CN115225441A - Unmanned aerial vehicle cluster communication waveform identification method in complex environment - Google Patents
Unmanned aerial vehicle cluster communication waveform identification method in complex environment Download PDFInfo
- Publication number
- CN115225441A CN115225441A CN202210858726.3A CN202210858726A CN115225441A CN 115225441 A CN115225441 A CN 115225441A CN 202210858726 A CN202210858726 A CN 202210858726A CN 115225441 A CN115225441 A CN 115225441A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- signal
- cluster communication
- channel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0012—Modulated-carrier systems arrangements for identifying the type of modulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/18502—Airborne stations
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Aviation & Aerospace Engineering (AREA)
- Astronomy & Astrophysics (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses an unmanned aerial vehicle cluster communication waveform identification method under a complex environment, which comprises the steps of establishing an unmanned aerial vehicle cluster communication multipath fading channel under Alpha noise interference, and acquiring a received signal passing through the channel; preprocessing a received signal, including nonlinear conversion, down-conversion and band-pass sampling; extracting generalized cyclic mean values and generalized cyclic spectrum characteristics of the preprocessed signals, and constructing unmanned aerial vehicle cluster communication waveform characteristic matrixes under different signal-to-noise ratios; and inputting the characteristic matrix into an SAE neural network for training and testing, and outputting the type of the cluster communication waveform of the unmanned aerial vehicle, so as to realize the identification of the cluster communication waveform of the unmanned aerial vehicle in a complex environment with Alpha noise interference, multipath fading and frequency shift. The method has stronger robustness in a complex environment, and can still ensure the identification accuracy rate of more than 80 percent when the signal-to-noise ratio is-10 dB.
Description
Technical Field
The invention belongs to the technical field of digital communication, and particularly relates to an unmanned aerial vehicle cluster communication waveform identification method in a complex environment.
Background
Because the unmanned aerial vehicle cluster possesses the characteristics of high harmony, multifunctionality, strong survivability, its status in military fields such as reconnaissance location, electronic countermeasure, communication information is increasing outstanding. Unmanned aerial vehicle cluster operation becomes one of the main forms of future war, and for maintaining national security, the research of unmanned aerial vehicle cluster countermeasures technology is urgent, and the realization of unmanned aerial vehicle cluster communication waveform identification is a key link of the research of unmanned aerial vehicle cluster countermeasures technology.
In the non-cooperative communication field, a communication signal of an enemy needs to be intercepted first, analysis and processing are carried out, and the enemy information is obtained, and automatic modulation identification of the signal is the basis for obtaining the enemy information. Modulation identification occurs between signal detection and demodulation, the basic tasks of which are: and analyzing, judging and classifying the modulation type of the intercepted unknown signal.
The current modulation identification method mainly comprises two categories: modulation identification based on maximum likelihood function and modulation identification based on feature extraction. The modulation identification process is mainly divided into three steps: signal preprocessing, feature extraction and classification identification. The signal preprocessing process comprises down-conversion, noise suppression, estimation of carrier frequency or symbol period and the like; the feature extraction is to extract features which can represent the modulation type of the signal from the signal, such as a cyclic spectrum, high-order cumulant, wavelet transformation features and the like; the classification identification is to combine the extracted features with classification performance and select a proper mode to perform decision classification, such as a decision tree, a neural network and the like.
The research of the existing communication signal automatic modulation identification method is mostly considered under the Gaussian channel, and the assumed channel condition of the research is too ideal. In practical communication, especially in battlefield environment, multipath effect and doppler shift occur inevitably in signal during transmission, and user interference may exist.
The existing modulation identification algorithm has over-ideal channel conditions and poor robustness with low signal-to-noise ratio, and is difficult to meet the requirement of unmanned aerial vehicle cluster communication waveform identification.
Disclosure of Invention
The invention aims to solve the technical problem of the prior art, provides an unmanned aerial vehicle cluster communication waveform identification method in a complex environment, fully considers the problems of small-scale fading and interference among users of unmanned aerial vehicle cluster communication signals in the transmission process, extracts the cyclostationarity of the unmanned aerial vehicle communication signals, and realizes the identification of the unmanned aerial vehicle cluster waveform through a Sparse Auto Encoder (SAE) neural network.
In order to achieve the technical purpose, the technical scheme adopted by the invention is as follows:
an unmanned aerial vehicle cluster communication waveform identification method under a complex environment comprises the following steps:
step 1: establishing an unmanned aerial vehicle cluster communication multipath fading channel under Alpha noise interference, and acquiring a received signal passing through the channel;
step 2: preprocessing a received signal, including nonlinear conversion, down-conversion and band-pass sampling;
and step 3: extracting generalized cyclic mean values and generalized cyclic spectrum characteristics of the preprocessed signals, and constructing unmanned aerial vehicle cluster communication waveform characteristic matrixes under different signal-to-noise ratios;
and 4, step 4: and inputting the characteristic matrix into an SAE neural network for training and testing, and outputting the type of the cluster communication waveform of the unmanned aerial vehicle, so as to realize the identification of the cluster communication waveform of the unmanned aerial vehicle in a complex environment with Alpha noise interference, multipath fading and frequency shift.
In order to optimize the technical scheme, the specific measures adopted further comprise:
the unmanned aerial vehicle trunking communication waveform types comprise BPSK, QPSK, 2FSK, 4FSK, 2ASK, and MSK.
In the above step 1, a TDL model and Alpha stable distributed noise are used to establish an unmanned aerial vehicle cluster communication multipath fading channel, and the channel parameters are set based on a 3gpp TR 901.38 technical report, so that the intercepted unmanned aerial vehicle communication signal, that is, the acquired received signal passing through the channel is represented as:
wherein, x (t) is a transmitted modulation signal, and n (t) is Alpha stable distributed noise;
h l (t) and τ l Respectively corresponding channel coefficient and time delay of the first multipath, wherein L is more than or equal to 0 and less than or equal to L-1, and L is the distinguishable path number of the multipath fading channel.
The above-mentioned channel coefficient h l (t) is obtained by multiplying the output of the L flat fading signal generators by the power of each tap.
The step 2 includes:
1) Nonlinear transformation of the received signal r (t):
wherein, delta is a normal number;
2) Down-converting the signal after the nonlinear transformation to 140MHz;
3) And performing band-pass sampling and low-pass filtering on the down-converted signal, and moving the center frequency of the signal to 4MHz.
The step 3 includes:
1) Extracting generalized cyclostationary feature, the generalized cyclostationary mean of the preprocessed signal r' (t)
Is defined as follows:
wherein ε = k/T is the cycle frequency, M r' (t) is the mean of the signals r' (t);
2) Generalized cyclic spectral density of signal r' (t)
Expressed as:
wherein the content of the first and second substances,
a cyclic autocorrelation function that is the signal r' (t);
3) Selecting the generalized cyclic mean value, the discrete peak values of the generalized cyclic spectrum and the number of the discrete peak values as features, and expressing the feature matrix under a certain mixed signal-to-noise ratio as follows:
wherein the content of the first and second substances,
feature ρ representing the nth sample signal of drone user a with BPSK modulation 1 。
In the step 4, an SAE neural network is used for identifying cluster communication waveforms of the unmanned aerial vehicle;
the SAE neural network has a sparse characteristic, and the forward propagation equation is as follows:
S in ＝σ[U(u 1 ,...,u m )×X t +a] (11)
S out ＝O in ＝σ(η 1 ×S in ×V t (v 1 ,v 2 ...v n )+b) (12)
O out ＝f(η 2 ×S out ×W t (w 1 ,w 2 …w p )+c) (13)
wherein S is in ，S out And O out Respectively, the input value, output value and final output value of the hidden layer, U, V and W are weight matrix of the corresponding connection layer, X t A feature matrix constructed for step 3;
and sigma and f are activation functions, sigma is a tanh function or a sigmoid function, f is a Softmax function, eta is a sparse coefficient of the hidden layer, and a, b and c are offsets of each layer.
The invention has the following beneficial effects:
the unmanned aerial vehicle cluster communication electromagnetic environment is complex, phenomena such as user interference, multipath fading and frequency shift exist, the performance of a waveform identification algorithm under a traditional Gaussian channel is greatly reduced under the scene, and aiming at the problem, the invention provides an unmanned aerial vehicle cluster communication waveform identification method under the complex environment. The invention establishes an unmanned aerial vehicle cluster communication multipath fading channel under Alpha noise interference, and obtains a received signal passing through the channel; preprocessing a received signal, including nonlinear conversion, down-conversion and band-pass sampling; extracting generalized cyclic mean values and generalized cyclic spectrum characteristics of the preprocessed signals, and constructing unmanned aerial vehicle cluster communication waveform characteristic matrixes under different signal-to-noise ratios; and inputting the characteristic matrix into an SAE neural network for training and testing, and outputting the type of the cluster communication waveform of the unmanned aerial vehicle, so as to realize the identification of the cluster communication waveform of the unmanned aerial vehicle in a complex environment with Alpha noise interference, multipath fading and frequency shift. Simulation results show that: the method has stronger robustness in a complex environment, and can still ensure more than 80% of recognition accuracy when the signal-to-noise ratio is-10 dB.
Drawings
Fig. 1 is a diagram of a cluster communication scenario.
Fig. 2 is a diagram of the position relationship between the unmanned aerial vehicle cluster user and the receiver.
Fig. 3 is a generalized cyclic average of a 2FSK signal.
Fig. 4 is a generalized cyclic average of the 2ASK signal.
Fig. 5 is a generalized cyclic spectrum of a BPSK signal.
Fig. 6 is a generalized cyclic spectrum of a QPSK signal.
Fig. 7 is a graph of the path gain of the TDL-a model as a function of channel path and sample time.
Fig. 8 is a graph of the path gain of the TDL-D model as a function of channel path and sample time.
Fig. 9 shows a signal modulation scheme prediction model based on SAE.
Fig. 10 is a signal modulation identification performance curve for the TDL-a channel.
Fig. 11 is a signal modulation identification performance curve under a TDL-D channel.
FIG. 12 is a TDL-A channel signal modulation identification performance curve of different scenes;
fig. 13 is a flow chart of the method of the present invention.
Detailed Description
Embodiments of the present invention are described in further detail below with reference to the accompanying drawings.
The invention establishes an unmanned aerial vehicle cluster communication multipath fading channel under Alpha noise interference, designs an unmanned aerial vehicle cluster communication scene, intercepts unmanned aerial vehicle cluster communication signals, preprocesses the signals, extracts the cyclic mean value and cyclic cumulant characteristics (namely generalized cyclic mean value and generalized cyclic spectrum characteristics), establishes unmanned aerial vehicle cluster communication waveform characteristic vectors, generates multidimensional characteristic matrixes under different signal-to-noise ratios, establishes SAE neural network as an unmanned aerial vehicle cluster communication waveform identification model, and realizes identification of 6 unmanned aerial vehicle cluster waveforms.
As shown in fig. 13, the method for identifying cluster communication waveforms of unmanned aerial vehicles provided by the present invention includes the following steps:
step 1: establishing an unmanned aerial vehicle cluster communication multipath fading channel under Alpha noise interference according to 3GPP TR 901.38, and acquiring a received signal passing through the channel;
multipath effect, doppler shift and user interference inevitably exist in the cluster communication process of the unmanned aerial vehicle, the method establishes a Rice/Rayleigh fading channel based on a Tapped Delay Line (TDL) model, and adopts Alpha stable distributed noise as interference. The impulse response of the TDL channel can be expressed as:
wherein h is l (t) and τ l Respectively the channel coefficient and the time delay corresponding to the first multipath, wherein L is more than or equal to 0 and less than or equal to L-1, and L is the distinguishable path number of the multipath channel.
Channel systemNumber h l (t) is obtained by multiplying the output of L flat fading signal generators (implemented by Jakes model) by the power of each tap.
The invention adopts Alpha stable distribution to simulate high-intensity pulse noise:
wherein alpha is a characteristic index, gamma is a dispersion coefficient, beta is a deviation parameter, upsilon is a position parameter, and sgn (t) is a sign function.
The invention adopts symmetrical Alpha stable distribution noise with Alpha =1.5, beta =0, gamma =1 and upsilon =0, and defines a mixed signal-to-noise ratio:
in the formula (I), the compound is shown in the specification,
is the average power of the signal.
To this end, the received signal may be expressed as:
where x (t) is the transmitted modulation signal, and n (t) is Alpha stationary distributed noise.
Step 2: preprocessing a received signal, including nonlinear conversion, down conversion, and bandpass sampling, comprises:
1) To suppress the effect of Alpha noise, the received signal is first subjected to a nonlinear transformation:
wherein, delta is a normal number;
2) Then, in order to reduce the cost of radio frequency sampling, the signal after nonlinear conversion is converted to 140MHz in a down-conversion mode;
3) And then performing band-pass sampling and low-pass filtering on the down-converted signal, and moving the center frequency of the signal to 4MHz.
And 3, step 3: aiming at the problem that high-order statistics does not exist in Alpha stable distribution noise, extracting generalized cyclic mean values and generalized cyclic spectrum characteristics of preprocessed signals, and constructing unmanned aerial vehicle cluster communication waveform characteristic matrixes under different signal-to-noise ratios;
and extracting the preprocessed cluster communication waveform characteristics of the unmanned aerial vehicle.
After preprocessing, extracting the circulation stability characteristics of the intercepted unmanned aerial vehicle cluster communication waveform, including the generalized circulation mean characteristic and the generalized circulation spectrum characteristic.
The generalized cyclic mean of the preprocessed signal r' (t) is defined as:
wherein ε = k/T is the cycle frequency, M r' (t) is the mean value of the signal r' (t).
Generalized cyclic spectral density of signal r' (t)
Expressed as:
wherein the content of the first and second substances,
a cyclic autocorrelation function for the signal r' (t), defined as:
when ε =0, the cyclic spectral density degrades to the power spectral density.
And step 3: and establishing an unmanned aerial vehicle cluster communication waveform characteristic vector. In the invention, assuming that two users in an unmanned aerial vehicle cluster communicate by adopting BPSK waveform under a certain mixed signal-to-noise ratio, and the generalized cyclostationary features of the two users are extracted, the feature vector of the sample is as follows:
wherein the content of the first and second substances,
and
respectively representing the num-th features of drone user a and B.
According to table 4, the generalized cyclic mean value, the discrete peaks of the generalized cyclic spectrum, and the number thereof are selected as features, and then under a certain mixed signal-to-noise ratio, the feature matrix of the structure of intercepting N samples by each waveform is represented as:
wherein the content of the first and second substances,
feature ρ representing the nth sample signal of drone user a with BPSK modulation 1 ，N＝1000。
And 4, step 4: inputting the characteristic matrix into an SAE neural network for training and testing, outputting the type of the cluster communication waveform of the unmanned aerial vehicle, and realizing the identification of 6 types of cluster communication waveforms of the unmanned aerial vehicle under the complex environment with Alpha noise interference, multipath fading and frequency shift.
And establishing an SAE neural network as an unmanned aerial vehicle cluster communication waveform identification model.
And (4) forming a multi-dimensional characteristic matrix under different mixed signal-to-noise ratios by using the extracted unmanned aerial vehicle communication waveform characteristics, and inputting the multi-dimensional characteristic matrix into SAE neural network training.
The forward propagation equation for SAE with sparse properties is as follows:
S in ＝σ[U(u 1 ,...,u m )×X t +a] (11)
S out ＝O in ＝σ(η 1 ×S in ×V t (v 1 ,v 2 ...v n )+b) (12)
O out ＝f(η 2 ×S out ×W t (w 1 ,w 2 ...w p )+c) (13)
wherein S is in ，S out And O out Respectively the input value of the hidden layer, the output value and the final output value of the hidden layer, U, V and W respectively are the weight matrix of the corresponding connection layer, X t Is the constructed feature vector. And sigma and f are activation functions, sigma is usually a tanh function or a sigmoid function, f is a Softmax function, eta is a sparse coefficient of the hidden layer, and a, b and c are offsets of each layer.
And finally, inputting the waveform characteristics of the unmanned aerial vehicle cluster into the trained neural network, and judging the communication waveform of the unmanned aerial vehicle cluster.
The invention has completed MATLAB software simulation and verification.
The following gives the specific implementation steps of the invention:
(1) And (2) according to the step 1, establishing an unmanned aerial vehicle cluster communication multipath fading channel by adopting a TDL model and Alpha stable distributed noise.
The channel parameter setting in The present invention is based on The channel model technical report 3GPP TR 901.38 (V17.0.0) of 0.5-100 GHz issued by The third Generation Partnership project (The 3rd Generation Partnership project,3 GPP).
The 3GPP TR 38.901 report provides 5 TDL channel model parameters according to measured data, wherein, TDL-A, TDL-B and TDL-C are suitable for Non-Line of Sight (NLOS) scenes, and TDL-D and TDL-E are suitable for Line of Sight (LOS) scenes.
The technical report specifies the fading obeying Rayleigh distribution of the TDL channel, provides the multipath number of the channel under each model parameter and the standardized time delay and tap power corresponding to each path, and configures the reference report of the specific parameters from table 7.7.2-1 to table 7.7.2-5. The time delays of the above contents are all standardized time delays, i.e. Root Mean Square (RMS) time Delay Spread (DS) thereof is 1.
The specification of 3GPP TR 38.901 states that the tap time delay under a specified scene can be obtained by adjusting the RMS time delay expansion, and the RMS time delay expansion under different scenes refers to a report table 7.7.3-1.
Also, the report specifies that for channel models containing LOS paths, such as TDL-D and TDL-E, the user can obtain the specified K by adjusting the K factor desired (dB), 3GPP TR 38.901 provides the mean and standard deviation of the K factor for different scenarios, see tables 7.5-6.
After adjusting the K factor, the power corresponding to each tap of the model should be adjusted accordingly. The specific adjustment modes of the time delay and the K factor refer to section 7.7.2 in 3GPP TR 38.901.
(2) And setting parameters of the unmanned aerial vehicle cluster communication scene.
Consider that multiple users in a drone cluster communicate using orthogonal resources (time, frequency, code, space), such as user U in the cluster A And U B Communication is carried out, assuming U A And U B The time/frequency/code/space domain are orthogonal, and the detector can respectively obtain the signals transmitted by two users.
The invention considers that all users in the cluster adopt the same modulation mode, the possible adopted modulation mode types comprise BPSK, QPSK, 2FSK, 4FSK, 2ASK and MSK, and the users communicating with each other in the cluster are two unmanned aerial vehicles.
When the communication frequency of the unmanned aerial vehicle is 5.8GHz, the atmospheric loss is extremely low, and the electromagnetic wave can be considered to be transmitted in free space, and the transmission loss L is P (dB) is defined as:
wherein, P T To transmit power, P R To receive power, f c (GHz) is the operating frequency, and d (km) is the distance between the transmit and receive antennas.
Assume that any two users in the drone cluster have the same parameters except for a different distance from the receiver. Set up unmanned aerial vehicle user A apart from listening machine d 1 Km, unmanned aerial vehicle user B distance detecting and receiving machine d 2 km。
Defining a path loss difference DeltaL between two drone users within a cluster p (dB), table 1 is the distance ratio and path loss difference between two users in a cluster of drones.
TABLE 1 distance ratio between two users in a cluster versus pathloss
| d 1 /d 2 | 0.71 | 0.74 | 0.77 | 0.79 | 0.83 | 0.86 | 0.89 | 0.92 | 0.95 | 1.00 |
| ΔL(dB) | 3.0 | 2.6 | 2.3 | 2 | 1.6 | 1.3 | 1.0 | 0.7 | 0.4 | 0 |
Two unmanned aerial vehicle users are in communication in the cluster, the communication frequency is 5.8GHz, and the code element rate is 2M symbol/s. Taking into account its path loss difference Δ L p The scenes are respectively 0dB and 3dB, correspondingly, the distance ratio of the two unmanned aerial vehicle users relative to the receiver is respectively 0.71 and 1, and other parameter settings are shown in a table 2.
The invention takes 2dB as step length to respectively generate data sets with the mixed signal-to-noise ratio of-20 dB to 0 dB.
The drone and channel parameters are set according to table 2, and 1000 signal samples are generated for drone user a and user B under each modulation mode.
Table 2 unmanned aerial vehicle user parameter table
The flying speed of the unmanned aerial vehicle can influence the value of the maximum Doppler frequency shift, so that the identification accuracy of the cluster waveform of the unmanned aerial vehicle is influenced.
Consider the path loss difference Δ L of two drone users p In a 3dB scene, a TDL-A model is adopted, the communication frequency of the unmanned aerial vehicle is 5.8GHz, and other parameters refer to a table 3.
Table 3 user parameter table for unmanned aerial vehicle
(3) And (4) preprocessing the intercepted unmanned aerial vehicle cluster communication waveform signals according to the step (2).
As the statistics of second order and above orders of Alpha stable distributed noise tend to be infinite, the received signal needs to be subjected to nonlinear transformation, and the infinite amplitude of the noise is limited in a limited interval so as to obtain effective signal characteristics.
The signal after nonlinear transformation is:
wherein, delta is a normal number.
The features of the signal after nonlinear processing are called generalized cyclic mean and generalized cyclic spectrum.
The communication frequency of the unmanned aerial vehicle is set to be 5.8GHz, in order to reduce the radio frequency sampling cost, the signal is subjected to down-conversion to 140MHz, then band-pass sampling is carried out, and the sampling rate f s Is 16M sample/s.
After low-pass filtering, the signal center frequency is shifted to 4MHz.
Currently, there are many pairs of signal carrier frequencies f c And symbol rate R b The invention relates to an equal parameter estimation method, which uses the communication frequency and the code element rate of an unmanned aerial vehicle as known information.
(4) And (4) extracting the communication waveform characteristics of the unmanned aerial vehicle according to the step (3) and constructing an unmanned aerial vehicle cluster communication waveform characteristic matrix.
Table 4 shows the cyclic mean and cyclic spectrum characteristics of the signals with different modulation types.
According to table 4, input characteristic parameters of the neural network are selected: for the cyclic mean characteristics of different modulation signals, the discrete peak number rho of the cyclic mean is used 1 And average cyclic mean value ρ 2 As a characteristic parameter;
discrete peak values ρ with f =0 cross section and ∈ > 0 for cyclic spectral characteristics of different modulation signals 3 ～ρ 7 As characteristic parameter, i.e. epsilon =2f c ，ε＝2f c ±R b And e =2f c ±R b The peak of the circulating spectral density at/2.
TABLE 4 different modulation signal cycle characteristics
Fig. 3 to 6 are generalized cyclostationary features of a part of a modulated signal, respectively. After the unmanned aerial vehicle communication signal is preprocessed, the center frequency is shifted to 4MHz. As can be seen from fig. 3 and 4, the generalized cyclic mean of the 2FSK signal has two discrete peaks, while the generalized cyclic mean of the 2ASK signal has only one discrete peak; as can be seen from fig. 5 and 6, in the generalized cyclic spectrum of the BPSK signal, = ± (2 f) on the f =0 cross section c ±R b ) There are discrete peaks, while the f =0 cross section of QPSK does not.
Two communicating users exist in the unmanned plane cluster, and the constructed characteristic matrix is as follows:
wherein, the first and the second end of the pipe are connected with each other,
feature ρ representing the nth sample signal of drone user a with BPSK modulation 1 ，N＝1000。
Fig. 1 and 2 depict the presence of two communicating users U within a certain drone cluster A And U B The two communication parties adopt the same modulation mode, the detecting and receiving machines can respectively obtain the signals sent by the two users, and the unmanned aerial vehicle user A is a distance detecting and receiving machine d 1 Km, unmanned aerial vehicle user B distance detecting and receiving machine d 2 km。
(5) And (4) identifying the unmanned aerial vehicle cluster waveform by adopting an SAE neural network according to the step 4.
And identifying the unmanned aerial vehicle cluster waveform according to the neural network described in the step 4, wherein the adopted SAE neural network structural parameters are shown in a table 5. The size ratio of the training set to the test set in the SAE neural network is 7: and 3, according to different scenes and user parameters set by the content, providing a performance curve for identifying the cluster communication waveform of the unmanned aerial vehicle under different signal-to-noise ratios.
TABLE 5 SAE neural network architecture parameter configuration
Fig. 7 and 8 show the path gain of the TDL-a model and TDL-D as a function of channel path and sample time, respectively. The TDL-A channel is an NLOS scene, and the TDL-D channel contains an LOS path. Channel parameters considered for TDL-a: maximum doppler shift 1160hz, rms delay spread 100ns, sample point 8000sample, and the model has 23 resolvable paths. Channel parameters considered for TDL-D: the maximum Doppler shift is 1160Hz, the RMS delay spread is 30ns, the number of sample points is 8000sample, the K factor is 9dB, and the model has 13 resolvable paths, wherein the first path is an LOS path, and the gain of the LOS path is larger than that of other paths.
Fig. 9 is a signal modulation scheme prediction model based on the SAE neural network. The cluster communication waveform characteristics of the unmanned aerial vehicle are input on the left side of the model, the cluster communication waveform types of the unmanned aerial vehicle are output on the right side of the model, and a structural parameter configuration reference table 5 of the SAE neural network is shown.
Fig. 10 and 11 are signal modulation identification performance curves for the TDL-a and TDL-D channels, respectively. The channel parameters and drone user parameters are in accordance with table 2. As can be seen from fig. 10 and 11: according to channel analysis, due to the fact that LOS paths exist in the TDL-D channel, the recognition accuracy of cluster communication waveforms of the unmanned aerial vehicle is higher than that of the TDL-A channel; according to analysis on the method, the accuracy of identifying the unmanned aerial vehicle cluster waveform through the SAE neural network is higher than that of a decision tree-based method; from the path loss difference analysis, the larger the path loss difference is, the lower the identification accuracy rate of the unmanned aerial vehicle cluster waveform is.
Fig. 12 is a signal modulation identification performance curve of different scenes under a DL-A channel. The flight speed of the unmanned aerial vehicle can influence the value of the maximum Doppler frequency shift, thereby influencing the identification accuracy of the unmanned aerial vehicle cluster waveform. Consider path loss difference Δ L of two drone users p In a 3dB scene, a TDL-A model is adopted, the communication frequency of the unmanned aerial vehicle is 5.8GHz, and other parameters refer to a table 5. As can be seen from fig. 12: with the increase of the flight speed and the Doppler frequency shift of the unmanned aerial vehicle, the accuracy of unmanned aerial vehicle cluster communication waveform identification is reduced.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and adaptations to those skilled in the art without departing from the principles of the present invention may be apparent to those skilled in the relevant art and are intended to be within the scope of the present invention.
Claims (7)
1. An unmanned aerial vehicle cluster communication waveform identification method under a complex environment is characterized by comprising the following steps:
step 1: establishing an unmanned aerial vehicle cluster communication multipath fading channel under Alpha noise interference, and acquiring a received signal passing through the channel;
and 2, step: preprocessing a received signal, including nonlinear conversion, down-conversion and band-pass sampling;
and step 3: extracting generalized cyclic mean values and generalized cyclic spectrum characteristics of the preprocessed signals, and constructing unmanned aerial vehicle cluster communication waveform characteristic matrixes under different signal-to-noise ratios;
and 4, step 4: and inputting the characteristic matrix into an SAE neural network for training and testing, and outputting the type of the cluster communication waveform of the unmanned aerial vehicle, so as to realize the identification of the cluster communication waveform of the unmanned aerial vehicle in a complex environment with Alpha noise interference, multipath fading and frequency shift.
2. The method according to claim 1, wherein the type of the UAV cluster communication waveform comprises BPSK, QPSK, 2FSK, 4FSK, 2ASK, and MSK.
3. The method according to claim 1, wherein in step 1, a TDL model and Alpha stable distributed noise are used to establish a multipath fading channel for cluster communication of the unmanned aerial vehicles, and channel parameters are set based on a report of 3gpp TR 901.38 technology, so that the intercepted unmanned aerial vehicle communication signals, that is, the acquired received signals passing through the channel, are represented as:
wherein, x (t) is a transmitted modulation signal, and n (t) is Alpha stable distributed noise;
h l (t) and τ l Respectively corresponding channel coefficient and time delay of the first multipath, wherein L is more than or equal to 0 and less than or equal to L-1, and L is the distinguishable path number of the multipath fading channel.
4. The unmanned aerial vehicle cluster communication waveform identification method in the complex environment of claim 4, wherein the channel coefficient h is l (t) is obtained by multiplying the output of the L flat fading signal generators by the power of each tap.
5. The unmanned aerial vehicle cluster communication waveform identification method in the complex environment according to claim 1, wherein the step 2 includes:
1) Nonlinear transformation is performed on the received signal r (t):
wherein, delta is a normal number;
2) Down-converting the signal after the nonlinear transformation to 140MHz;
3) And performing band-pass sampling and low-pass filtering on the down-converted signal, and moving the central frequency of the signal to 4MHz.
6. The unmanned aerial vehicle cluster communication waveform identification method in the complex environment according to claim 1, wherein the step 3 comprises:
1) Extracting generalized cyclostationary feature, the generalized cyclostationary mean of the preprocessed signal r' (t)
Is defined as follows:
wherein ε = k/T is the cycle frequency, M r' (t) is the mean of the signal r' (t);
2) Generalized cyclic spectral density of signal r' (t)
Expressed as:
wherein the content of the first and second substances,
as a cyclic autocorrelation of the signal r' (t)A function;
3) Selecting the generalized cyclic mean value, the discrete peak values of the generalized cyclic spectrum and the number of the discrete peak values as features, and expressing the feature matrix under a certain mixed signal-to-noise ratio as follows:
wherein the content of the first and second substances,
characteristic ρ representing the Nth sample signal of drone user A with BPSK modulation 1 。
7. The unmanned aerial vehicle cluster communication waveform identification method under the complex environment of claim 1,
in the step 4, an SAE neural network is used for identifying cluster communication waveforms of the unmanned aerial vehicle;
the SAE neural network has sparse characteristics, and the forward propagation equation is as follows:
S in ＝σ[U(u 1 ,...,u m )×X t +a] (11)
S out ＝O in ＝σ(η 1 ×S in ×V t (v 1 ,v 2 …v n )+b) (12)
O out ＝f(η 2 ×S out ×W t (w 1 ,w 2 ...w p )+c) (13)
wherein S is in ，S out And O out Respectively, the input value, output value and final output value of the hidden layer, U, V and W are weight matrix of the corresponding connection layer, X t A feature matrix constructed for step 3;
and sigma and f are activation functions, sigma is a tanh function or a sigmoid function, f is a Softmax function, eta is a sparse coefficient of the hidden layer, and a, b and c are offsets of each layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210858726.3A CN115225441A (en) | 2022-07-21 | 2022-07-21 | Unmanned aerial vehicle cluster communication waveform identification method in complex environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210858726.3A CN115225441A (en) | 2022-07-21 | 2022-07-21 | Unmanned aerial vehicle cluster communication waveform identification method in complex environment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115225441A true CN115225441A (en) | 2022-10-21 |
Family
ID=83613776
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210858726.3A Pending CN115225441A (en) | 2022-07-21 | 2022-07-21 | Unmanned aerial vehicle cluster communication waveform identification method in complex environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115225441A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115902804A (en) * | 2022-11-07 | 2023-04-04 | 南京航空航天大学 | Unmanned aerial vehicle cluster type identification method and system |
-
2022
- 2022-07-21 CN CN202210858726.3A patent/CN115225441A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115902804A (en) * | 2022-11-07 | 2023-04-04 | 南京航空航天大学 | Unmanned aerial vehicle cluster type identification method and system |
| CN115902804B (en) * | 2022-11-07 | 2024-01-05 | 南京航空航天大学 | Unmanned aerial vehicle cluster type identification method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pinto et al. | Communication in a Poisson field of interferers-Part II: Channel capacity and interference spectrum | |
| Yang et al. | Specific emitter identification with limited samples: A model-agnostic meta-learning approach | |
| CN101895494B (en) | Stochastic resonance preprocessing-based digital modulation mode automatic identification method | |
| CN110166387A (en) | A kind of method and system based on convolutional neural networks identification signal modulation system | |
| CN107018101A (en) | Based on the varying Channels method of estimation for simplifying basis expansion model | |
| CN115225441A (en) | Unmanned aerial vehicle cluster communication waveform identification method in complex environment | |
| CN108234044B (en) | Preprocessing method of wireless equipment identification classifier based on mixed input information | |
| US20030236072A1 (en) | Method and apparatus for estimating a channel based on channel statistics | |
| CN116318478A (en) | Energy efficiency-based normalized unmanned aerial vehicle frequency spectrum sensing method and system | |
| CN115987739A (en) | Communication perception integrated signal processing method based on double time scales | |
| CN115409056A (en) | Automatic modulation identification method for large dynamic signal-to-noise ratio | |
| CN112653497B (en) | Signal transceiving method for reducing MIMO multichannel phase noise influence | |
| CN110266624B (en) | Modulation mode blind identification method under interference | |
| Zhai et al. | Waveform Recognition of UAV Swarm Communication in Multipath Channel | |
| Hu et al. | Signal Classification in Real-time Based on SDR using Convolutional Neural Network | |
| CN111796263B (en) | Low-altitude target distributed ranging method based on 5G base station signals | |
| CN113381792A (en) | Waveform generation method based on signal-to-interference-and-noise ratio lower bound | |
| CN112232120A (en) | Radar radiation source signal classification system and method based on software radio | |
| Nerandžić et al. | Impact of AWGN estimation on classification performance of AMC algorithms based on higher order cumulants | |
| CN116866116B (en) | Time-delay mixed linear blind separation method | |
| Liu et al. | Specific Emitter Identification Method based on Deep Ensemble Learning | |
| Xu et al. | Wireless channel scenario recognition based on neural networks | |
| Kumar et al. | Comprehensive analysis of cyclo-stationary feature detection technique for efficient spectrum usage: Future research and recent advantages | |
| CN115426017B (en) | Multiple access method and system based on harmonic signal virtual non-orthogonal decomposition | |
| Taran et al. | Performance evaluation of 802.15. 3a UWB channel model with antipodal, orthogonal and DPSK modulation scheme |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |