Frequency hopping signal detection and sorting method based on signal fingerprint mechanism matching

Abstract

The invention provides a frequency hopping signal detection and sorting method based on signal fingerprint mechanism matching, and relates to the technical field of signal processing. According to the invention, through the analysis and feature extraction of the transmitter fingerprint mechanism, the identification of different transmitter individuals is realized, the individuals to which the signals belong are distinguished, and the frequency hopping signals are effectively detected and sorted; a reconnaissance frequency band is subdivided into a plurality of sub-channels, signal fingerprint features of intermediate frequency data are extracted after existence of signals is determined through energy judgment, feature matching is carried out, and signal detection is completed in combination with signal bandwidth and occurrence time information; according to the invention, the problem that the frequency hopping signal is falsely detected as a normal signal when being continuously transmitted at the same frequency point is solved, and the detection accuracy and speed are improved; by utilizing the uniqueness of the fingerprint characteristics of the transmitter, the problem of sorting errors when multiple groups of frequency hopping signals are overlapped in time or frequency spectrum is solved, accurate distinguishing and detection of the multiple groups of frequency hopping signals are realized, and the overall performance of frequency hopping signal detection is improved.

Description

Technical Field

[0001] The invention relates to the technical field of signal processing, in particular to a frequency hopping signal detection and sorting method based on signal fingerprint mechanism matching. Background Art

[0002] Frequency hopping signals are typical non-stationary signals. The current mainstream frequency hopping detection methods mainly include energy detection method, autocorrelation detection method and time-frequency analysis detection method. Among them, the energy detection algorithm is designed under strict premise, and has some prior information about the frequency hopping signal to be detected. The specific frequency hopping signal is detected according to the energy and prior information of the signal. The autocorrelation detection method performs maximum correlation processing on the frequency hopping signal, and detects a frequency hopping signal with a specific hopping speed according to the threshold of the correlation value in the autocorrelation detection. Frequency hopping detection based on time-frequency analysis can display the time, frequency and energy characteristics of the signal by converting the time domain frequency hopping signal into a two-dimensional signal matrix in the time-frequency domain. The existence of the frequency hopping signal is judged by using some characteristics of the signal, such as based on the edge, the peak value of the signal, the phase, etc. It is widely used in modern frequency hopping communication detection.

[0003] Among the three existing mainstream frequency hopping detection methods, the energy detection algorithm has poor detection performance in blind detection of frequency hopping signals when the parameter information of the acquired signal is less known, and cannot be applied to the detection of frequency hopping signals that are actually lacking the required detection. The autocorrelation detection method requires maximum correlation processing of the frequency hopping signal, so the threshold of the correlation value in the autocorrelation detection is difficult to give, and one correlation can only extract a signal with a specific hopping rate. When there are signals with different hopping rates, it is difficult to effectively detect the correct signal. The time-frequency analysis detection method, in its time-frequency analysis method, the short-time Fourier transform cannot solve the problem of time resolution and frequency resolution, and the processing process of the wavelet transform is computationally intensive. In the presence of constant signal interference and multiple groups of frequency hopping signals, the time-frequency analysis detection method will confuse the identification of constant signals and frequency hopping signals. In addition, the time-frequency analysis detection method requires data accumulation of the signal to be detected for a certain period of time, and the frequency hopping signal is identified according to the distribution characteristics of the signal.

[0004] Therefore, it is necessary to provide a frequency hopping signal detection and sorting method based on signal fingerprint mechanism matching to solve the technical problem that the existing frequency hopping detection method cannot effectively and quickly detect the frequency hopping signal in the presence of interference such as a constant signal. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching, so as to solve the technical problem that the existing frequency hopping detection method cannot effectively and quickly detect the frequency hopping signal in the presence of interference such as a constant signal; the purpose of the present invention is achieved by the following technical solutions:

[0006] The present invention provides a method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching, and performs frequency hopping signal detection through the following steps:

[0007] Step 1: Divide the frequency band to be detected of the frequency hopping signal into sub-channels;

[0008] Step 2: According to the set detection energy threshold, compare with the energy of each sub-channel; if the current sub-channel energy is greater than the detection energy threshold, there is a signal in the current sub-channel; otherwise, there is no signal in the current sub-channel;

[0009] Step 3: Merge each subchannel with a signal; if adjacent subchannels have signals, the bandwidth of the current signal is determined as the bandwidth of the adjacent subchannel;

[0010] Step 4: According to the number of sub-channels where the signal exists obtained in step 3, obtain the intermediate frequency data corresponding to the sub-channels where the signal exists;

[0011] Step 5: Extract fingerprint features through the intermediate frequency data of the sub-channel where the signal exists, and obtain the signal fingerprint features;

[0012] Step 6: According to the sub-channel where the signal exists, the signal bandwidth, signal start and end time and duration of the current signal are calculated to obtain the corresponding signal parameter characteristics;

[0013] Step 7: Merge the signal fingerprint features and signal parameter features extracted in steps 5 and 6 to form a corresponding signal feature vector;

[0014] Step 8: Determine whether the current signal feature vector is the first batch of signal feature vectors in the current detection; if so, directly store the signal feature vector in the library; if not, match the current signal feature vector with the signal feature vector in the fingerprint feature library and calculate the correlation coefficient R F，P ;

[0015] Step 9: Get the matching results of step 8; if the correlation coefficient R F，P If the signal fingerprint feature is greater than the preset correlation threshold, the signal fingerprint feature matching is successful, indicating that the signal transmitted by the same signal transmitter is obtained at different times; otherwise, it means that the signal transmitted by the same transmitter is not received at the current time, and the signal feature vector extracted in step 7 is stored in the database;

[0016] Step 10: According to the result of step 9, if the frequency hopping signals transmitted by multiple signal transmitters are matched, it means that the signal sorting operation is performed on the multiple groups of frequency hopping signals, and the results are displayed in the detection results according to the serial numbers of the matched multiple transmitters.

[0017] As a further solution, in step 1, the bandwidth of the frequency band to be detected is N, and the number of sub-channel divisions is M; then the bandwidth occupied by each sub-channel is B w =N / M, the center frequency of each sub-channel is: i is the current subchannel number.

[0018] As a further solution, after the sub-channels are divided, the sub-channel data is obtained through the following steps:

[0019] Obtain the original intermediate frequency data before channelization, and perform down-conversion operation according to the center frequency of each sub-channel;

[0020]

[0021] Among them, x i is the original intermediate frequency data before channelization, f c is the center frequency of the current subchannel that needs to be channelized, o i is the data after down-conversion, t is the signal time parameter, and j is the imaginary unit;

[0022] The down-converted intermediate frequency data is low-pass filtered according to the bandwidth of each sub-channel to obtain the channelized sub-channel data; the frequency domain expression is:

[0023]

[0024] Y = O(f)·H(f);

[0025] Where, f is the frequency, f b is the cut-off frequency B w / 2, O(f) is the frequency domain expression of the intermediate frequency data after down-conversion, and Y is the frequency domain expression of the sub-channel data after channelization.

[0026] As a further solution, in step 2, the detection energy threshold th is set and compared with the energy En of each sub-channel; wherein,

[0027]

[0028] In the formula, r i represents the energy of each signal sampling point in the subchannel, k represents the number of sampling points in the current subchannel, Channel iRepresents the subchannel signal presence parameter, 1 means the signal exists, 0 means the signal does not exist.

[0029] As a further solution, in step 3, the bandwidth of the subchannels after merging is:

[0030]

[0031] Among them, Bd signal It represents the bandwidth after the sub-channels are merged, kd is the number of adjacent channels, and i is the number of the current sub-channel.

[0032] As a further solution, in step 5, fingerprint feature extraction is performed by the following steps:

[0033] Perform filtering operation on the received intermediate frequency data;

[0034] The filtering operation is completed by acquiring and normalizing the envelope signal through the intermediate frequency data;

[0035] Extract the time domain features of the envelope signal and extract the RJ parameters of the envelope signal;

[0036] Obtain the transient energy envelope curve, extract the curve area, duration, kurtosis, slope, and variance of the transient energy envelope curve, and combine them as characteristic parameters

[0037] Perform time-frequency domain conversion on the intermediate frequency signal to obtain matrix D;

[0038] Get the first k singular values ​​of matrix D as the eigenvalues ​​extracted from the signal fingerprint feature

[0039] Perform feature fusion on the obtained features to obtain the final extracted signal fingerprint features

[0040] As a further solution, the time-frequency domain conversion of the intermediate frequency signal is performed by the following steps:

[0041] Find all local maximum and local minimum points of the original intermediate frequency data;

[0042] Use cubic spline curve to fit the envelope w of upper and lower extreme points max (t) and envelope e min (t);

[0043] Find the average value of the upper and lower envelopes;

[0044]

[0045] Calculate the difference d(t) between the original intermediate frequency data and the average values ​​of the upper and lower envelopes;

[0046] d(t)=x(t)-m(t)

[0047] Among them, x(t) represents the original intermediate frequency data, and m(t) represents the average value of the upper and lower envelopes;

[0048] Determine whether the obtained subtraction difference d(t) satisfies the following IMF conditions:

[0049] Condition 1: The difference between the number of extreme points and the number of zero-crossing points in the decomposed time series does not exceed 1;

[0050] Condition 2: In the analyzed time series, the mean of the extreme envelope fitted by interpolation is 0.

[0051] Repeat the calculation of the difference d(t) until the IMF condition is met. k1 (t) is an IMF component, and so on, we can get several orders of IMF components; where k1 represents the number of repeated calculations; i1 represents the maximum order of the IMF component;

[0052]

[0053] Each time an IMF component is obtained, it is subtracted from the original intermediate frequency data to obtain the residual component r1(t), and then the next-order IMF component is subtracted from the residual component r1(t) until the final residual component r i1 (t) is just a monotone sequence or a constant sequence, and the modal components of each order of the signal are obtained; among them, i1 is the maximum order

[0054]

[0055] Perform HHT transform on each IMF component, that is, perform Hilbert transform on each IMF component:

[0056]

[0057] Then find the instantaneous frequency ω of the corresponding IMF component i1 (t):

[0058] θ i1 (t) = arctan(d i1 (t) / imf i1 (t))

[0059]

[0060] Yes i1The time-frequency matrix W composed of and t is subjected to singular value decomposition; among them,

[0061] SVD(W)=UDV T

[0062] Where SVD(·) is the singular value decomposition function, U is an M×M matrix, D is an M×k matrix, all elements of matrix D are 0 except the elements on the main diagonal, each element on the main diagonal is a singular value, and the size of the singular value decreases from the upper left to the lower right, and V is a k×k matrix.

[0063] As a further solution, in step 6:

[0064] The time when the current sub-channel exceeds the energy threshold for the first time is counted as the signal starting time T start The moment when the current subchannel falls below the energy threshold for the first time after the signal start time is taken as the signal cutoff time T end ;

[0065] Calculate the signal duration T hold =T end -T start ;

[0066] The signal bandwidth is obtained by counting the number of adjacent sub-channels n where the signal appears: BW signal =B w ×n.

[0067] As a further solution, in step 7, the fingerprint features and parameter features extracted in steps 5 and 6 are combined to form a signal feature vector F;

[0068] F=[F signal , B.W. signal , T hold ]

[0069] Among them, F signal Indicates signal fingerprint characteristics, BW signal represents the signal bandwidth, T hold Indicates the signal duration.

[0070] As a further solution, in step 8, the correlation coefficient R is calculated by the following formula F，P :

[0071]

[0072] Among them, F v The vth eigenvalue in the quantity representing the current signal characteristics, represents the mean of the eigenvalues ​​of the current signal eigenvector, Represents the vth eigenvalue in the signal feature vector of the uth individual in the feature library, represents the mean eigenvalue of the u-th individual signal feature vector in the feature, and l+7 represents the dimension of the individual fingerprint feature.

[0073] Compared with the related art, the frequency hopping signal detection and sorting method based on signal fingerprint mechanism matching provided by the present invention has the following advantages:

[0074] 1. Based on the existing technology, the present invention analyzes the fingerprint mechanism of the signal transmitter and extracts the fingerprint features, identifies the transmitter individuals through the fingerprint features of different transmitters, distinguishes the individual to which the current signal belongs in combination with the individual identification information, and realizes the detection and sorting of frequency hopping signals;

[0075] 2. The present invention divides the reconnaissance frequency band into multiple sub-channels to ensure that the bandwidth of each sub-channel is the minimum frequency resolution of the reconnaissance detection. After performing energy judgment on each sub-channel to determine whether there is a signal in the current sub-channel, the intermediate frequency data of the channel with the signal is obtained, and the signal fingerprint feature is extracted and matched on the intermediate frequency data. The feature matching result is used as a main basis for sorting the frequency hopping signal, and then combined with the information of the signal bandwidth and the occurrence time, the frequency hopping signal is detected;

[0076] 3. The present invention solves the problem that the existing detection method detects such signals as constant signals because the existing frequency hopping signals continuously transmit signals at the same frequency point at two adjacent hopping points by introducing the method of extracting and detecting the fingerprint features of the signal transmitter. At the same time, due to the uniqueness of the fingerprint of the signal transmitter, the frequency hopping signal can be quickly detected and identified for the signal to be detected, further improving the detection accuracy and speed of the frequency hopping signal detection;

[0077] 4. The present invention solves the problem that when multiple groups of frequency hopping signals appear simultaneously or crosswise in a time period, the existing frequency hopping signal detection and sorting methods will make mistakes in sorting frequency hopping signals that are continuous in time and belong to different groups at the same frequency point by combining the signal transmitter fingerprint feature extraction and matching method. In addition, since the uniqueness of the transmitter fingerprint feature information is used, different groups of frequency hopping signals that overlap in the time domain or spectrum can be accurately distinguished, thereby realizing the detection and sorting of multiple groups of frequency hopping signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 A schematic diagram of the steps of a frequency hopping signal detection and sorting method based on signal fingerprint mechanism matching provided by the present invention;

[0079] Figure 2 A schematic diagram of a sub-channel data acquisition process provided by the present invention;

[0080] Figure 3 A schematic diagram of the fingerprint feature extraction process provided by the present invention;

[0081] Figure 4 A schematic diagram of the pulse signal envelope curve provided by the present invention;

[0082] Figure 5 A schematic diagram of the signal envelope of an individual radiation source provided by the present invention. DETAILED DESCRIPTION

[0083] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions 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 part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0084] See also Figure 1 The present invention provides a method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching, and the frequency hopping signal detection is performed through the following steps:

[0085] Step 1: Divide the frequency band to be detected of the frequency hopping signal into sub-channels;

[0086] Step 2: According to the set detection energy threshold, compare with the energy of each sub-channel; if the current sub-channel energy is greater than the detection energy threshold, there is a signal in the current sub-channel; otherwise, there is no signal in the current sub-channel;

[0087] Step 3: Merge each subchannel with a signal; if adjacent subchannels have signals, the bandwidth of the current signal is determined as the bandwidth of the adjacent subchannel;

[0088] Step 4: According to the number of sub-channels where the signal exists obtained in step 3, obtain the intermediate frequency data corresponding to the sub-channels where the signal exists;

[0089] Step 5: Extract fingerprint features through the intermediate frequency data of the sub-channel where the signal exists, and obtain the signal fingerprint features;

[0090] Step 6: According to the sub-channel where the signal exists, the signal bandwidth, signal start and end time and duration of the current signal are calculated to obtain the corresponding signal parameter characteristics;

[0091] Step 7: Merge the signal fingerprint features and signal parameter features extracted in steps 5 and 6 to form a corresponding signal feature vector;

[0092] Step 8: Determine whether the current signal feature vector is the first batch of signal feature vectors in the current detection; if so, directly store the signal feature vector in the library; if not, match the current signal feature vector with the signal feature vector in the fingerprint feature library and calculate the correlation coefficient R F，P ;

[0093] Step 9: Get the matching results of step 8; if the correlation coefficient R F，P If the signal fingerprint feature is greater than the preset correlation threshold, the signal fingerprint feature matching is successful, indicating that the signal transmitted by the same signal transmitter is obtained at different times; otherwise, it means that the signal transmitted by the same transmitter is not received at the current time, and the signal feature vector extracted in step 7 is stored in the database;

[0094] Step 10: According to the result of step 9, if the frequency hopping signals transmitted by multiple signal transmitters are matched, it means that the signal sorting operation is performed on the multiple groups of frequency hopping signals, and the results are displayed in the detection results according to the serial numbers of the matched multiple transmitters.

[0095] It should be noted that: when executing step 1, the frequency band width of the frequency band to be detected is N, and the number of sub-channel divisions is M; then the bandwidth occupied by each sub-channel is B w =N / M, the center frequency of each sub-channel is: i is the current subchannel number.

[0096] For example, if the frequency band width is 60 (MHz), the sub-channels are divided into 1000 sub-channels, and the bandwidth occupied by each sub-channel is The center frequency of each subchannel is f c =(i-1)×0.06+0.03(MHz), i=1,...,M.

[0097] The steps for acquiring data after sub-channel division are as follows: Figure 2 As shown,

[0098] Obtain the original intermediate frequency data before channelization, and perform down-conversion operation according to the center frequency of each sub-channel;

[0099]

[0100] Among them, x i is the original intermediate frequency data before channelization, f c is the center frequency of the current subchannel that needs to be channelized, o i is the data after down-conversion, t is the signal time parameter, and j is the imaginary unit;

[0101] The down-converted intermediate frequency data is low-pass filtered according to the bandwidth of each sub-channel to obtain the channelized sub-channel data; the frequency domain expression is:

[0102]

[0103] Y = O(f)·H(f);

[0104] Where, f is the frequency, f b is the cut-off frequency B w / 2, O(f) is the frequency domain expression of the intermediate frequency data after down-conversion, and Y is the frequency domain expression of the sub-channel data after channelization.

[0105] In step 2, the set detection energy threshold th is compared with each sub-channel energy En; if the sub-channel energy is greater than the detection energy threshold th, there is a signal in the sub-channel; otherwise, there is no signal in the sub-channel; wherein,

[0106]

[0107] In the formula, r i represents the energy of each signal sampling point in the subchannel, k represents the number of sampling points in the current subchannel, Channel i Represents the subchannel signal presence parameter, 1 means the signal exists, 0 means the signal does not exist.

[0108] As a further solution, in step 3, the bandwidth of the subchannels after merging is:

[0109] Channel i =Channel i+1 =…=Channel i+k-1 =1

[0110] Among them, Bd signal It represents the bandwidth after the sub-channels are merged, kd is the number of adjacent channels, and i is the number of the current sub-channel.

[0111] Step 4 is to obtain the corresponding intermediate frequency data according to the number of sub-channels of the signal obtained in step 3;

[0112] y=y1+…+y i

[0113] Among them, y i is the number of adjacent sub-channels with signals, and y is the intermediate frequency data of the merged sub-channels.

[0114] The feature extraction process of step 5 is as follows Figure 3 As shown, the process of generating a signal from a transmitter to receiving it by a receiver is expressed as:

[0115] x i (t) = e(d i , m i ), i = 1, ..., K

[0116] r i (t) = g(d i , m i , v), i = 1, ..., K

[0117] Among them, d i represents the intentional modulation information sent by the i-th transmitter, m i The characteristic information of the transmitter module. The characteristics of each module of the transmitter are the result of the combined effects of internal components and the external working environment. It contains information that can be used to identify the individual transmitter, that is, the RF fingerprint information. i With m i Independent of each other, x i (t) represents the signal transmitted by the transmitter, and e(·) represents the process of the transmitter generating the transmitted signal. i (t) represents the received signal, g(·) represents the process of the RF signal generated by the transmitter and transmitted through the channel, v is the noise in the channel, and K is the number of transmitters.

[0118] In step 5, fingerprint feature extraction is performed through the following steps:

[0119] For signal r i (t) Perform filtering operations to reduce noise interference during signal transmission through the channel;

[0120] x = filter(r)

[0121] Acquiring and normalizing the signal envelope of the filtered signal;

[0122]

[0123] Among them, x I I-channel signal representing the intermediate frequency signal, x Q Represents the Q channel signal of the intermediate frequency signal.

[0124] Extract the time domain features of the envelope signal and extract the RJ parameters of the envelope signal;

[0125] The R parameter is the ratio of the variance of the signal envelope to the square of the envelope mean;

[0126]

[0127] The J parameter is the ratio of the difference between the square of the fourth-order moment and the second-order moment of the signal envelope to the square of the envelope mean:

[0128]

[0129] For the transient energy envelope curve of the signal, extract the curve area, duration, kurtosis, slope, and variance of the transient energy envelope curve to obtain the characteristic parameters of the above five characteristic data.

[0130] like Figure 4 The pulse signal envelope curve shown in the figure is the integral S of the transient energy envelope curve in the time period t0 to t1. 0～1 , duration t hold =t1-t0, kurtosis μ4 is the fourth-order central moment of the transient energy envelope curve, σ is the standard deviation of the transient energy envelope curve, and the slope variance k is the number of sampling points in the time period from t0 to t1.

[0131] Perform time-frequency domain conversion on the intermediate frequency signal to obtain matrix D;

[0132] Get the first k singular values ​​of matrix D as the eigenvalues ​​extracted from the signal fingerprint feature

[0133] Perform feature fusion on the obtained features to obtain the final extracted signal fingerprint features

[0134] As a further solution, the time-frequency domain conversion of the intermediate frequency signal is performed by the following steps:

[0135] Find all local maximum and local minimum points of the original intermediate frequency data;

[0136] Use cubic spline curve to fit the envelope of upper and lower extreme points max (t) and envelope e min (t);

[0137] Find the average value of the upper and lower envelopes;

[0138]

[0139] Calculate the difference d(t) between the original intermediate frequency data and the average values ​​of the upper and lower envelopes;

[0140] d(t)=x(t)-m(t)

[0141] Among them, x(t) represents the original intermediate frequency data, and m(t) represents the average value of the upper and lower envelopes;

[0142] Determine whether the obtained subtraction difference d(t) satisfies the following IMF conditions:

[0143] Condition 1: The difference between the number of extreme points and the number of zero-crossing points in the decomposed time series does not exceed 1;

[0144] Condition 2: In the analyzed time series, the mean of the extreme envelope fitted by interpolation is 0.

[0145] Repeat the calculation of the difference d(t) until the IMF condition is met. k1 (t) is an IMF component, and so on, we can get several orders of IMF components; where k1 represents the number of repeated calculations; i1 represents the maximum order of the IMF component;

[0146]

[0147] Each time an IMF component is obtained, it is subtracted from the original intermediate frequency data to obtain the residual component r1(t), and then the next-order IMF component is subtracted from the residual component r1(t) until the final residual component r i1 (t) is just a monotone sequence or a constant sequence, and the modal components of each order of the signal are obtained; among them, i1 is the maximum order

[0148]

[0149] Perform HHT transform on each IMF component, that is, perform Hilbert transform on each IMF component:

[0150]

[0151] Then find the instantaneous frequency ω of the corresponding IMF component i1 (t):

[0152] θ i1 (t) = arctan(d i1 (t) / imfi i1 (t))

[0153]

[0154] Yes i1 The time-frequency matrix W composed of and t is subjected to singular value decomposition; among them,

[0155] SVD(W)=UDV T

[0156] Where SVD(·) is the singular value decomposition function, U is an M×M matrix, D is an M×k matrix, all elements of matrix D are 0 except the elements on the main diagonal, each element on the main diagonal is a singular value, and the size of the singular value decreases from the upper left to the lower right, and V is a k×k matrix.

[0157] As a further solution, in step 6:

[0158] The time when the current sub-channel exceeds the energy threshold for the first time is counted as the signal starting time T start The moment when the current subchannel falls below the energy threshold for the first time after the signal start time is taken as the signal cutoff time T end ;

[0159] Calculate the signal duration T hold =T end -T start ;

[0160] The signal bandwidth is obtained by counting the number of adjacent sub-channels n where the signal appears: BW signal =B w ×n.

[0161] As a further solution, in step 7, the fingerprint features and parameter features extracted in steps 5 and 6 are combined to form a signal feature vector F;

[0162] F=[F signal , B.W. signal , T hold ]

[0163] Among them, F signal Indicates signal fingerprint characteristics, BW signal represents the signal bandwidth, T hold Indicates the signal duration.

[0164] As a further solution, in step 8, the correlation coefficient R is calculated by the following formula F，P :

[0165]

[0166] Among them, F v The vth eigenvalue in the quantity representing the current signal characteristics, represents the mean of the eigenvalues ​​of the current signal eigenvector, Represents the vth eigenvalue in the signal feature vector of the uth individual in the feature library, represents the mean eigenvalue of the u-th individual signal feature vector in the feature, and l+7 represents the dimension of the individual fingerprint feature.

[0167] In summary, the present invention aims at the problem of extracting transmitter fingerprint features of signals. According to the difference in features exhibited by the signal in the transient stage and the steady-state stage, the time domain signal analysis that can reflect the features of the signal in the transient stage and the time-frequency domain signal analysis in the steady-state stage are feature-fused, and the transient features of the energy envelope of the signal are extracted using the time domain analysis technology. The focus is on extracting seven types of parameters, including RJ parameters, curve area, duration, kurtosis, slope, and variance, of the transient part of the signal envelope. In the steady-state stage of the signal, the EMD-HHT time-frequency conversion method is used to improve the analysis and processing speed of the signal while ensuring the time resolution and frequency resolution of the time-frequency signal. Through the SVD decomposition method, the time-frequency signal matrix with relatively rich feature parameters is represented by a few eigenvalues, and the individual differences of the signal transmitters are amplified to the greatest extent.

[0168] The present invention extracts fingerprint features by fusing time domain features with time-frequency domain features. It not only analyzes the fingerprint features of the signal from the signal envelope, but also extracts the transient and steady-state features of the signal from the high-dimensional time-frequency domain. While ensuring the speed of signal feature extraction, it analyzes the features of the signal transmitter from multiple dimensions and performs feature fusion, thereby improving the accuracy of subsequent signal transmitter fingerprint feature extraction and matching. The signal envelope diagrams of two different radiation source individuals are shown in Figure 2. Figure 5 As shown (the left is the signal envelope diagram of radiation source individual 1, and the right is the signal envelope diagram of radiation source individual 2).

[0169] The existing frequency hopping detection method based on time-frequency characteristics detects the characteristics of the frequency hopping signal such as bandwidth and duration, and is easily interfered by burst signals or constant signals with similar bandwidth.

[0170] Based on the above-mentioned frequency hopping detection method, the present invention combines the characteristics of the signal transmitter fingerprint mechanism, utilizes the unique characteristics of the transmitter fingerprint mechanism and the characteristics that the same group of frequency hopping signals are transmitted from the same transmitter, performs feature extraction and feature fusion on the signal, and then pre-sorts the signals belonging to the same transmitter through feature matching, and performs frequency hopping signal detection based on the signals transmitted by the same transmitter, which can effectively avoid the interference problem between the above-mentioned burst signal and the constant signal. Due to the addition of the new dimension of the transmitter fingerprint feature, in the process of frequency hopping signal detection, the frequency hopping signal can be quickly judged based on the current signal fingerprint feature matching result and the bandwidth and duration result, which is effectively improved in detection speed compared to the existing frequency hopping detection method based on time-frequency features.

[0171] The above embodiments only express preferred implementation modes, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention.

Claims

1. A method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching, characterized in that: Perform frequency hopping signal detection through the following steps: Step 1: Divide the frequency band to be detected of the frequency hopping signal into sub-channels; Step 2: According to the set detection energy threshold, compare with the energy of each sub-channel; if the current sub-channel energy is greater than the detection energy threshold, there is a signal in the current sub-channel; otherwise, there is no signal in the current sub-channel; Step 3: Merge each subchannel with a signal; if adjacent subchannels have signals, the bandwidth of the current signal is determined as the bandwidth of the adjacent subchannel; Step 4: According to the number of sub-channels where the signal exists obtained in step 3, obtain the intermediate frequency data corresponding to the sub-channels where the signal exists; Step 5: Extract fingerprint features through the intermediate frequency data of the sub-channel where the signal exists, and obtain the signal fingerprint features; Step 6: According to the sub-channel where the signal exists, the signal bandwidth, signal start and end time and duration of the current signal are calculated to obtain the corresponding signal parameter characteristics; Step 7: Merge the signal fingerprint features and signal parameter features extracted in steps 5 and 6 to form a corresponding signal feature vector; Step 8: Determine whether the current signal feature vector is the first batch of signal feature vectors in the current detection; if so, directly store the signal feature vector in the library; if not, match the current signal feature vector with the signal feature vector in the fingerprint feature library and calculate the correlation coefficient R F,P ; Step 9: Get the matching results of step 8; if the correlation coefficient R F,P If the signal fingerprint feature is greater than the preset correlation threshold, the signal fingerprint feature matching is successful, indicating that the signal transmitted by the same signal transmitter is obtained at different times; otherwise, it means that the signal transmitted by the same transmitter is not received at the current time, and the signal feature vector extracted in step 7 is stored in the database; Step 10: According to the result of step 9, if the frequency hopping signals transmitted by multiple signal transmitters are matched, it means that the signal sorting operation is performed on the multiple groups of frequency hopping signals, and the results are displayed in the detection results according to the serial numbers of the matched multiple transmitters.

2. A method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching according to claim 1, characterized in that: In step 1, the bandwidth of the frequency band to be detected is N, and the number of sub-channel divisions is M; then the bandwidth occupied by each sub-channel is B w =N / M, the center frequency of each sub-channel is: i is the current subchannel number.

3. The method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching according to claim 2, characterized in that: After the sub-channels are divided, the sub-channel data is obtained through the following steps: Obtain the original intermediate frequency data before channelization, and perform down-conversion operation according to the center frequency of each sub-channel; Among them, x i is the original intermediate frequency data before channelization, f c is the center frequency of the current subchannel that needs to be channelized, o i is the data after down-conversion, t is the signal time parameter, and j is the imaginary unit; The down-converted intermediate frequency data is low-pass filtered according to the bandwidth of each sub-channel to obtain the channelized sub-channel data; the frequency domain expression is: Y = O(f)·H(f); Where, f is the frequency, f b is the cut-off frequency B w / 2, O(f) is the frequency domain expression of the intermediate frequency data after down-conversion, and Y is the frequency domain expression of the sub-channel data after channelization.

4. The method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching according to claim 2, characterized in that: In step 2, the detection energy threshold th is set and compared with the energy En of each sub-channel; wherein, In the formula, r i represents the energy of each signal sampling point in the subchannel, k represents the number of sampling points in the current subchannel, Channel i Represents the subchannel signal presence parameter, 1 means the signal exists, 0 means the signal does not exist.

5. A method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching according to claim 4, characterized in that: In step 3, the bandwidth of the combined subchannels is: Channel i =Channel i+1 =...=Channel i+k-1 =1 Among them, Bd signal It represents the bandwidth after the sub-channels are merged, kd is the number of adjacent channels, and i is the number of the current sub-channel.

6. A method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching according to claim 1, characterized in that: In step 5, fingerprint feature extraction is performed through the following steps: Perform filtering operation on the received intermediate frequency data; The filtering operation is completed by acquiring and normalizing the envelope signal through the intermediate frequency data; Extract the time domain features of the envelope signal and extract the RJ parameters of the envelope signal; Obtain the transient energy envelope curve, extract the curve area, duration, kurtosis, slope, and variance of the transient energy envelope curve, and combine them as characteristic parameters Perform time-frequency domain conversion on the intermediate frequency signal to obtain matrix D; Get the first k singular values ​​of matrix D as the eigenvalues ​​extracted from the signal fingerprint feature Perform feature fusion on the obtained features to obtain the final extracted signal fingerprint features 7. A method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching according to claim 6, characterized in that: The time-frequency domain conversion of the intermediate frequency signal is carried out through the following steps: Find all local maximum and local minimum points of the original intermediate frequency data; Use cubic spline curve to fit the envelope w of upper and lower extreme points max (t) and envelope e min (t); Find the average value of the upper and lower envelopes; Calculate the difference d(t) between the original intermediate frequency data and the average values ​​of the upper and lower envelopes; d(t)=x(t)-m(t) Among them, x(t) represents the original intermediate frequency data, and m(t) represents the average value of the upper and lower envelopes; Determine whether the obtained subtraction difference d(t) satisfies the following IMF conditions: Condition 1: The difference between the number of extreme points and the number of zero-crossing points in the decomposed time series does not exceed 1; Condition 2: In the analyzed time series, the mean of the extreme envelope fitted by interpolation is 0. Repeat the calculation of the difference d(t) until the IMF condition is met. k1 (t) is an IMF component, and so on, we can get several orders of IMF components; where k1 represents the number of repeated calculations; i1 represents the maximum order of the IMF component; Each time an IMF component is obtained, it is subtracted from the original intermediate frequency data to obtain the residual component r1(t), and then the next-order IMF component is subtracted from the residual component r1(t) until the final residual component r i1 (t) is just a monotone sequence or a constant sequence, and the modal components of each order of the signal are obtained; among them, i1 is the maximum order Perform HHT transform on each IMF component, that is, perform Hilbert transform on each IMF component: Then find the instantaneous frequency ω of the corresponding IMF component i1 (t): θ i1 (t)=arctan(d i1 (t) / imf i1 (t)) Yes i1 The time-frequency matrix W composed of and t is subjected to singular value decomposition; among them, SVD(W)=UDV T Where SVD(·) is the singular value decomposition function, U is an M×M matrix, D is an M×k matrix, all elements of matrix D are 0 except the elements on the main diagonal, each element on the main diagonal is a singular value, and the size of the singular value decreases from the upper left to the lower right, and V is a k×k matrix.

8. The method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching according to claim 1, characterized in that: In step 6: The time when the current sub-channel exceeds the energy threshold for the first time is counted as the signal starting time T start The moment when the current subchannel falls below the energy threshold for the first time after the signal start time is taken as the signal cutoff time T end ; Calculate the signal duration T hold =T end -T start ; The signal bandwidth is obtained by counting the number of adjacent sub-channels n where the signal appears: BW signal =B w ×n.

9. The method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching according to claim 1, characterized in that: In step 7, the fingerprint features and parameter features extracted in steps 5 and 6 are combined to form a signal feature vector F; F=[F signal ,BW signal ,T hold ] Among them, F signal Indicates signal fingerprint characteristics, BW signal represents the signal bandwidth, T hold Indicates the signal duration.

10. The method for detecting and sorting frequency hopping signals based on signal fingerprint mechanism matching according to claim 1, characterized in that: In step 8, the correlation coefficient R is calculated by the following formula: F,P : Among them, F v The vth eigenvalue in the quantity representing the current signal characteristics, represents the mean of the eigenvalues ​​of the current signal eigenvector, Represents the vth eigenvalue in the signal feature vector of the uth individual in the feature library, represents the mean eigenvalue of the u-th individual signal feature vector in the feature, and l+7 represents the dimension of the individual fingerprint feature.

Images