Disclosure of Invention
Aiming at the defects in the background technology, the invention provides a main lobe anti-interference method combining a JADE (feature matrix joint diagonalization) blind source separation algorithm and a CLEAN algorithm. Firstly, estimating a steering vector and a waveform of an interference signal by using a JADE algorithm, and further reconstructing an interference array signal; then calculating the MUSIC spectrum of the received signal to obtain the MUSIC spectrum of the received signal, and calculating the reconstructed MUSIC spectrum of the interference array signal; and finally, canceling the strong interference signal on the MUSIC spectrum of the received signal through a space domain CLEAN algorithm, thereby obtaining the direction of arrival (DOA) estimation of the target signal. The simulation shows the effectiveness of the method, the algorithm does not need prior information of interference signals, various types of interference can be suppressed, and the method has universal applicability.
The invention provides a JADE and CLEAN combined main lobe anti-interference method, which comprises the following steps:
step 1: it is assumed that M target signals and P high-power interference signals which are mutually independent exist in a space, the difference of the arrival directions of the target signals and the interference signals is within the angle range of a main lobe, and the target signals and the interference signals simultaneously enter a space array, wherein the array consists of L array elements, the target signals and the interference signals are all far-field narrow-band signals, and the t-th array element receives signals:
wherein s is
m(t), M is 1,2, …, M is the mth target signal, θ
TmIs the direction of arrival, DOA angle, J, of the mth target signal
p(t), P is 1,2, …, P is the P-th interference signal, θ
JpIs the DOA angle of the arrival direction of the p-th interference signal, n (T) represents the noise signal at the time T, T represents the total number of samples, d represents the array element interval, lambda is the working wavelength,
then the antenna array receives the signal as:
X(t)=[x1(t),x2(t),…,xL(t)]T,t=1,2,…,T (2)
wherein, (.)TRepresenting a transpose operator, T representing a total number of samples;
step 2: calculating the MUSIC spectrum of the antenna array receiving signal X (t):
step 2-1: computing a spatial correlation matrix of a received signal
Wherein, (.)HRepresenting a conjugate transpose operator;
step 2-2: for correlation matrix
Decomposing the characteristic values, arranging the characteristic values according to a monotone increasing sequence, and arranging the characteristic vectors u corresponding to the last L-M-P characteristic values
M+P+1,u
M+P+2,…,u
LForming a matrix G:
G=[uM+P+1uM+P+2… uL](4)
step 2-3: the MUSIC spectral formula of x (t) is:
wherein the content of the first and second substances,
is an array guide vector, d is an array element interval, and lambda is a working wavelength; since the direction of arrival information of the target source is unknown, the spatial angle theta is divided
K represents the number of grids divided by space angle, and the
function 1/a is calculated in sequence
H(θ)GG
Ha (theta) value to obtain a MUSIC spectrum P of X (t)
X(θ);
And step 3: separating the target signal and the interference signal by adopting a blind source separation algorithm of feature matrix joint diagonalization (JADE):
step 3-1: pre-whitening the received signal x (t) to obtain a whitened signal z (t), that is:
Z(t)=WX(t) (6)
wherein W is a whitening matrix;
step 3-2: obtaining the fourth-order cumulant matrix Q of the whitening signal Z (t)z:
Wherein E [. C]Denotes an averaging operation, zi(t) i, j, k, l in the ith row of the whitening signal Z (t) are 1-M + P; to QzDecomposing the characteristic value to obtainTo the first M + P maximum eigenvalues lambda1,λ2,…,λM+PAnd its corresponding feature vector v1,v2,…,vM+PWherein v isiI is 1,2, …, M + P is (M + P)2X 1-dimensional column vectors, thus resulting in an object matrix { M } requiring approximate joint diagonalization1,M2,…,MM+P}; wherein Vec (M)i)=λiviI 1,2, …, M + P, Vec (·) denotes a vectorization operator, i.e., a column vector of a matrix is arranged into column vectors in the order of arrangement in the matrix;
step 3-3: using unitary matrix V pair { M1,M2,…,MM+PPerforming approximate joint diagonalization;
step 3-4: obtaining a separation signal and an array flow pattern estimation:
wherein, W
#Is the pseudo-inverse of the whitening matrix W; y (t) is a separate signal including an estimate of the waveform of the target signal
And interference signal waveform estimation
Array flow pattern estimation including target signal for array flow pattern estimation
Array flow pattern estimation with interference signal
And 4, step 4: suppose that the interference signal waveform estimated in step 3 is
The interference array flows as
The interference contribution in the reconstructed received signal is:
wherein the content of the first and second substances,
the estimated interference component;
and 5: calculating the MUSIC spectrum of the target by using a CLEAN algorithm in a space domain:
step 5-1: calculating the MUSIC spectrum of the interference component estimated in the step 4 according to the method in the step 2 to obtain the MUSIC spectrum P of the interference componentJ(θ);
Step 5-2: to obtain the MUSIC spectrum of the target, the cost function needs to be minimized:
using P obtained in step 5-1
J(theta) and P obtained in step 2
X(theta) calculating the weight coefficient
Step 5-3: using the weight coefficients obtained in step 4-2
The MUSIC spectrum of the target was calculated as:
wherein, PT(theta) is the target MUSIC spectrum after interference suppression; before finding outThe M maximum values correspond to θ, which is the target direction of arrival.
The invention has the advantages that
The invention provides a main lobe anti-interference method combining JADE and CLEAN, compared with the existing interference suppression algorithm, the method does not need to know the prior information of interference, can be suitable for various types of interference, and realizes DOA estimation on a target.
Firstly, estimating a steering vector and a waveform of an interference signal by using a JADE algorithm, and further reconstructing an interference array signal; then calculating and reconstructing the MUSIC spectrum of the interference array signal to obtain the MUSIC spectrum only containing the interference signal; and finally, canceling the MUSIC spectrum of the interference signal on the MUSIC spectrum of the received signal by utilizing a CLEAN algorithm in a space domain, thereby obtaining the estimation of the direction of arrival of the target signal. Simulation results show that the method can well complete interference suppression and estimate the target DOA.
Detailed Description
Step 1:
it is assumed that M target signals and P high-power interference signals which are mutually independent exist in a space, the difference of the arrival directions of the target signals and the interference signals is within the angle range of a main lobe, and the target signals and the interference signals simultaneously enter a space array, wherein the array consists of L array elements, the target signals and the interference signals are all far-field narrow-band signals, and the t-th array element receives signals:
wherein s ism(t), M is 1,2, …, M is the mth target signal, θTmFor the direction of arrival of the corresponding target signal, Jp(t), P is 1,2, …, P is the pth interference signal, θJpAnd n (T) represents a noise signal at the time T, T represents the total number of samples, d represents the array element interval, and lambda is the working wavelength.
Then the antenna array receives the signal as:
X(t)=[x1(t),x2(t),…,xL(t)]T,t=1,2,…,T (10)
wherein, (.)TDenotes the transpose operator and T denotes the total number of samples.
Step 2: calculating the MUSIC spectrum of the antenna array receiving signal X (t):
step 2-1: computing a spatial correlation matrix of a received signal
Wherein, (.)HRepresenting a conjugate transpose operator.
Step 2-2: for correlation matrix
Decomposing the characteristic values, arranging the characteristic values according to a monotone increasing sequence, and arranging the characteristic vectors u corresponding to the last L-M-P characteristic values
M+P+1,u
M+P+2,…,u
LForming a matrix G:
G=[uM+P+1uM+P+2… uL](12)
step 2-3: the MUSIC spectral formula of x (t) is:
wherein the content of the first and second substances,
and d is the array guide vector, d is the array element interval, and lambda is the working wavelength. Since the direction of arrival information of the target source is unknown, the spatial angle theta is divided
K represents the number of grids divided by space angle, and the
function 1/a is calculated in sequence
H(θ)GG
Ha (theta) value to obtain a MUSIC spectrum P of X (t)
X(θ)。
And step 3: separating the target signal and the interference signal by adopting a blind source separation algorithm of feature matrix joint diagonalization (JADE):
step 3-1: pre-whitening the received signal x (t) to obtain a whitened signal z (t), that is:
Z(t)=WX(t) (14)
wherein the content of the first and second substances,
in order to whiten the matrix, the matrix is,
U
max=[u
1u
2… u
M+P], λ
1,λ
2,…,λ
M+Pfor the correlation matrix in step 2-1
The first M + P eigenvalues, u
1,u
2,…,u
M+PIs its corresponding characteristic vector.
Step 3-2: obtaining the fourth-order cumulant matrix Q of the whitening signal Z (t)z:
Wherein E [. C]Denotes an averaging operation, zi(t) i, j, k, l of the ith row of the whitened signal Z (t), respectively1 to M + P. To QzDecomposing the eigenvalue to obtain the first M + P maximum eigenvalues lambda1,λ2,…,λM+PAnd its corresponding feature vector v1,v2,…,vM+PWherein v isiI is 1,2, …, M + P is (M + P)2X 1-dimensional column vectors, thus resulting in an object matrix { M } requiring approximate joint diagonalization1,M2,…,MM+P}. Wherein Vec (M)i)=λiviI 1,2, …, M + P, Vec (·) denotes a vectorization operator, i.e., a column vector of a matrix is arranged into column vectors in the order of arrangement in the matrix.
Step 3-3: finding a unitary matrix V pair { M }1,M2,…,MM+PPerforming joint diagonalization, which comprises the following specific steps:
step 3-3-1: given an initial matrix V ═ IM+P,IM+PRepresents an (M + P) × (M + P) -dimensional identity matrix, and M + P object matrices M in step 3-2nN is 1,2, …, M + P, threshold ρ.
Step 3-3-2: for matrix
Decomposing the eigenvalue to obtain the eigenvector [ x, y, z ] corresponding to the maximum eigenvalue]
TWherein h (M)
n)=[m
ii-m
jjm
ij+m
jii(m
ji-m
ij)],m
ijRepresentation matrix M
nThe ith row and the jth column of elements,
step 3-3-3: using [ x, y, z ] obtained in 3-3-2]TC, s is calculated as follows:
where c, s are elements in a Givens rotation matrix G, G
(i,j,
c,s)The (i, i), (i, j), (j, i), (j, j) th elements of the representation matrix are respectively
The other elements are the same as the unit array, and a matrix G is obtained according to c and s
(i,j,c,s)。
Step 3-3-4: judging whether s is greater than or equal to rho, and if so, performing the step 3-3-5; if not, the obtained V is the unitary matrix V.
Step 3-3-5: updating matrix V ═ VG
(i,j,c,s)And an object matrix
n is 1,2, …, M + P until i, j has traversed 1-M + P. The algorithm flow is shown in fig. 2.
Step 3-4: obtaining a separation signal and an array flow pattern estimation:
wherein, W
#Is the pseudo-inverse of the whitening matrix W; y (t) is a separate signal including an estimate of the waveform of the target signal
And interference signal waveform estimation
Array flow pattern estimation including target signal for array flow pattern estimation
Array flow pattern estimation with interference signal
And 4, step 4: suppose that the interference signal waveform estimated in step 3 is
The interference array flows as
The interference contribution in the reconstructed received signal is:
wherein the content of the first and second substances,
is the estimated interference component.
And 5: calculating the MUSIC spectrum of the target by using a CLEAN algorithm in a space domain:
step 5-1: calculating the MUSIC spectrum of the interference component estimated in the step 3 according to the method in the step 2 to obtain the MUSIC spectrum P of the interference componentJ(θ)。
Step 5-2: to obtain the MUSIC spectrum of the target, the cost function needs to be minimized:
suppose the MUSIC spectrum of the received signal obtained in
step 1 is P
X(θ) by P
J(theta) and P
X(theta) calculating the weight coefficient
Step 5-3: using the weight coefficients obtained in step 4-2
The MUSIC spectrum of the target was calculated as:
wherein, PTAnd (theta) is the target MUSIC spectrum after interference suppression. Find the first M maximumsThe values are each associated with θ, which is the direction of arrival of the target.
Simulation verification and analysis
Simulation parameters:
here, simulation verification is performed by taking the example that two targets and two interferences exist in the space. Assuming that the target signals are respectively a chirp signal and a complex sine signal, the expressions are respectively as follows:
s2(t)=exp(j2π(f0t)),0<t≤τp(22)
wherein, the chirp rate K of the chirp signal is B/taupWhere B is 10MHz as the working bandwidth and pulse width τ p10 mus, carrier frequency f0=1GHz。
The interference signal considers spectrum dispersion (SMSP) interference and noise amplitude modulation interference, and the expressions are respectively as follows:
J2(t)=(U0+Un(t))exp(j(2πfjt+φ)) (24)
wherein j issmsp(t)=exp(j2πf0t+jπk′t2),k′=nK,t∈[0,τp/n]And k 'is the frequency modulation slope of the interference signal, the value of k' is n times of the frequency modulation slope of the radar transmission signal, n is the number of interference sub-pulses, and n is set to be 5. U shapen(t) is white Gaussian noise with mean value of zero and variance of 1, fj1GHz is the interference carrier frequency and phi is 0,2 pi) a uniformly distributed random variable.
Let two target signals be respectively located at thetaT1=30°,θT244.5 °, DOA of the two interference signals is θJ1=30.4°,θJ245 deg. is equal to. The signal-to-noise ratio of the two target signals is SNR1=SNR210dB, the dry-to-noise ratio is JNR160dB and JNR2=70dB。
Simulation analysis:
as can be seen from fig. 3, on the MUSIC spectrum, the two interfering signal amplitudes are much higher than the target amplitude, resulting in difficulty in target DOA estimation. As can be seen from fig. 4 and 5, the JADE blind source separation algorithm has a good separation performance, and can effectively separate the target signal from the interference signal. The interference component is reconstructed by using the interference signal waveform and the interference guide vector obtained by JADE, and the MUSIC spectrum is calculated and shown in figure 6, so that the DOA of the interference signal can be accurately estimated. The estimated interfering MUSIC spectrum is cancelled from the MUSIC spectrum of the received signal by using the CLEAN algorithm to obtain the MUSIC spectrum of the target shown in fig. 7, and it can be seen from fig. 7 that the interference is suppressed and the DOA of the target can be well estimated. The effectiveness of the present invention is illustrated by the above results.