CN109472239B - Individual identification method of frequency hopping radio station - Google Patents
Individual identification method of frequency hopping radio station Download PDFInfo
- Publication number
- CN109472239B CN109472239B CN201811328578.4A CN201811328578A CN109472239B CN 109472239 B CN109472239 B CN 109472239B CN 201811328578 A CN201811328578 A CN 201811328578A CN 109472239 B CN109472239 B CN 109472239B
- Authority
- CN
- China
- Prior art keywords
- frequency
- time
- equal
- energy spectrum
- frequency hopping
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000001228 spectrum Methods 0.000 claims abstract description 79
- 239000011159 matrix material Substances 0.000 claims description 26
- 238000004422 calculation algorithm Methods 0.000 claims description 22
- 238000009826 distribution Methods 0.000 claims description 20
- 239000013598 vector Substances 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 16
- 238000005457 optimization Methods 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 8
- 238000012706 support-vector machine Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 239000013256 coordination polymer Substances 0.000 claims description 3
- 230000014759 maintenance of location Effects 0.000 claims description 3
- 230000000087 stabilizing effect Effects 0.000 claims description 3
- 238000012549 training Methods 0.000 abstract description 9
- 238000002474 experimental method Methods 0.000 abstract description 2
- 238000000638 solvent extraction Methods 0.000 abstract 1
- 230000008859 change Effects 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 238000004891 communication Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000000739 chaotic effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000001052 transient effect Effects 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2136—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on sparsity criteria, e.g. with an overcomplete basis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Signal Processing (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The method comprises the steps of firstly sparsely reconstructing a frequency hopping signal to obtain a time frequency spectrum, then partitioning the time frequency energy spectrum under different scales, and respectively extracting three characteristics of a Rayleigh entropy, a difference box dimension and a multi-fractal dimension of the time frequency energy spectrum. The classification and identification experiments show that the identification performance of the method is less influenced by the number of training samples, and the method has higher identification accuracy under the condition of less training samples.
Description
Technical Field
The invention relates to wireless communication and signal processing technology, in particular to a frequency hopping radio station individual identification method.
Background
The sorting of the frequency hopping communication network station is the first premise of intercepting enemy communication and generating the best interference signal. The existing frequency hopping signal network station sorting mainly utilizes the duration, the azimuth information, the power, the signal time correlation and the like of a frequency hopping signal to realize the network station sorting identification of the frequency hopping signal. However, with the increase of the frequency hopping patterns, it is difficult to accurately sort the frequency hopping signals only by the above features. Due to the random discreteness of the component performance, the production process, the debugging and the like of each frequency hopping radio station, the frequency hopping signals radiated by the frequency hopping radio stations have individual characteristics which are different from those of other frequency hopping radio station signals, and therefore, the network station sorting and identification of the frequency hopping signals can be realized by utilizing the unique fine characteristics of different frequency hopping radio station signals, namely the fingerprint characteristics. At present, the commonly used fine feature extraction methods mainly include: the classification and identification are realized by extracting the characteristics of the frequency conversion transient signal of the frequency hopping communication equipment; sorting frequency hopping radio stations by using different characteristics of the transient response process of the transmitting power amplifier; the individual identification of the radiation source is realized by extracting the multi-dimensional characteristics of the time-frequency energy spectrum, but the identification accuracy is low due to more information redundancy of the characteristic set extracted by the method; the first moment, the second moment and the energy spectrum entropy characteristics of the time-frequency energy spectrum with fixed scale are extracted, but the time-frequency energy spectrum has different distribution information characteristics under different blocking conditions, so that the classification and identification accuracy is low. In addition, most of the existing algorithms for researching the fine feature recognition of the frequency hopping radio station are carried out under the conditions of large samples and high signal-to-noise ratio, and the adverse effects of the number of samples and the signal-to-noise ratio on the classification recognition rate cannot be effectively overcome.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a frequency hopping radio station individual identification method, which comprises the following steps:
the first step is as follows: extracting frequency hopping signal time frequency energy spectrum
Assuming a finite set of frequencies for the frequency hopping signal as w, the received signal frequency is set asM is more than or equal to 0 and less than or equal to P-1, a frequency hopping section is divided into P frequency points at equal intervals according to the requirement on time precision, a received frequency hopping signal y is divided into G sections at equal intervals, and each section of data ykHas a length of P, then
yk=y(k·L:k·L+P-1) (1)
Wherein k is more than or equal to 1 and less than or equal to G, and L is a segment interval;
observation matrix Y ═ Y1,y2,...,yG]Can be expressed as
Y=WX+V (2)
Wherein W is [ omega ]0,...,ωP-1],ωi=[ejω1,...,ejωP]X is a time-frequency matrix, V is a noise matrix, and X, V are belonged to CP×GC represents a complex matrix;
due to the sparsity of the frequency hopping signals in the time-frequency domain, time-frequency points in the X are sparse, non-zero points are all on the corresponding rows of each frequency hopping point, and the X is also sparse; therefore, an unconstrained optimization function with a penalty function is assumed:
wherein x iskWhich represents the k-th row of X,denotes the value of X, μ, which minimizes L (X)1Is an X-point sparsity penalty factor, μ2Is an X row sparse penalty factor;
the AL0 algorithm introduces a gaussian function to approximate l0Norm ofWhere s represents an arbitrary D-dimensional vector and σ represents a normal number close to zero; when σ is approximately 0, there are
Wherein s isiRepresents the ith row of vector s, D represents the dimension of vector s; approximating formula (4) to l0The norm represents the sparsity of the time-frequency domain of the frequency hopping signal, and then the sparse reconstruction problem can be converted into the minimization problem for solving the Gaussian sum functionWherein Fσ(x) Indicating a frequency hopping signal0Norm, introducing a penalty factor lambda, and dividing l0The norm minimization problem translates into an optimization problem, i.e.
the steepest descent direction is the direction of the conjugate gradient of the unconstrained optimization function L (x)
Wherein x*Is the conjugate of x; the definition of the conjugate gradient of the complex variable can be obtained
Wherein x isRRepresenting the real part, x, of a signal xIRepresenting the imaginary part of the signal x, i being the unit of an imaginary number, diagonal matrix Ai,i=exp(-|xk|2/2σ2)/σ2;
by substituting the formulae (7) and (8) into the formula (6)
Then the direction of the conjugate gradient of L (x) in equation (3) is calculated as
Wherein A is1=fσ(X)/σ2;A2Is a diagonal matrix whose k-th diagonal element isK is not less than 1 and not more than G, and X (k, k) represents the kth line of X;
the second step is that: fractal feature extraction
(1) Differential box dimension feature extraction
And (3) regarding the time-frequency axis of the frequency-hopping signal time-frequency energy spectrum subjected to sparse reconstruction as a plane, regarding the energy spectrum value as an image gray value, and calculating the difference box dimension of the time-frequency energy spectrum by the following steps:
step 1: assuming that the size of the time-frequency energy spectrum F is NxN, S is a grid of MxM, wherein M is more than or equal to 1 and less than or equal to N/2, and M belongs to Z+,Z+Representing a set of positive integers; f is divided into a plurality of sub-blocks by S, each grid S is provided with a box column with the size of M multiplied by M ', M' is more than or equal to 1 and less than or equal to N/2,e represents the maximum energy spectrum value and,represents a round-down operator;
step 2: setting a (mu, v) th grid, wherein mu is more than or equal to 1 and less than or equal to M, and v is more than or equal to 1 and less than or equal to M; the maximum and minimum energy values of the energy spectrum are denoted g, respectivelymaxAnd gminAt the scale M, the box count of the (μ, ν) th grid is:
step 3: by analogy, the number of all grid boxes under the M scale is calculated
Step 4: changing the scale M of the grid S, repeating the Step1-3, and calculating N under the condition of different block scales MM;
Step 5: estimating differential box dimensions
(2) multi-fractal feature extraction
In order to more effectively represent the local characteristics of the time-frequency energy spectrum, extracting a multi-fractal dimension of the time-frequency energy spectrum as a second-dimensional characteristic;
the fractal MD can be expressed as:
in the formula, q represents a multi-fractal parameter, and in practical application, different choices of q values influence the stability of MD; according to difference boxStep1-3 of the dimension extraction algorithm, calculating N under the condition of different block scales MMThen, combining with the formula (14), obtaining the multi-fractal characteristic of the frequency-hopping signal time-frequency energy spectrum;
(3) time-frequency energy spectrum Rayleigh entropy feature extraction
The time-frequency energy distribution rules of the two frequency hopping radio station signals have certain difference, and the difference of the distribution can be measured by counting the Rayleigh entropy of the time-frequency energy spectrum, namely the difference is converted from the intuitive difference to the difference in numerical value;
let the probability density distribution of the random variable J be P ═ { P ═ P1,p2,...,pTWhere T represents the dimension of the random variable J, satisfying the conditionT is more than or equal to 1 and less than or equal to T, the Rayleigh entropy of J is defined as follows:
wherein alpha is the Rayleigh entropy order, and the Rayleigh entropy is defined by applying one-dimensional probability density distribution in the formula (15);
the continuous form of the rayleigh entropy of order α is as follows:
wherein f (χ, γ) is a continuous two-dimensional probability density distribution, fα(χ, γ) represents f (χ, γ) to the power α;
after the frequency hopping signal y is sparsely reconstructed, the obtained time-frequency energy spectrum P (t, f) has time-frequency edge characteristics and energy retention characteristics, as shown in formulas (17) and (18):
∫P(t,f)df=|y(t)|2,∫P(t,f)dt=|y(f)|2 (17)
wherein y (f) is the Fourier transform of y (t), and f represents the signal frequency; therefore, the two-dimensional probability density distribution f (χ, γ) of the time-frequency energy spectrum P (t, f) and the equation (16) has similar properties, so that the rayleigh entropy of the time-frequency energy spectrum of the frequency hopping signal can be defined by P (t, f) as shown in the formula:
in the formula Pα(t, f) represents P (t, f) to the power of alpha;
the stabilizing condition of formula (19) is ^ jeppeα(t, f) dtdf > 0; for convenience of calculation, the discrete expression of the rayleigh entropy of the time-frequency energy spectrum is as follows:
wherein X is a time-frequency matrix, k and k 'represent the number of sampling points of observation data, k is more than or equal to 1 and less than or equal to G, and k' is more than or equal to 1 and less than or equal to G;
psi and psi' denote frequency set dictionary, omega0≤ψ′≤ωP-1,ω0≤ψ≤ωP-1;
And finally, forming a characteristic vector V [ FD, MD (q), R from the three characteristics of the Rayleigh entropy, the difference box dimension and the multi-fractal dimension of the calculated time-frequency energy spectrumα]And carrying out classification and identification by using a support vector machine classifier.
In the individual identification method of the frequency hopping radio station, the AL0 algorithm comprises the following steps:
inputting: matrix W, measured value vector y;
step (1): let initial value x(0)=WT(WWT)-1y;
Step (2): selection of descending sequence [ sigma ]1σ2...σJ]J ═ 6; let the convergence criterion be ε, ε representing a mean of 0 and a variance of σ2White gaussian noise of (1);
step (3): iterating an algorithm;
(1) j takes the values of 1, 2,.. J in sequence;
(2) for each value of j, let σ be σ ═ σj,
(6) At this time, let x(j)Continuing to take the next value of j as x, and carrying out iterative operation until all values are taken in sequence;
In a specific embodiment of the present invention, σ takes on values of [2, 1, 0.5, 0.2, 0.1, 0.05] in sequence.
The method provided by the invention has the advantages that the recognition accuracy is not influenced by the change of the number of training samples and is always kept at a higher level. The main reason is that the algorithm respectively extracts the difference box dimension, the multi-fractal dimension and the Rayleigh entropy of the time-frequency energy spectrum as the feature vectors, quantitatively describes the energy change of the time-frequency energy spectrum, the complexity and the regularity of the time-frequency energy spectrum, and avoids the problem of misjudgment caused by the similarity of single features. The classification features of the method are only slightly overlapped, and the method has obvious clustering effect.
Drawings
FIG. 1 shows a flow chart of the method of the present invention;
FIG. 2 shows a time-frequency energy diagram for two frequency hopping stations; wherein 2(a) shows a first station time-frequency energy graph, and 2(b) shows a second station time-frequency energy graph;
FIG. 3 illustrates a fractal variation law;
FIG. 4 shows the law of change of time-frequency energy spectrum entropy with order;
FIG. 5 shows the recognition rate as a function of the number of training samples;
fig. 6 shows a classification effect diagram.
Detailed Description
The technical scheme and the implementation process of the invention are described in detail by combining specific examples.
The process flow of the method of the invention is shown in figure 1.
First by approximating l0And obtaining a time-frequency energy spectrum curved surface of the frequency hopping signal by a norm (AL0) algorithm, and then extracting a difference box dimension, a multi-fractal dimension and a time-frequency Rayleigh entropy under different block scales to form a feature vector. And finally, training, identifying and classifying through a support vector machine classifier.
The first step is as follows: extracting frequency hopping signal time frequency energy spectrum
Assuming a finite set of frequencies for the frequency hopping signal as w, the received signal frequency is set asM is more than or equal to 0 and less than or equal to P-1, according to the requirement of the invention on time precision, the frequency hopping section is divided into P frequency points at equal intervals, the received frequency hopping signal y is divided into G sections at equal intervals, and each section of data ykHas a length of P, then
yk=y(k·L:k·L+P-1) (1)
Wherein k is more than or equal to 1 and less than or equal to G, and L is a segmentation interval.
Observation matrix Y ═ Y1,y2,...,yG]Can be expressed as
Y=WX+V (2)
Wherein W is [ omega ]0,...,ωP-1],ωi=[ejω1,...,ejωP]X is a time-frequency matrix, V is a noise matrix, and X, V are belonged to CP×GC represents a complex momentAnd (5) arraying.
Due to the sparsity of the frequency hopping signals in the time-frequency domain, time-frequency points in X are sparse, non-zero points are all on the corresponding rows of the frequency hopping points, and X is also sparse. An unconstrained optimization function with a penalty function can therefore be assumed:
wherein x iskWhich represents the k-th row of X,denotes the value of X, μ, which minimizes L (X)1Is an X-point sparsity penalty factor, μ2Is an X-row sparse penalty factor.
The AL0 algorithm introduces a gaussian function to approximate l0Norm ofWhere s represents an arbitrary D-dimensional vector and σ represents a normal number close to zero. When σ is approximately 0, there are
Wherein s isiThe ith row of the vector s is indicated and D indicates the dimension of the vector s. Can approximate the formula (4) to l0The norm represents the sparsity of the time-frequency domain of the frequency hopping signal, and then the sparse reconstruction problem can be converted into the minimization problem for solving the Gaussian sum functionWherein Fσ(x) Indicating a frequency hopping signal0Norm, introducing a penalty factor lambda, and dividing l0The norm minimization problem translates into an optimization problem, i.e.
The steepest descent direction is the direction of the conjugate gradient of the unconstrained optimization function L (x)
Wherein x*Is the conjugate of x. The definition of the conjugate gradient of the complex variable can be obtained
Wherein x isRRepresenting the real part, x, of a signal xIRepresenting the imaginary part of the signal x, i being the unit of an imaginary number, diagonal matrix Λi,i=exp(-|xk|2/2σ2)/σ2。
By substituting the formulae (7) and (8) into the formula (6)
Then the direction of the conjugate gradient of L (x) in equation (3) is calculated as
Wherein, Λ1=fσ(X)/σ2;Λ2Is a diagonal matrix whose k-th diagonal element isK is not less than 1 and not more than G, and X (k,: represents the kth line of X. The detailed algorithm steps of the AL0 algorithm are as follows, where σ takes the values of [2, 1, 0.5, 0.2, 0.1, 0.05]:
Inputting: matrix W, measurement vector y.
Step 1: let initial value x(0)=WT(WWT)-1y。
Step 2: selection of descending sequence [ sigma ]1σ2...σJ]In one embodiment of the invention is the sequence [2, 1, 0.5, 0.2, 0.1, 0.05]]And J is 6. Let the convergence criterion be ε, ε representing a mean of 0 and a variance of σ2White gaussian noise.
Step 3: and (6) iterating the algorithm.
(1) J takes the values of 1, 2,.. J in sequence;
(2) for each value of j, let σ be σ ═ σj,
(6) At this time, let x(j)And continuing to take the next value of j and carrying out iterative operation until the values are sequentially subjected to the iterative operationAll values are taken out.
The time-frequency energy spectrum of two synchronous networking frequency hopping radio stations obtained through sparse reconstruction of the AL0 algorithm is shown in figure 2.
As can be seen from fig. 2, two frequency hopping stations of the same type and operation carry respective fine feature information due to their different radiation sources. The individual identification of the frequency hopping radio station can be realized by extracting the characteristic rules of energy change, frequency point distribution and the like of the time-frequency energy spectrum of each radio station signal.
The second step is that: extracting fractal characteristics;
(1) differential box dimension feature extraction
And (3) regarding the time-frequency axis of the frequency-hopping signal time-frequency energy spectrum subjected to sparse reconstruction as a plane, regarding the energy spectrum value as an image gray value, and calculating the difference box dimension of the time-frequency energy spectrum by the following steps:
step 1: assuming that the size of the time-frequency energy spectrum F is NxN, S is a grid of MxM, wherein M is more than or equal to 1 and less than or equal to N/2, and M belongs to Z+,Z+Representing a set of positive integers. F is divided into a plurality of sub-blocks by S, each grid S is provided with a box column with the size of M multiplied by M ', M' is more than or equal to 1 and less than or equal to N/2,e represents the maximum energy spectrum value and,indicating the rounding-down operator.
Step 2: setting the (mu, v) th grid, wherein mu is more than or equal to 1 and less than or equal to M, and v is more than or equal to 1 and less than or equal to M. The maximum and minimum energy values of the energy spectrum are denoted g, respectivelymaxAnd gminAt the scale M, the box count of the (μ, ν) th grid is:
step 3: by analogy, the number of all grid boxes under the M scale is calculated
Step 4: changing the scale M of the grid S, repeating the Step1-3, and calculating N under the condition of different block scales MM。
Step 5: estimating differential box dimensions
(2) multi-fractal feature extraction
Many scholars find that when single fractal describes most of objectively existing fractal objects, the complexity and nonlinear characteristics of the fractal objects cannot be completely measured, and if the overall characteristics of a system are described by using the single fractal dimension, all the characteristics of the fractal objects cannot be fully reflected to a certain extent. The multi-fractal can study the overall characteristics of the multi-fractal from local parts, and improves the fine degree of describing the geometric characteristics and local scale behaviors of the object. In order to more effectively represent the local characteristics of the time-frequency energy spectrum, the multi-fractal dimension of the time-frequency energy spectrum is extracted as a second-dimensional characteristic.
The fractal MD can be expressed as:
in the formula, q represents a multi-fractal parameter, and in practical application, different choices of q values influence the stability of MD. According to Step1-3 of the difference box dimension extraction algorithm, calculating N under the condition of different block scales MMThen, combining with the formula (14), obtaining the multi-fractal characteristic of the frequency-hopping signal time-frequency energy spectrum。
(3) Time-frequency energy spectrum Rayleigh entropy feature extraction
Entropy can be defined as a quantity that measures chaotic irregularities, uncertainty, etc. chaotic states. As can be seen from fig. 2, the time-frequency energy distribution rules of the two frequency hopping radio station signals have a certain difference, and the difference of the distribution can be measured by counting the rayleigh entropy of the time-frequency energy spectrum, that is, the difference is converted from the intuitive difference to the numerical difference.
Let the probability density distribution of the random variable J be P ═ { P ═ P1,p2,...,pTWhere T represents the dimension of the random variable J, satisfying the conditionT is more than or equal to 1 and less than or equal to T, the Rayleigh entropy of J is defined as follows:
where α is the rayleigh entropy order, and equation (15) defines rayleigh entropy using a one-dimensional probability density distribution.
The continuous form of the rayleigh entropy of order α is as follows:
wherein f (χ, γ) is a continuous two-dimensional probability density distribution, fα(χ, γ) represents f (χ, γ) to the α -power.
After the frequency hopping signal y is sparsely reconstructed, the obtained time-frequency energy spectrum P (t, f) has time-frequency edge characteristics and energy retention characteristics, as shown in formulas (17) and (18):
∫P(t,f)df=|y(t)|2,∫P(t,f)dt=|y(f)|2 (17)
where y (f) is the Fourier transform of y (t), and f represents the signal frequency. Therefore, the two-dimensional probability density distribution f (χ, γ) of the time-frequency energy spectrum P (t, f) and the equation (16) has similar properties, so that the rayleigh entropy of the time-frequency energy spectrum of the frequency hopping signal can be defined by P (t, f) as shown in the formula:
in the formula Pα(t, f) represents P (t, f) to the power of alpha.
The stabilizing condition of formula (19) is ^ jeppeα(t, f) dtdf > 0. For convenience of calculation, the discrete expression of the rayleigh entropy of the time-frequency energy spectrum is as follows:
wherein X is a time-frequency matrix, k and k 'represent the number of sampling points of observation data, k is more than or equal to 1 and less than or equal to G, and k' is more than or equal to 1 and less than or equal to G;
psi and psi' denote frequency set dictionary, omega0≤ψ′≤ωP-1,ω0≤ψ≤ωP-1。
And finally, forming a characteristic vector V [ FD, MD (q), R from the three characteristics of the Rayleigh entropy, the difference box dimension and the multi-fractal dimension of the calculated time-frequency energy spectrumα]And carrying out classification and identification by using a support vector machine classifier.
Example testing
The experimental data of the invention is collected from four frequency hopping radio stations with the same model, the working frequency of the radio stations is set to 150MHz, the frequency hopping bandwidth is 6.4MHz, the sampling rate is 600MHz, and the sampling duration is 5 seconds. Considering that when the time-frequency energy spectrum after sparse reconstruction is partitioned, if the value of the dimension L is too large, the boundary information of the energy spectrum cannot be completely utilized, so that the extracted features cannot completely reflect the essential information of the signal; if the value of the scale L is too small, the amount of calculation increases. Therefore, in the simulation experiment of the invention, the block scale L takes five values of 4, 8, 12, 16 and 20.
As can be seen from equation (14), the parameter q affects the stability of the md (q) values, and md (q) under different q values is calculated by using the data samples of the frequency hopping signal of one of the stations, and the change rule is shown in fig. 3.
As can be seen from FIG. 3, MD (q) is less affected by the value of q only when 10 < q. Through repeated experimental observation, the md (q) values of the hopping signals of the 4 stations all become stable when q is 13, and q is 13 in order to ensure the stability of the extracted features.
The rayleigh entropy of the time-frequency energy spectrum can well reflect the time-frequency energy spectrum rule and complexity of each frequency hopping signal, but as can be known from the formula (20), the extraction of the rayleigh entropy characteristics of the time-frequency energy spectrum is greatly influenced by the value of the order alpha. The invention selects the frequency hopping signal sample data of 4 radio stations, and respectively calculates the Rayleigh entropy of the energy spectrum of the integer order of 0-30 of the order alpha, as shown in figure 4.
As can be seen from fig. 4, as the order α increases, the rayleigh entropy of the time-frequency energy spectrum of the hopping signal of the 4 radio stations tends to become substantially smaller, and the rayleigh entropy of the time-frequency energy spectrum of the hopping signal of the 4 radio stations gradually converges to the same value. When alpha is 14, the rayleigh entropy of the time-frequency energy spectrum of the hopping signals of different stations has a certain discrimination, and when alpha is more than 16, the discrimination becomes more and more fuzzy. Fig. 4 shows that when α is 3, the discrimination is the best, and as the order α increases, the discrimination is continuously reduced, so the invention selects the rayleigh entropy of the 3-order time-frequency energy spectrum with higher discrimination as the characteristic of individual identification of the frequency hopping radio station.
Respectively intercepting 4 radio station frequency hopping signal data, starting from the head of the sampled data, intercepting 1024 sampling points as training samples at intervals of 200 sampling points, and intercepting 200 sections. Then, in the same manner, 200 pieces of data are cut out as test samples from the end of the sampling data. Under the condition of different training sample numbers, the algorithm is adopted to extract three characteristics of the time-frequency energy spectrum Rayleigh entropy, the difference box dimension and the multi-fractal dimension, then a support vector machine classifier is used for classification and identification, and the average identification accuracy of 50 experiments is counted as shown in figure 5.
As can be seen from FIG. 5, the recognition accuracy of the algorithm of the present invention is not affected by the variation of the number of training samples, and is always kept at 85.6%. The main reason is that the algorithm respectively extracts the difference box dimension, the multi-fractal dimension and the Rayleigh entropy of the time-frequency energy spectrum as the feature vectors, quantitatively describes the energy change of the time-frequency energy spectrum, the complexity and the regularity of the time-frequency energy spectrum, and avoids the problem of misjudgment caused by the similarity of single features.
In order to facilitate visualization of the classification effect, the number of the training samples is 200, and the classification effect of the three-dimensional features is drawn by using the algorithm of the invention and is shown in fig. 6.
As can be seen from FIG. 6, the classification features of the algorithm of the present invention are only rarely partially overlapped, and have an obvious clustering effect, which is in accordance with the recognition result of FIG. 5.
Claims (3)
1. A frequency hopping radio station individual identification method comprises the following steps:
the first step is as follows: extracting frequency hopping signal time frequency energy spectrum
Assuming a finite set of frequencies for the frequency hopping signal as w, the received signal frequency is set asM is more than or equal to 0 and less than or equal to P-1, a frequency hopping section is divided into P frequency points at equal intervals according to the requirement on time precision, a received frequency hopping signal y is divided into G sections at equal intervals, and each section of data ykHas a length of P, then
yk=y(k·L:k·L+P-1) (1)
Wherein k is more than or equal to 1 and less than or equal to G, and L is a segment interval;
observation matrix Y ═ Y1,y2,...,yG]Can be expressed as
Y=WX+V (2)
Wherein W is [ omega ]0,...,ωP-1],ωi=[ejω1,...,ejωP]X is a time-frequency matrix, V is a noise matrix, and X, V are belonged to CP ×GC represents a complex matrix;
due to the sparsity of the frequency hopping signals in the time-frequency domain, time-frequency points in the X are sparse, non-zero points are all on the corresponding rows of each frequency hopping point, and the X is also sparse; therefore, an unconstrained optimization function with a penalty function is assumed:
wherein x iskWhich represents the k-th row of X,denotes the value of X, μ, which minimizes L (X)1Is an X-point sparsity penalty factor, μ2Is an X row sparse penalty factor;
the AL0 algorithm introduces a gaussian function to approximate l0Norm ofWhere s represents an arbitrary D-dimensional vector and σ represents a normal number close to zero; when σ is approximately 0, there are
Wherein s isiRepresents the ith row of vector s, D represents the dimension of vector s; approximating formula (4) to l0The norm represents the sparsity of the time-frequency domain of the frequency hopping signal, and then the sparse reconstruction problem can be converted into the minimization problem for solving the Gaussian sum functionWherein Fσ(x) Indicating a frequency hopping signal0Norm, introducing a penalty factor lambda, and dividing l0The norm minimization problem translates into an optimization problem, i.e.
the steepest descent direction is the direction of the conjugate gradient of the unconstrained optimization function L (x)
Wherein x*Is the conjugate of x; the definition of the conjugate gradient of the complex variable can be obtained
Wherein x isRRepresenting the real part, x, of a signal xIRepresenting the imaginary part of the signal x, i being the unit of an imaginary number, diagonal matrix Λi,i=exp(-|xk|2/2σ2)/σ2;
by substituting the formulae (7) and (8) into the formula (6)
Then the direction of the conjugate gradient of L (x) in equation (3) is calculated as
Wherein, Λ1=fσ(X)/σ2;Λ2Is a diagonal matrix whose k-th diagonal element isK is not less than 1 and not more than G, and X (k, k) represents the kth line of X;
the second step is that: fractal feature extraction
(1) Differential box dimension feature extraction
And (3) regarding the time-frequency axis of the frequency-hopping signal time-frequency energy spectrum subjected to sparse reconstruction as a plane, regarding the energy spectrum value as an image gray value, and calculating the difference box dimension of the time-frequency energy spectrum by the following steps:
step 1: assuming that the size of the time-frequency energy spectrum F is NxN, S is a grid of MxM, wherein M is more than or equal to 1 and less than or equal to N/2, and M belongs to Z+,Z+Representing a set of positive integers; f is divided into a plurality of sub-blocks by S, each grid S is provided with a box column with the size of M multiplied by M ', M' is more than or equal to 1 and less than or equal to N/2,e represents the maximum energy spectrum value and,represents a round-down operator;
step 2: setting the (mu, v) grid, wherein mu is more than or equal to 1 and less than or equal to M, and v is more than or equal to 1 and less than or equal to M; the maximum and minimum energy values of the energy spectrum are denoted g, respectivelymaxAnd gminAt scale M, the number of boxes in the (μ, v) th grid is:
step 3: by analogy, the number of all grid boxes under the M scale is calculated
Step 4: changing the scale M of the grid S, repeating the Step1-3, and calculating N under the condition of different block scales MM;
Step 5: estimating differential box dimensions
(2) multi-fractal feature extraction
In order to more effectively represent the local characteristics of the time-frequency energy spectrum, extracting a multi-fractal dimension of the time-frequency energy spectrum as a second-dimensional characteristic;
the fractal MD can be expressed as:
in the formula, q represents a multi-fractal parameter, and in practical application, different choices of q values influence the stability of MD; according to Step1-3 of the difference box dimension extraction algorithm, calculating N under the condition of different block scales MMThen, combining with the formula (14), obtaining the multi-fractal characteristic of the frequency-hopping signal time-frequency energy spectrum;
(3) time-frequency energy spectrum Rayleigh entropy feature extraction
The time-frequency energy distribution rules of the two frequency hopping radio station signals have certain difference, and the difference of the distribution can be measured by counting the Rayleigh entropy of the time-frequency energy spectrum, namely the difference is converted from the intuitive difference to the difference in numerical value;
let the probability density distribution of the random variable J be P ═ { P ═ P1,p2,...,pTWhere T represents the dimension of the random variable J, satisfying the conditionT is more than or equal to 1 and less than or equal to T, the Rayleigh entropy of J is defined as follows:
wherein alpha is the Rayleigh entropy order, and the Rayleigh entropy is defined by applying one-dimensional probability density distribution in the formula (15);
the continuous form of the rayleigh entropy of order α is as follows:
wherein f (χ, γ) is a continuous two-dimensional probability density distribution, fα(χ, γ) represents f (χ, γ) to the power α;
after the frequency hopping signal y is sparsely reconstructed, the obtained time-frequency energy spectrum P (t, f) has time-frequency edge characteristics and energy retention characteristics, as shown in formulas (17) and (18):
∫P(t,f)df=|y(t)|2,∫P(t,f)dt=|y(f)|2 (17)
wherein y (f) is the Fourier transform of y (t), and f represents the signal frequency; therefore, the two-dimensional probability density distribution f (χ, γ) of the time-frequency energy spectrum P (t, f) and the equation (16) has similar properties, so that the rayleigh entropy of the time-frequency energy spectrum of the frequency hopping signal can be defined by P (t, f) as shown in the formula:
in the formula Pα(t, f) represents P (t, f) to the power of alpha;
the stabilizing condition of formula (19) is ^ jeppeα(t, f) dtdf > 0; for convenience of calculation, the discrete expression of the rayleigh entropy of the time-frequency energy spectrum is as follows:
wherein X is a time-frequency matrix, k and k 'represent the number of sampling points of observation data, k is more than or equal to 1 and less than or equal to G, and k' is more than or equal to 1 and less than or equal to G; psi and psi' denote frequency set dictionary, omega0≤ψ′≤ωP-1,ω0≤ψ≤ωP-1;
And finally, forming a characteristic vector V [ FD, MD (q), R from the three characteristics of the Rayleigh entropy, the difference box dimension and the multi-fractal dimension of the calculated time-frequency energy spectrumα]And carrying out classification and identification by using a support vector machine classifier.
2. The individual identification method of frequency hopping radio of claim 1, wherein AL0 algorithm steps are as follows:
inputting: matrix W, measured value vector y;
step (1): let initial value x(0)=WT(WWT)-1y;
Step (2): selection of descending sequence [ sigma ]1σ2...σJ]J ═ 6; let the convergence criterion be ε, ε representing a mean of 0 and a variance of σ2White gaussian noise of (1);
step (3): iterating an algorithm;
(1) j takes the values of 1, 2,.. J in sequence;
(2) for each value of j, let σ be σ ═ σj,
(6) At this time, let x(j)Continuing to take the next value of j as x, and carrying out iterative operation until all values are taken in sequence;
3. The individual identification method of a frequency hopping radio station as claimed in claim 2, wherein σ takes values of [2, 1, 0.5, 0.2, 0.1, 0.05] in sequence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811328578.4A CN109472239B (en) | 2018-10-28 | 2018-10-28 | Individual identification method of frequency hopping radio station |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811328578.4A CN109472239B (en) | 2018-10-28 | 2018-10-28 | Individual identification method of frequency hopping radio station |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109472239A CN109472239A (en) | 2019-03-15 |
CN109472239B true CN109472239B (en) | 2021-10-01 |
Family
ID=65672201
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811328578.4A Active CN109472239B (en) | 2018-10-28 | 2018-10-28 | Individual identification method of frequency hopping radio station |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109472239B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP4227691A1 (en) * | 2022-02-10 | 2023-08-16 | Rohde & Schwarz GmbH & Co. KG | Method of classifying a radio frequency signal |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110472584A (en) * | 2019-08-16 | 2019-11-19 | 四川九洲电器集团有限责任公司 | A kind of communication equipment personal identification method, electronic equipment and computer program product |
CN112994740B (en) * | 2021-04-23 | 2021-07-23 | 成都天锐星通科技有限公司 | Frequency hopping signal parameter estimation method and device, electronic equipment and readable storage medium |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102437984A (en) * | 2011-11-07 | 2012-05-02 | 哈尔滨工程大学 | Modulation signal identification method based on complexity characteristic under low signal-to-noise ratio condition |
CN203151487U (en) * | 2013-04-11 | 2013-08-21 | 中国人民解放军装甲兵工程学院 | Performance test system for frequency hopping radio station |
CN103340625A (en) * | 2013-06-18 | 2013-10-09 | 中国人民解放军第四军医大学 | Regularization method of fast optimization in electrical impedance tomography |
CN103959839A (en) * | 2011-08-12 | 2014-07-30 | 黑莓有限公司 | Methods of channel state information feedback and transmission in coordinated multi-point wireless communications system |
CN104485979A (en) * | 2014-12-09 | 2015-04-01 | 西安电子科技大学 | Blind estimation method for underdetermined hybrid frequency hopping parameters based on time frequency diagram correction |
CN105891825A (en) * | 2016-03-29 | 2016-08-24 | 西安电子科技大学 | Multiple-input multiple-output array radar staring imaging method based on tensor compression perception |
CN106908774A (en) * | 2017-01-06 | 2017-06-30 | 南京航空航天大学 | Based on the sparse one-dimensional range profile recognition methods for keeping projecting of multiple dimensioned core |
CN107526074A (en) * | 2017-07-19 | 2017-12-29 | 上海无线电设备研究所 | A kind of distance of sparse Frequency Hopping Signal and Speed Two Dimensions high resolution processing method |
CN107894591A (en) * | 2017-09-30 | 2018-04-10 | 沈阳航空航天大学 | Through-wall radar diffraction tomography method based on compressed sensing |
CN108700652A (en) * | 2015-12-09 | 2018-10-23 | 欧利景无线有限公司 | The methods, devices and systems for detecting and monitoring for wireless event |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140181166A1 (en) * | 2012-12-26 | 2014-06-26 | Industrial Technology Research Institute | Apparatus for low complexity sub-nyquist sampling of sparse wideband signals |
-
2018
- 2018-10-28 CN CN201811328578.4A patent/CN109472239B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103959839A (en) * | 2011-08-12 | 2014-07-30 | 黑莓有限公司 | Methods of channel state information feedback and transmission in coordinated multi-point wireless communications system |
CN102437984A (en) * | 2011-11-07 | 2012-05-02 | 哈尔滨工程大学 | Modulation signal identification method based on complexity characteristic under low signal-to-noise ratio condition |
CN203151487U (en) * | 2013-04-11 | 2013-08-21 | 中国人民解放军装甲兵工程学院 | Performance test system for frequency hopping radio station |
CN103340625A (en) * | 2013-06-18 | 2013-10-09 | 中国人民解放军第四军医大学 | Regularization method of fast optimization in electrical impedance tomography |
CN104485979A (en) * | 2014-12-09 | 2015-04-01 | 西安电子科技大学 | Blind estimation method for underdetermined hybrid frequency hopping parameters based on time frequency diagram correction |
CN108700652A (en) * | 2015-12-09 | 2018-10-23 | 欧利景无线有限公司 | The methods, devices and systems for detecting and monitoring for wireless event |
CN105891825A (en) * | 2016-03-29 | 2016-08-24 | 西安电子科技大学 | Multiple-input multiple-output array radar staring imaging method based on tensor compression perception |
CN106908774A (en) * | 2017-01-06 | 2017-06-30 | 南京航空航天大学 | Based on the sparse one-dimensional range profile recognition methods for keeping projecting of multiple dimensioned core |
CN107526074A (en) * | 2017-07-19 | 2017-12-29 | 上海无线电设备研究所 | A kind of distance of sparse Frequency Hopping Signal and Speed Two Dimensions high resolution processing method |
CN107894591A (en) * | 2017-09-30 | 2018-04-10 | 沈阳航空航天大学 | Through-wall radar diffraction tomography method based on compressed sensing |
Non-Patent Citations (7)
Title |
---|
Frequency hopping radio individual identification based on energy spectrum blended subtle characteristics;Yang Xin 等;《Journal of Physics: Conference Series》;20190707;第1325卷;第1-7页 * |
Specific emitter identification based on normalized frequency spectrum;Degang Sun 等;《2016 2nd IEEE International Conference on Computer and Communications》;20161017;第1192-1205页 * |
Specific Emitter Identification via Hilbert–Huang Transform in Single-Hop and Relaying Scenarios;Jingwen Zhang 等;《IEEE Transactions on Information Forensics and Security》;20160630;第11卷(第6期);第1192-1205页 * |
Structure-Aware Bayesian Compressive Sensing for Frequency-Hopping Spectrum Estimation With Missing Observations;Shengheng Liu 等;《IEEE Transactions on Signal Processing》;20180415;第66卷(第8期);第2153-2166页 * |
块稀疏贝叶斯模型下的跳频信号时频分析;李雷;《信号处理》;20180131;第34卷(第01期);第107-113页 * |
基于ST-RFT算法的信号调制方式识别与参数估计方法研究;刘丹;《中国优秀硕士学位论文全文数据库 信息科技辑》;20171015(第10期);第I136-60页 * |
基于时频能量谱特征的跳频电台个体识别;杨鑫 等;《信号处理》;20191031;第35卷(第10期);第1671-1679页 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP4227691A1 (en) * | 2022-02-10 | 2023-08-16 | Rohde & Schwarz GmbH & Co. KG | Method of classifying a radio frequency signal |
Also Published As
Publication number | Publication date |
---|---|
CN109472239A (en) | 2019-03-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109472239B (en) | Individual identification method of frequency hopping radio station | |
CN105137498B (en) | A kind of the underground objects detection identifying system and method for feature based fusion | |
Wang et al. | Fractal complexity-based feature extraction algorithm of communication signals | |
CN109307862A (en) | A kind of target radiation source individual discrimination method | |
Iorkyase et al. | Improving RF-based partial discharge localization via machine learning ensemble method | |
Liu et al. | Deep learning and recognition of radar jamming based on CNN | |
CN114201988A (en) | Satellite navigation composite interference signal identification method and system | |
CN112213688B (en) | Feature extraction method for identifying low-altitude airspace low-small slow aircraft target individuals | |
CN108734228A (en) | The polarimetric SAR image random forest classification method of comprehensive multiple features | |
CN111680737B (en) | Radar radiation source individual identification method under differential signal-to-noise ratio condition | |
CN112859014A (en) | Radar interference suppression method, device and medium based on radar signal sorting | |
CN116047427B (en) | Small sample radar active interference identification method | |
CN110244275A (en) | The reconstruct of sea clutter bispectrum and emulation mode | |
CN115034261A (en) | Method and equipment for extracting features between pulses of radar radiation source signal and storage medium | |
Wang et al. | Classification methods with signal approximation for unknown interference | |
CN109446910A (en) | A kind of communication emitter Signals classifying identification method | |
CN115951315B (en) | Radar spoofing interference identification method and system based on improved wavelet packet energy spectrum | |
Liu et al. | Graph signal coarsening: Dimensionality reduction in irregular domain | |
CN114841195B (en) | Avionics space signal modeling method and system | |
Yang et al. | Radar target recognition using contourlet packet transform and neural network approach | |
Xu et al. | Individual identification of electronic equipment based on electromagnetic fingerprint characteristics | |
CN112213698B (en) | Deception jamming identification method based on sparse representation classification | |
Shreyamsha Kumar et al. | Target identification using harmonic wavelet based ISAR imaging | |
Melinda et al. | Implementation of segmentation scheme based on wavelet transform in multi-spectral fluctuation patterns | |
Shan et al. | Wavelet based recognition for pulsar signals |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |