Invention content
The present invention is directed to the deficiency in background technology, it is proposed that a kind of by JADE (eigenmatrix Joint diagonalization) blind source point
The main lobe anti-interference method being combined with CLEAN algorithms from algorithm.This method estimates interference signal first with JADE algorithms
Steering vector and waveform, and then reconstruct interference array signal;Then the MUSIC spectrums for receiving signal are calculated, obtain receiving letter
Number MUSIC spectrum, and calculate reconstruct come interference array signal MUSIC spectrum;It will be strong finally by spatial domain CLEAN algorithms
Interference signal eliminates in the MUSIC spectrums for receiving signal, to obtain direction of arrival (DOA) estimation of echo signal.Emulation
The validity of this method is shown, and the algorithm does not need the prior information of interference signal, and can inhibit a plurality of types of
Interference has general applicability.
The present invention provides a kind of main lobe anti-interference methods of joint JADE and CLEAN, it includes the following steps:
Step 1:If there is M echo signal being mutually independent and P high-power interference signals in space, and target
The direction of arrival difference of signal and interference signal is incident on space an array within the scope of main lobe angle, the array by
L array element composition, and it is far field narrow band signal to assume echo signal and interference signal all, first of array element t moment receives signal
For:
Wherein, sm(t), m=1,2 ..., M are m-th of echo signal, θTmDirection of arrival for m-th of echo signal is
DOA angles, Jp(t), p=1,2 ..., P are p-th of interference signal, θJpFor direction of arrival, that is, angles DOA of p-th of interference signal
Degree, n (t) indicate that t moment noise signal, T indicate that sampling sum, d indicate array element spacing, and λ is operation wavelength,
Then antenna array receiver signal is:
X (t)=[x1(t),x2(t),…,xL(t)]T, t=1,2 ..., T (2)
Wherein, ()TIndicate that transposition operator, T indicate sampling sum;
Step 2:Calculate the MUSIC spectrums of antenna array receiver signal X (t):
Step 2-1:Calculate the spatial correlation matrix for receiving signal
Wherein, ()HIndicate conjugate transposition operation symbol;
Step 2-2:To correlation matrixEigenvalues Decomposition is done, and characteristic value is ranked sequentially by monotonic increase, by rear L-
The corresponding feature vector u of M-P characteristic valueM+P+1,uM+P+2,…,uLConstitute matrix G:
G=[uM+P+1 uM+P+2 … uL] (4)
Step 2-3:The MUSIC of X (t) composes formula:
Wherein,For array steering vector, d is array element spacing, and λ is work
Wavelength;Due to not knowing the direction of arrival information of target source, space angle θ is dividedK
The grid number of representation space angular divisions calculates function 1/a successivelyH(θ)GGHThe value of a (θ) obtains the MUSIC spectrums P of X (t)X
(θ);
Step 3:Believed with interference using the blind source separation algorithm separation echo signal of eigenmatrix Joint diagonalization (JADE)
Number:
Step 3-1:It docks collection of letters X (t) and carries out prewhitening, obtain whitened signal Z (t), i.e.,:
Z (t)=WX (t) (6)
Wherein, W is whitening matrix;
Step 3-2:Seek the fourth order cumulant matrix Q of whitened signal Z (t)z:
Wherein E [] indicates operation of averaging, zi(t) the i-th row of whitened signal Z (t) is indicated, i, j, k, l are belonging respectively to 1
~M+P;To QzEigenvalues Decomposition is carried out, M+P maximum eigenvalue λ before obtaining1,λ2,…,λM+PFeature vector corresponding with its
v1,v2,…,vM+P, wherein vi, i=1,2 ..., M+P is (M+P)2× 1 dimension column vector, therefore obtain needing approximate joint diagonal
Objective matrix { the M of change1,M2,…,MM+P};Wherein Vec (Mi)=λivi, i=1,2 ..., M+P, Vec () expression vectorization calculations
Symbol, i.e., line up column vector by a matrix column vector according to the ordering in matrix;
Step 3-3:Using unitary matrice V to { M1,M2,…,MM+PCarry out approximately joint diagonalization;
Step 3-4:Separation signal is obtained with array manifold to estimate:
Wherein, W#For the pseudoinverse of whitening matrix W;Y (t) is separation signal, including echo signal waveform is estimatedWith interference
Signal waveform is estimated Estimate for array manifold, includes the array manifold estimation of echo signalWith interference signal
Array manifold is estimated
Step 4:Assuming that the interference signal waveform that step 3 estimates isInterference array manifold beThen weigh
Structure receive signal in interference component be:
Wherein,For the interference component estimated;
Step 5:The MUSIC that target is calculated using CLEAN algorithms on spatial domain is composed:
Step 5-1:The MUSIC spectrums that the interference component that step 4 estimates is calculated by the method for step 2, obtain interfering into
The MUSIC divided composes PJ(θ);
Step 5-2:The MUSIC spectrums of target in order to obtain, need to minimize cost function:
The P obtained using step 5-1JThe P that (θ) and step 2 obtainX(θ) calculates weight coefficient
Step 5-3:The weight coefficient obtained using step 4-2The MUSIC of calculating target, which is composed, is:
Wherein, PT(θ) is the target MUSIC spectrums after AF panel;The corresponding θ of preceding M maximum value difference is found out, is
The direction of arrival of target.
The beneficial effects of the invention are as follows
The present invention proposes a kind of main lobe anti-interference method of joint JADE and CLEAN, and the method is dry compared to existing
Restrainable algorithms are disturbed, the prior information of interference is required no knowledge about, are applicable to a plurality of types of interference, and realize to target
DOA estimation.
The invention firstly uses steering vectors and waveform that JADE algorithms estimate interference signal, and then reconstruct interference battle array
Column signal;Then the MUSIC spectrums for calculating the interference array signal that reconstruct obtains obtain the only MUSIC comprising interference signal and compose;
Finally the MUSIC spectrums of interference signal are offseted in the MUSIC spectrums for receiving signal using CLEAN algorithms on spatial domain, to
To the Mutual coupling of echo signal.Simulation result shows that this method can be very good to complete AF panel, and by target
DOA is estimated.
Specific implementation mode
Step 1:
If there is M echo signal being mutually independent and P high-power interference signals in space, and echo signal and
The direction of arrival difference of interference signal is incident on space an array within the scope of main lobe angle, and the array is by L battle array
Member composition, and it is far field narrow band signal to assume echo signal and interference signal all, first of array element t moment receives signal and is:
Wherein, sm(t), m=1,2 ..., M are m-th of echo signal, θTmFor the direction of arrival of corresponding echo signal,
Jp(t), p=1,2 ..., P are the P interference signal, θJpFor the direction of arrival of corresponding interference signal, n (t) indicates that t moment is made an uproar
Acoustical signal, T indicate that sampling sum, d indicate array element spacing, and λ is operation wavelength.
Then antenna array receiver signal is:
X (t)=[x1(t),x2(t),…,xL(t)]T, t=1,2 ..., T (10)
Wherein, ()TIndicate that transposition operator, T indicate sampling sum.
Step 2:Calculate the MUSIC spectrums of antenna array receiver signal X (t):
Step 2-1:Calculate the spatial correlation matrix for receiving signal
Wherein, ()HIndicate conjugate transposition operation symbol.
Step 2-2:To correlation matrixEigenvalues Decomposition is done, and characteristic value is ranked sequentially by monotonic increase, by rear L-
The corresponding feature vector u of M-P characteristic valueM+P+1,uM+P+2,…,uLConstitute matrix G:
G=[uM+P+1 uM+P+2 … uL] (12)
Step 2-3:The MUSIC of X (t) composes formula:
Wherein,For array steering vector, d is array element spacing, and λ is work
Wavelength.Due to not knowing the direction of arrival information of target source, space angle θ is dividedK
The grid number of representation space angular divisions calculates function 1/a successivelyH(θ)GGHThe value of a (θ) obtains the MUSIC spectrums P of X (t)X
(θ)。
Step 3:Believed with interference using the blind source separation algorithm separation echo signal of eigenmatrix Joint diagonalization (JADE)
Number:
Step 3-1:It docks collection of letters X (t) and carries out prewhitening, obtain whitened signal Z (t), i.e.,:
Z (t)=WX (t) (14)
Wherein,For whitening matrix,Umax=[u1 u2
… uM+P], λ1,λ2,…,λM+PFor correlation matrix in step 2-1Preceding M+P characteristic value, u1,u2,…,uM+PIt is corresponded to for it
Feature vector.
Step 3-2:Seek the fourth order cumulant matrix Q of whitened signal Z (t)z:
Wherein E [] indicates operation of averaging, zi(t) the i-th row of whitened signal Z (t) is indicated, i, j, k, l are belonging respectively to 1
~M+P.To QzEigenvalues Decomposition is carried out, M+P maximum eigenvalue λ before obtaining1,λ2,…,λM+PFeature vector corresponding with its
v1,v2,…,vM+P, wherein vi, i=1,2 ..., M+P is (M+P)2× 1 dimension column vector, therefore obtain needing approximate joint diagonal
Objective matrix { the M of change1,M2,…,MM+P}.Wherein Vec (Mi)=λivi, i=1,2 ..., M+P, Vec () expression vectorization calculations
Symbol, i.e., line up column vector by a matrix column vector according to the ordering in matrix.
Step 3-3:A unitary matrice V is found to { M1,M2,…,MM+PJoint diagonalization is carried out, the specific steps are:
Step 3-3-1:Given initial matrix V=IM+P, IM+PIndicate (M+P) × (M+P) dimension unit matrixs and step
M+P objective matrix M in 3-2n, n=1,2 ..., M+P, threshold value ρ.
Step 3-3-2:To matrixEigenvalues Decomposition is carried out, its maximum eigenvalue pair is obtained
The feature vector [x, y, z] answeredT, wherein h (Mn)=[mii-mjj mij+mji i(mji-mij)], mijRepresenting matrix MnI-th row j row
Element,
Step 3-3-3:Utilize [x, y, the z] obtained in 3-3-2TC, s is calculated as follows:
Wherein c, s are the element in Givens spin matrixs G, G(i,j,c,s)(i, i) of representing matrix, (i, j), (j,
I), (j, j) element is respectivelyRemaining element is identical as unit matrix, and matrix G is obtained according to c, s(i,j,c,s)。
Step 3-3-4:Judge whether s >=ρ is true, if set up, carries out step 3-3-5;If invalid, gained
V is required unitary matrice V.
Step 3-3-5:Update matrix V=VG(i,j,c,s)And objective matrixN=1,
2 ..., M+P, until i, j have traversed 1~M+P.Algorithm flow is as shown in Fig. 2.
Step 3-4:Separation signal is obtained with array manifold to estimate:
Wherein, W#For the pseudoinverse of whitening matrix W;Y (t) is separation signal, including echo signal waveform is estimatedWith interference
Signal waveform is estimated Estimate for array manifold, includes the array manifold estimation of echo signalWith interference signal
Array manifold estimation
Step 4:Assuming that the interference signal waveform that step 3 estimates isInterference array manifold beThen weigh
Structure receive signal in interference component be:
Wherein,For the interference component estimated.
Step 5:The MUSIC that target is calculated using CLEAN algorithms on spatial domain is composed:
Step 5-1:The MUSIC spectrums that the interference component that step 3 estimates is calculated by the method for step 2, obtain interfering into
The MUSIC divided composes PJ(θ)。
Step 5-2:The MUSIC spectrums of target in order to obtain, need to minimize cost function:
Assuming that the MUSIC spectrums of the reception signal obtained in step 1 are PX(θ), utilizes PJ(θ) and PX(θ) calculates weight coefficient
Step 5-3:The weight coefficient obtained using step 4-2The MUSIC of calculating target, which is composed, is:
Wherein, PT(θ) is the target MUSIC spectrums after AF panel.The corresponding θ of preceding M maximum value difference is found out, is
The direction of arrival of target.
Simulating, verifying and analysis
Simulation parameter:
Here simulating, verifying is done so that space is there are two targets, two interference as an example.Assuming that echo signal is respectively linear adjusts
Frequency signal and multiple sinusoidal signal, expression formula difference are as follows:
s2(t)=exp (j2 π (f0), t) 0 < t≤τp (22)
Wherein, linear FM signal chirp rate K=B/ τp, B=10MHz is bandwidth of operation, pulse width τp=10 μ s,
Carrier frequency f0=1GHz.
Interference signal considers that frequency spectrum disperse (SMSP) is interfered and amplitude modulated jamming, expression formula difference are as follows:
J2(t)=(U0+Un(t))exp(j(2πfjt+φ)) (24)
Wherein, jsmsp(t)=exp (j2 π f0t+jπk′t2), k '=nK, t ∈ [0, τp/ n], wherein k ' is interference signal
Chirp rate, value are n times of radar emission signal chirp rate, and n is interference subpulse number, and n=5 is arranged here.Un
(t) be mean value it is zero, the white Gaussian noise that variance is 1, fj=1GHz be interference carrier frequency, φ be [0,2 π) it is equally distributed with
Machine variable.
If two echo signals are located at θT1=30 °, θT2=44.5 °, the DOA of two interference signals is respectively θJ1=
30.4°,θJ2=45 °.The signal-to-noise ratio of two echo signals is SNR1=SNR2=10dB, it is dry to make an uproar than being respectively JNR1=60dB and
JNR2=70dB.
Simulation analysis:
From attached drawing 3 as can be seen that in MUSIC spectrums, two interference signal amplitudes are far above target amplitude, lead to target
DOA estimations are difficult.By attached drawing 4 and attached drawing 5 it is found that JADE blind source separation algorithms have good separating property, can effectively by
Echo signal is separated with interference signal.Utilize the obtained interference signal waveforms of JADE and interference steering vector reconstruct interference
Ingredient it is as shown in Fig. 6 to calculate its MUSIC spectrums, it can be seen that the DOA of interference signal can be accurately estimated out.It utilizes
The interference MUSIC spectrums estimated are eliminated to obtain as shown in Fig. 7 by CLEAN algorithms from the MUSIC spectrums for receiving signal
The MUSIC of target is composed, and from attached drawing 7 it can be seen that interference is suppressed, the DOA of target can be come out by good estimation.By tying above
Fruit illustrates effectiveness of the invention.