Summary of the invention
For overcome the deficiencies in the prior art, the present invention provides a kind of removable based on Space-time domain Combined Treatment and singly stands erectly
Localization method is connect, it can sampling number of snapshots few in array element number, array received data be smaller, noise of incoming wave signal compares
In the case where low, it is ensured that the performance of location estimation does not decline.
The technical solution adopted by the present invention to solve the technical problems the following steps are included:
A) assume that antenna array elements number is M, number of source Q, and Q > M, incoming wave signal S (t) are narrow band signal, letter
Road noise N (t) is the additive white Gaussian noise that mean value is zero, and sampling number of snapshots are K, and airspace orientation vector is Al, be delayed tap
Number is G, and the time interval of delay process is τ, sampled data matrix v of the array in observation point l moment t- τ every timel=[v1
(t-τ),…,vM(t-τ)]T=AlS(t-τ)+N(t-τ);
System passes through G delay process, and signal data matrix is received at observation point
In formula
AST-l(p)=[aST-l(P1), aST-l(P2) ..., aST-l(PQ)]∈CM×Q
fiIndicate incoming wave signal frequency, i=1,2 ..., Q, dmIndicate the distance of m-th of array element to reference point, m=1 ...,
M, aST-lIndicate Space-time domain orientation vector;
B) covariance matrix of collection of letters number is followed by by G delay process
C) feature decomposition is carried out to covariance matrix, and calculates noise subspace
Q in formulaSIt is characterized vector matrix, ∑SIndicate diagonal matrix,Indicate signal subspace,Indicate noise
Subspace;
D) initial each particle x is uniformly distributed in observation areai,j(0), allowing respective optimum position is Pi,j(0)=
xi,j(0);Calculate average optimum position C (t);
E) by the corresponding Space-time domain orientation vector a of each point in target areaST-l(x, y, z) is projected to noise subspace respectively
And sum, the cost function of Space-time domain Combined Treatment direct location
The optimal value of search cost function in monitoring region, finds cost function using quantum-behaved particle swarm optimization
Optimal value;
F) according to displacement renewal equationCalculate particle
New position, wherein
I indicates i-th of population;M indicates total number of particles;The dimension of j expression particle;The dimension of N expression search space;t
Indicate evolutionary generation;ui,j(t) andIt is equally distributed random number on [01] section;xi,j(t) position is indicated, most preferably
Position Pi,j(t) it indicates, pi,jIt (t) is attractor position;Gi,j(t) optimum position of group is indicated;C (t) indicates the flat of particle
Equal optimum position;α is expansion-contraction factor;
G) step (e), (f) are repeated, until the difference between the value of global optimum position is no more than 0.001, current grain
Optimal estimation value of the optimum position of son as target position.
The beneficial effects of the present invention are: being measured without intermediate parameters, target position is directly parsed from observation data, is had
Effect improves the precision and resolution capability of location estimation, receives signal by increasing and improves antenna array in the delay tap of time domain
The freedom degree of column substantially increases the performance of location estimation.
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and examples, and the present invention includes but are not limited to following implementations
Example.
The present invention the following steps are included:
A) assume that antenna array elements number is M, number of source Q, and Q > M, incoming wave signal S (t) are narrow band signal, letter
Road noise N (t) is the additive white Gaussian noise that mean value is zero, and sampling number of snapshots are K, and airspace orientation vector is Al.Be delayed tap
Number is G, and the time interval of delay process is τ every time.Sampled data matrix of the array in observation point l moment t- τ can indicate
Are as follows:
vl=[v1(t-τ),…,vM(t-τ)]T=AlS(t-τ)+Nl(t-τ) (1)
System passes through G delay process, and each time interval is τ, then signal data matrix is received at observation point
Are as follows:
In formula
AST-l(p)=[aST-l(p1),aST-l(p2),…,aST-l(pQ)]∈CM×Q (4)
Wherein fi, i=1,2 ..., Q indicates incoming wave signal frequency.dm, m=1 ..., M indicates m-th of array element to reference point
Distance, aST-lIndicate Space-time domain orientation vector.
B) covariance matrix of collection of letters number is followed by by G rank delay process are as follows:
R in formulaSSFor the autocorrelation matrix of incoming signal,For Kronecker product, σ2Indicate noise power, I indicates unit
Matrix,For delay factors at different levels composition column vector, i.e.,
In the concrete realization, data covariance matrix sample covariance matrixInstead of that is,
C) feature decomposition is carried out to covariance matrix, and calculates noise subspace
Q in formulaSIt is characterized vector matrix, ∑SIndicate diagonal matrix,Indicate signal subspace,Indicate noise
Subspace.
D) initial each particle x is uniformly distributed in observation areai,j(0), allowing respective optimum position is Pi,j(0)=
xi,j(0), then average optimum position C (t) is calculated;
E) then by the corresponding Space-time domain orientation vector a of each point in target areaST-l(x, y, z) is respectively to noise subspace
It projects and sums, the cost function of Space-time domain Combined Treatment direct location are as follows:
The optimal value of search cost function, finds generation using quantum-behaved particle swarm optimization (QPSO) in monitoring region
The optimal value of valence function.Its cost function is calculated to each particle, if bigger than the average optimum position C (t-1) of last iteration
Just more new individual optimum position Pi,j(t)=xi,j(t);If than the global optimum position G of last iterationi,j(t-1) greatly, just more
New overall situation optimum position Gi,j(t)=xi,j(t);
F) the new position of particle, the displacement renewal equation of QPSO are calculated according to displacement renewal equation are as follows:
Wherein
I indicates i-th of population;M indicates total number of particles;The dimension of j expression particle;The dimension of N expression search space;t
Indicate evolutionary generation;ui,j(t) andIt is equally distributed random number on [0 1] section;xi,j(t) position is indicated, most preferably
Position Pi,j(t) it indicates, pi,jIt (t) is attractor position;Gi,j(t) optimum position of group is indicated;C (t) indicates the flat of particle
Equal optimum position;α is expansion-contraction factor.
G) step (e), (f) are repeated, by successive ignition until the difference between the value of global optimum position is no more than
Until 0.001, using the optimum position of current particle as the optimal estimation value of target position.
The embodiment of the present invention provides a kind of direct localization method based on Space-time domain Combined Treatment, and Fig. 1 is this method stream
Cheng Tu, Fig. 2 are location model schematic diagrames, and Fig. 3 is the Space-time domain Combined Treatment block diagram of this method, and array antenna is 10 in this example
Array element linear array, antenna aperature are half-wavelength, and observed object number Q is 4, the points of measurement L=20, sample number of snapshots K=512, time domain
Tap G=3.Execute following steps:
Step 1: establishing location model by Fig. 2, and there are 4 observed objects on ground, and true coordinate is respectively (20,200),
(50,150), (50,100), (20,50).Observation station with 20 meter per second of speed along rectilinear flight, in 20 observation point r1,…,r20
Data are acquired, the data for being 524 every the one group of data length of acquisition in 10 seconds, sampling number of snapshots are 512, according to Fig. 3 by three times
Delay process, be delayed four snaps every time, obtains the array received signal matrix being made of delay process signal
Step 2: calculating the covariance of all observation point array signal data after delay process, with sampling covariance
MatrixInstead of that is,
Step 3: to covariance matrixFeature decomposition is carried out, signal noise subspace is found out
Step 4: 25 initial particle x are uniformly distributed in observation areai,j(0), respectively (0,0), (0,50),
(0,100), (0,150), (0,200), (50,0) (50,50), (50,100), (50,150), (50,200), (100,0) (100
50), (100,100), (100,150), (100,200), (150,0) (150,50), (150,100), (150,150), (150,
200), (200,0), (200,50), (200,100), (200,150), (200,200) allow respective optimum position to be Pi,j(0)
=xi,j(0), then average optimum position C (t) is calculated;
Step 5: calculating its cost function to each particle, by the corresponding Space-time domain orientation vector point of each point in target area
Do not project and sum to noise subspace, if the average optimum position than last iteration greatly if more new individual optimum position;Such as
Fruit is bigger than the global optimum position of last iteration, just updates global optimum position;
Step 6: calculating the new position of particle according to displacement renewal equation, wherein the Spatial Dimension N=2 searched for, expansion-
Contraction factor α=0.5;
Step 7: it repeats Step 5: step 6, until the difference between the value of 10 iteration overall situation optimum positions is no more than
Until 0.001, using the optimum position of current particle as the optimal estimation value of target position.
Fig. 4 is positioning target in coordinate (20,200), (50,150), (50,100), when (20,50), is mentioned using the present invention
The direct localization method based on Space-time domain Combined Treatment out carries out the power spectrum chart of location estimation, it can be seen from the figure that this
The spectral peak based on the direct localization method of Space-time domain Combined Treatment localization method more direct than tradition of invention is more sharp, and error is more
Small, resolution ratio is more preferable.
Fig. 5 indicates that the relationship of algorithm root-mean-square error and number of snapshots is directly positioned based on Space-time domain in combination as seen from the figure
Method is more preferable than the estimation performance of pure airspace direct location, and in fewer snapshots, the method for space-time joint processing is being estimated
There is biggish promotion in performance.
Fig. 6 indicates the variation of the algorithm root-mean-square error under different signal-to-noise ratio, as seen from the figure, straight in combination using Space-time domain
It connects positioning mode and realizes that the snr threshold of successfully positioning is 5dB lower than using subspace to merge direct localization method.5dB's changes
It is kind to mean that observable distance double, keep long distance positioning more accurate, under low signal-to-noise ratio, the method for the present invention performance
More preferably.