping frequency in the UAV signal feature library to obtain a pre-classification cluster; Step 7: Use the repeated period transformation method to estimate the hopping period of the hopping time sequence in the pre-classification cluster to obtain the hopping period of each pre-classification cluster; Step 8: Use the sequence signal retrieval algorithm to screen out the hopping sequences that meet the hopping period estimation value in the pre-classification cluster to obtain the main classification cluster; Step 9: Match the parameter set of the main classification cluster with the time width, bandwidth, hopping frequency, hopping period, and the threshold of the number of hopping signals in the UAV feature library. If the match is successful, the specific UAV model is identified.

2. The method for sorting and identifying the frequency hopping signals of the unmanned aerial vehicle based on the spectral characteristics according to claim 1, wherein In Step 1, the short-time Fourier transform is The hopping signal of the UAV is x(t), and the time-frequency matrix is TFR(f,t) = |STFT(f,t)| 2 .

3. The method for sorting and identifying the frequency-hopping signals of the unmanned aerial vehicle based on spectral features according to claim 1, wherein In Step 2, the energy cancellation method is to subtract the mean value of each frequency component from the time-frequency matrix to obtain the time-frequency matrix after cancellation. K is the number of points in the frequency dimension of the time-frequency matrix, and N is the number of points in the time dimension of the time-frequency matrix.

4. The method for sorting and identifying the frequency-hopping signals of the unmanned aerial vehicle based on spectral features according to claim 1, wherein, In Step 3, the threshold method judges by setting the threshold of each frequency component. When the value of the time-frequency matrix is greater than the threshold, it is judged as useful data, and if it is less, it is filtered out. Specifically, Thresold(k)=(max(TFR1(k,:))+min(TFR1(k,:))) / 2 5. The method for sorting and identifying the frequency-hopping signals of the unmanned aerial vehicle based on spectral features according to claim 1, wherein In step 4, the morphological filtering method performs opening and closing operations on each frequency component of the time-frequency matrix TFR2(k,n), and performs the opening operation on each frequency component of the time-frequency matrix TFR2(k,n). b f is the structuring element when performing the closing operation on each frequency component of the time-frequency matrix TFR3(k,n), TFR4(k,n) = TFR3(k,n)·b f , k = 1, …, K, b f is the structuring element when performing the closing operation on each frequency component of the time-frequency matrix TFR3(k,n).

6. The method for sorting and identifying the FHSS signals of the UAV based on spectral features according to claim 1, characterized in that, In Step 5, using the method of time-frequency analysis to estimate the time width, bandwidth, hopping time, and hopping frequency of the UAV hopping signal to obtain a hopping signal parameter set specifically includes the following steps: Step 51: Extract the time-frequency ridge line V(k) of TFR4(k,n), that is, the maximum time corresponding to each frequency point. The extraction method is f (k), Step 52. Perform a difference operation on V f (k) to obtain a difference sequence d f (k). Locate the positions of 1 and -1 in the difference sequence to obtain the start positions FreqStartPos(h) of the frequency dimension of the frequency-hopping signal, where h = 1, …, H, and the end positions FreqEndPos(h) of the frequency dimension, where h = 1, …, H. Calculate the bandwidth B(h) = (FreqEndPos(h) - FreqStartPos(h)) × Δf, where h = 1, …, H, and the hopping frequency Freq(h) = (FreqEndPos(h) + FreqStartPos(h)) / 2 × Δf, where h = 1, …, H. Δf is the frequency resolution, and H is the number of frequency-hopping signals in the frequency dimension. Step 53: Extract the time-frequency ridge line V of TFR4(FreqStartPos(h):FreqEndPos(h), n), that is, the maximum frequency corresponding to each time point. The extraction method is t (n), and the extraction method is k = FreqStartPos(h), …, FreqEndPos(h), n = 1, … N, where h = 1, … H, and H is the number of frequency-hopping signals in the frequency dimension; Step 54, perform a difference operation on V t (n) to obtain a difference sequence d t (n). Locate the positions of 1 and -1 in the difference sequence to obtain the start positions TimeStartPos(l), l = 1, …, L and end positions TimeEndPos(l), l = 1, …, L of the frequency-hopping signal in the time dimension. Calculate the time width TP(l) = (TimeEndPos(l) - TimeStartPos(l)) × Δt, l = 1, …, L, and the start time of frequency hopping TimeStart(l) = TimeStartPos(l) × Δt, l = 1, …, L, where L is the number of frequency-hopping signals in the time dimension and Δt is the time resolution; Step 55: Combine the bandwidth, time width, hopping frequency, and hopping start time to form a hopping signal parameter set Feature=[B(i),Freq(i),TP(i),TimeStart(i)], i = 1,……I, where I is the number of hopping frequency signals.

7. The method for sorting and identifying UAV frequency hopping signals based on spectral features according to claim 1, characterized in that, In step 6, the frequency hopping signal parameter set Feature = [B(i), Freq(i), TP(i), TimeStart(i)], and the UAV feature library Feature uav = [B min , B max , Freq min , Freq max , TP min , TP max , P min , P max . The pre-classification cluster of the frequency hopping signal Feature = [B(i'), Freq(i'), TP(i'), t(i'), Label], where i = 1, …, I, i' = 1, …, I1, I is the number of frequency hopping signals, I1 is the number of frequency hopping signals successfully matched in the database, and Label is the UAV label of the pre-classification.

8. The method for sorting and identifying the frequency-hopping signals of an unmanned aerial vehicle based on spectral features according to claim 1, wherein In Step 7, using the repeated period transformation method to estimate the hopping period of the hopping time sequence in the pre-classification cluster to obtain the hopping period of each pre-classification cluster specifically includes the following steps: Step 71, set [τ min , τ max as the hopping period range to be studied, divide this range into K uniform small intervals, then the k-th interval is The width b of each interval is b = (τ max - τ min ) / K; Step 72, calculate N is the number of hopping time sequence Step 73, obtain the index k corresponding to the maximum value of , and τ k is the estimated repetition period.

9. The method for sorting and identifying UAV hopping signals based on spectral features according to claim 1, wherein In Step 8, using the sequence signal retrieval algorithm to screen out the hopping sequences that meet the hopping period estimation value in the pre-classification cluster to obtain the main classification cluster specifically includes the following steps: Step 81. Set the hopping time sequence {t1, t2, t3, …, t n} to be searched, estimate the hopping period as P, initialize the starting index start of its hopping time to 1, the maximum jitter rate a, and the minimum number pmin of retrieved hopping times; Step 82: Select the hopping time closest to its hopping time value t among all the hopping times within the search jitter range of the start-th hopping time as the reference hopping time m; Step 83: Search for the hopping time at a distance of P from m with P as the metric. Select the hopping time closest to the P value within the 2a jitter range and use this hopping time as the reference hopping time m, and then continue to search for the next interval of P to the right. As long as a hopping time is found, change the reference hopping time m, and so on; if there is no suitable hopping time at P, then execute Step 84; Step 84: Record that the number of consecutive missing hopping times missing is incremented by 1, search at an interval of 2P to the right. If the P at the interval of missing_max + 1 has not been selected as the hopping time after consecutive searches, it is determined that the sequence search for the start-th hopping time since sitting up is completed; Step 85: Judge the selected hopping times. If the length is less than pmin, return the searched hopping time value and the hopping time sequence to t and num_t respectively, increment start by 1, and execute step S802; if the length is sufficient for pmin, continue to retrieve the remaining hopping time sequence until the number of hopping times to the right of start is less than pmin, then end the search and form the main classification cluster parameter set Feature = [B(n), Freq(n), TP(n), t(n), P], where n is the number of hopping sequences with P as the hopping period.

10. The method for sorting and identifying the frequency hopping signals of the unmanned aerial vehicle based on the spectral features according to claim 1, wherein In step 9, the parameter set of the main classification cluster of the frequency hopping signal is Feature = [B(n), Freq(n), TP(n), t(n), P], and the UAV feature library is Feature uav =[B min ,B max ,Freq min ,Freq max ,TP min ,TP max ,P min ,P max ,ThresoldNum].

Images