CN104392115B - A kind of high-resolution two-dimensional parameter evaluation method - Google Patents

Abstract

The invention discloses a kind of high-resolution two-dimensional parameter evaluation method, belong to Radar Technology field.The data that the present invention is received by each sensor passage of sample record antenna, sampled data is lined up to the form of column vector, in time-domain, spatial domain and delay domain formation correlation matrix, denoising is carried out to correlation matrix, signal subspace is calculated using cycle specificity decomposition method, noise subspace is solved from the relation of signal and noise subspace, obtain the projection matrix of noise subspace, target bearing and pitch information are solved based on projection matrix is counter, classical MUSIC methods are avoided the problem of signal subspace is distinguished and drastically declining occurs in performance during noise subspace, improve the azimuth for determining information source, the accuracy during angle of pitch.

Description

A kind of high-resolution two-dimensional parameter evaluation method

Technical field

The present invention relates to Radar Technology field, more particularly to a kind of high-resolution two-dimensional parameter evaluation method.

Background technology

Spatialization function is one of key technology of antenna applications.In order to pursue higher positioning precision, according to orientation Estimation and the similitude of the Frequency Estimation of time signal, many Time-Domain Nonlinear localization methods are generalized in dimensional orientation estimation just Generate so-called High Resolution DOA.

The Schmidt in the U.S. proposes famous multiple signal classification (MUSIC) method, realizes modern high-resolution orientation The leap of estimation technique, so that the rise of proper subspace class method is promoted, the singular value point that this method passes through data matrix The Eigenvalues Decomposition (EVD) of (SVD) or space covariance matrix is solved to obtain mutually orthogonal signal subspace and noise sky Between, the algorithm for estimating of signal parameter is then constructed using the relation of aerial array steering vector and subspace, obtains to orientation and estimates The Asymptotic upbiased estimation of meter, breaches array aperture in conventional orientation algorithm for estimating and the Rayleigh of parameter Estimation performance is limited, mesh Mark resolution capability can reach the 1/3~1/5 of beam angle.

During the present invention is realized, inventor has found that prior art at least has problems with：

In actual applications, when signal to noise ratio is less than zero and smaller hits, classical MUSIC methods are distinguishing signal subspace sky Between and the problem of drastically declining occurs in performance during noise subspace, cause it is determined that information source azimuth, the angle of pitch when accuracy Decline.

The content of the invention

In order to solve problem of the prior art, the invention provides a kind of high-resolution two-dimensional parameter evaluation method, institute State including：

Step one, signal is sampled by sensor array, gets the first signal phasor x₁(t), secondary signal Vector x₂(t), wherein x₁(t)=A₁s(t)+n₁(t), x₂(t)=A₁Φs(t)+n₂(t)；

Step 2, first moment after sampling is started, construction is sweared with first signal phasor, the secondary signal The corresponding matrix of amount

Wherein R_sFor autocorrelation matrix,For the noise in the sampling process, second after sampling is started Moment, construction matrix corresponding with first signal phasor, the secondary signal vector

Two correlation matrixes are constructed at each moment successively, and according to time sequencing, whole correlation matrixes are combined as Correlation matrix groupWherein k represents moment number, Λ_m=Φ^m-1,

Whole correlation matrixes are carried out feature decomposition, the noise in whole correlation matrixes are carried out by step 3 Estimation, obtains the estimate of each noise It is right Whole eigenmatrixes carry out denoising, obtain denoising correlation matrix group

Step 4, construction initial matrix U (0), sets up the first cost function

Wherein, l represents cycle-index, carries out Eigenvalues Decomposition to first cost function, obtains the first breakdown

Wherein,It is characterized value matrix,Vector matrix is characterized, order matrix V (l-1) is matrixThe corresponding characteristic vector formation of the big characteristic value of preceding P matrix, set up the second cost function

Wherein, l represents cycle-index, carries out Eigenvalues Decomposition to second cost function, obtains the second breakdown

Wherein,It is characterized value matrix,It is characterized vector matrix, first cost function, the second cost letter Number arrives U (l) one cycle for U (l-1), to U (l) cycle calculations, until meeting

||U(l)U^H(l)-U(l-1)U^H(l-1)||_F＜ ε

When stop circulation, obtain stop circulation when l；

Step 5, l when being circulated according to the stopping, determining signal subspace U (l), makesProjected MatrixAngle estimation is obtained according to the projection matrix

The beneficial effect that the technical scheme that the present invention is provided is brought is：

The data received by each sensor passage of sample record antenna, sampled data are lined up the form of column vector, Time-domain, spatial domain and delay domain formation correlation matrix, carry out denoising to correlation matrix, utilize cycle specificity decomposition method Signal subspace is calculated, noise subspace is solved from the relation of signal and noise subspace, obtains the projection square of noise subspace Battle array, solves target bearing and pitch information based on projection matrix is counter, it is to avoid classical MUSIC methods are being distinguished signal subspace and made an uproar The problem of drastically declining occurs in performance during phonon space, accuracy when improving the azimuth for determining information source, the angle of pitch.

Brief description of the drawings

In order to illustrate more clearly of technical scheme, embodiment will be described below needed for the accompanying drawing to be used It is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the present invention, general for this area For logical technical staff, on the premise of not paying creative work, other accompanying drawings can also be obtained according to these accompanying drawings.

Fig. 1 is a kind of structural representation for fixing device that the present invention is provided；

Fig. 2 is that floor map is detected in the azimuth that the present invention is provided as a result for the special sections at 45 ° of places；

Fig. 3 (a) is that the space spectral peak contrast of MUSIC methods and the inventive method under the 0dB signal to noise ratios that provide of the present invention is shown It is intended to；

Fig. 3 (b) is that the space spectral peak contrast of MUSIC methods and the inventive method under the 5dB signal to noise ratios that provide of the present invention is shown It is intended to；

Fig. 3 (c) is that the space spectral peak contrast of MUSIC methods and the inventive method under the 10dB signal to noise ratios that provide of the present invention is shown It is intended to；

Fig. 4 (a) is the space spectral peak contrast signal that the present invention provides the lower MUSIC methods of 300 samplings and the inventive method Figure；

Fig. 4 (b) is that the space spectral peak contrast of the lower MUSIC methods of 500 samplings and the inventive method that the present invention is provided is shown It is intended to；

Fig. 4 (c) is that the space spectral peak contrast of the lower MUSIC methods of 700 samplings and the inventive method that the present invention is provided is shown It is intended to.

Embodiment

To make the structure and advantage of the present invention clearer, the structure of the present invention is made further below in conjunction with accompanying drawing Description.

Embodiment one

The invention provides a kind of high-resolution two-dimensional parameter evaluation method, as shown in figure 1, the device includes：

Step one, signal is sampled by sensor array, gets the first signal phasor x₁(t), secondary signal Vector x₂(t), wherein x₁(t)=A₁s(t)+n₁(t), x₂(t)=A₁Φs(t)+n₂(t)。

In force, the model that antenna receives signal is introduced first.Antenna is passed by the rectangle of M row N row proportional spacings Sensor array element is constituted, and adjacent array element spacing d is less than half-wavelength, coordinate of m-th sensor (m=1,2 ..., MN) in XOY faces For (x_m,y_m), the antenna, which is believed that, by two there is mutually isostructural submatrix to constitute, and detailed construction is as shown in figure 1, two submatrixs M rows are, the rectangular array of N-1 row, the 1st submatrix is arranged to M rows, 1 to N-1 for the 1 of former array and constituted, and the 2nd submatrix is original The 1 of array arranges composition to M rows, 2 to N.If the space two-dimensional angle of p-th of information source is (α_p,β_p).α represents azimuth, and β is represented and bowed The elevation angle, total reception signal of all P information sources in m-th of array element is under t (t=1,2 ..., T) secondary sampling

Whereinλ is the wavelength of information source, Δ_mp=x_mcosα_psinβ_p+y_msinα_psinβ_p, s_pP-th Information source complex envelope, n_mFor the noise shaken in member for m-th.The reception data vector of lower 1st submatrix of the t times sampling is represented by

x₁(t)=A₁s(t)+n₁(t)

The reception data vector of the t times sampling submatrix 2 is represented by

x₂(t)=A₁Φs(t)+n₂(t)

Wherein Φ is diagonal matrix,

Φ=diag (μ₁,μ₂,…,μ_P),

Step 2, start sampling after first moment, construct with the first signal phasor, secondary signal vector it is corresponding Matrix

Wherein R_sFor autocorrelation matrix,For the noise in the sampling process, second after sampling is started Moment, construction matrix corresponding with the first signal phasor, secondary signal vector

Two correlation matrixes are constructed at each moment successively, and according to time sequencing, whole correlation matrixes are combined as Correlation matrix groupWherein k represents moment number, Λ_m=Φ^m-1,

In force, because in each sampling instant, the first submatrix and the second submatrix obtain sampled signal values respectively, therefore Two signal phasor matrixes can be generated in same sampling instant, after sampling process terminates, by multiple signal phasors of generation Matrix is reassembled into correlation matrix group

Whole correlation matrixes are carried out feature decomposition, the noise in whole correlation matrixes are estimated, obtained by step 3 The estimate of each noiseTo whole features Matrix carries out denoising, obtains denoising correlation matrix group

In force, to correlation matrix groupIn each matrix carry out feature decomposition, matrix after disassembly is chosen M-P small characteristic values, estimate noise according to the average value of features described above value, obtain the estimate of each noiseFrom related matrix groupIn each matrix in remove After corresponding noise, obtained part denoising matrix is as follows：

Above-mentioned denoising matrix is combined as denoising matrix group

Step 4, construction initial matrix U (0), sets up the first cost function

Wherein, l represents cycle-index, carries out Eigenvalues Decomposition to the first cost function, obtains the first breakdown

Wherein,It is characterized value matrix,Vector matrix is characterized, order matrix V (l-1) is matrixThe corresponding characteristic vector formation of the big characteristic value of preceding P matrix, set up the second cost function

Wherein, l represents cycle-index, carries out Eigenvalues Decomposition to the second cost function, obtains the second breakdown

Wherein,It is characterized value matrix,Vector matrix is characterized, the first cost function, the second cost function are U (l-1) arrives U (l) one cycle, to U (l) cycle calculations, until meeting

||U(l)U^H(l)-U(l-1)U^H(l-1)||_F＜ ε, (5)

When stop circulation, obtain stop circulation when l.

In force, two cost functions are constructed to initial matrix U (0)WithIt is rightObtained after Eigenvalues Decomposition To withRelated multinomial, it is rightCarry out feature decomposition after obtain withRelated multinomial, now easily sends out Existing formula (1) to formula (2), be from U (l-1) multinomials toPolynomial process, formula (3) to (4) is from V (l-1) multinomial is arrivedPolynomial process, therefore, formula (1) to formula (2), formula (3) to formula (4) are a U (l-1) U (l) circulation is arrived, one cycle is often completed, the numerical value in () adds 1, until when formula (5) is set up, according to now l's Numerical value, determines matrix U (L).

Step 5, according to l when stopping circulating, determines signal subspace U (l), makesObtain projection matrixAngle estimation is obtained according to projection matrix

In force, should in order to make it easy to understand, sensor array used in the present invention is reduced into a line segment here One end points of line segment is fixed on the origin position of 3 d space coordinate system, in XOY plane, the range of movement of line segment projection For 0~180 °, in XOZ planes, the range of movement of line segment projection is 0~90 °, and the minimum move angle for defining the line segment is 1 °, then in measurement, for each azimuth angle alpha_p, there is 90 angle of pitch β_pCorrespond to therewith, due to azimuth angle alpha_pCan have 180 Plant value possible, thus there may be 16200 kinds in the presence of the dimensional orientation of combination in 180 × 90=16200, i.e. information source, Need to carry out signal sampling in 16200 orientation altogether, by calculating peak value highest sampled signal correspondence in sampled result Angle value, be used as the dimensional orientation of information source.In formula (6)It is the first signal arrow generated in step 2 Include azimuth in vector in amount, secondary signal vector in matrix in A1, the first signal phasor, secondary signal vector and bow Two, elevation angle angle, vectorIn also include two angles in azimuth and the angle of pitch, whereinIt is to being to p-th The azimuthal estimation of information source,It is the estimation to p-th of information source angle of pitch, andRepresentation vector's Conjugate transposition vector.

To further illustrate the superiority of the more classical high-resolution positioning side of the method for the present invention (such as MUSIC) method, do as Lower emulation experiment, takes three spacing waves in experiment, if three signals are located at (20 °, 20 °) respectively, (50 °, 50 °), (55 °, 55 °), each experiment takes 100 independent experiment average results.As shown in Fig. 2 for the simplicity of drawing, azimuth is only used here Special sections for 45 ° of places detect plane as a result, are not offered as this method and are only applicable to the orientation of information source in plane estimating Meter.

Experiment 1：Orientation and luffing angle estimation performance change with signal to noise ratio.

This experiment fixed sample number is 300 times, and Fig. 3 (a) is that signal to noise ratio is MUSIC methods and the inventive method under 0dB Space spectral peak, Fig. 3 (b) is the space spectral peak of MUSIC methods and the inventive method under 5dB, and Fig. 3 (c) is MUSIC methods under 10dB With the space spectral peak of the inventive method.From Fig. 3 (a), under 0dB classical MUSIC methods by two source signals (50 °, 50 °), (55 °, 55 °) are mistakenly considered a signal, and the angle estimated is not at any one actual value, and the inventive method can The signal nearer correctly to estimate the two spacing.From Fig. 3 (b), under 5dB classical MUSIC methods have one compared with High spectral peak occurs approximately in (50 °, 50 °) place, but lost target (55 °, 55 °).Fig. 3 (c) signal to noise ratio is 10dB, this hair Bright method and MUSIC methods can correctly estimate two echo signals, but the spectral peak of the inventive method is higher more sharp, with more Excellent resolution performance.

Experiment 2：Orientation and luffing angle estimation performance change with hits.

The fixed signal to noise ratio of this experiment is 5dB, and Fig. 4 (a) is that signal to noise ratio is the lower MUSIC methods of 300 samplings and side of the present invention The space spectral peak of method, Fig. 4 (b) is the space spectral peak of 500 lower MUSIC methods of sampling and the inventive method, and Fig. 4 (c) is 700 The space spectral peak of MUSIC methods and the inventive method under secondary sampling.From Fig. 4 (a), classical MUSIC under being sampled at 300 times Method is by two source signals (50 °, 50 °), and (55 °, 55 °) are mistakenly considered a signal, and the angle of estimation is not true at any one At real value, and the inventive method can correctly estimate the nearer signal of the two spacing.From Fig. 4 (b) and 4 (c), 700 and 900 lower the inventive method of sampling and MUSIC methods can correctly estimate two echo signals, but the inventive method Spectral peak is higher more sharp, with more excellent resolution performance.

A kind of high-resolution two-dimensional parameter evaluation method proposed in the present embodiment, is respectively sensed by sample record antenna The data of device channel reception, sampled data are lined up the form of column vector, in time-domain, spatial domain and delay domain formation Correlation Moment Battle array, denoising is carried out to correlation matrix, and signal subspace is calculated using cycle specificity decomposition method, empty from signal and noise Between relation solve noise subspace, obtain the projection matrix of noise subspace, target bearing and bowed based on projection matrix counter solve Face upward information, it is to avoid what drastically declining occurred in the classical MUSIC methods performance when distinguishing signal subspace and noise subspace asks Topic, accuracy when improving the azimuth for determining information source, the angle of pitch.

Embodiments of the invention are the foregoing is only, are not intended to limit the invention, it is all in the spirit and principles in the present invention Within, any modifications, equivalent substitutions and improvements made etc. should be included within the scope of the present invention.

Claims (1)

1. a kind of high-resolution two-dimensional parameter evaluation method, it is characterised in that the evaluation method includes：

Step one, signal is sampled by sensor array, gets the first signal phasor x₁(t), secondary signal vector x₂ (t), wherein x₁(t)=A₁s(t)+n₁(t), x₂(t)=A₁Φs(t)+n₂(t)；

Step 2, first moment after sampling is started, construction and first signal phasor, the secondary signal vector Corresponding matrix

$R_{11}^{\left(1\right)}=A_1{R}_s\left(1\right)A_1^H+{\mathrm{\σ}}_1^2\left(1\right)I,$

$R_{22}^{\left(1\right)}=A_1{\mathrm{\ΦR}}_s\left(1\right){\mathrm{\Φ}}^H{A}_1^H+{\mathrm{\σ}}_2^2\left(1\right)I,$

Wherein R_sFor autocorrelation matrix,For the noise in sampling process, second moment after sampling is started, structure Make matrix corresponding with first signal phasor, the secondary signal vector

$R_{11}^{\left(2\right)}=A_1{R}_s\left(2\right)A_1^H+{\mathrm{\σ}}_1^2\left(2\right)I,$

$R_{22}^{\left(2\right)}=A_1{\mathrm{\ΦR}}_s\left(2\right){\mathrm{\Φ}}^H{A}_1^H+{\mathrm{\σ}}_2^2\left(2\right)I,$

Two correlation matrixes are constructed at each moment successively, and according to time sequencing, whole correlation matrixes are combined as correlation Matrix groupWherein k represents moment number, Λ_m=Φ^m-1,

Whole correlation matrixes are carried out feature decomposition, the noise in whole correlation matrixes are estimated by step 3, Obtain the estimate of each noise To whole phases Close matrix and carry out denoising, obtain denoising correlation matrix group

Step 4, construction initial matrix U (0), sets up the first cost function

$\tilde{C}=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{m=1}{\overset{2}{\mathrm{\Σ}}}\left\{{\[R_{mm}^{\left(k\right)}\]}^{H}U(l-1)U^H(l-1)R_{mm}^{\left(k\right)}\right\},$

Wherein,Cycle-index is represented, Eigenvalues Decomposition is carried out to first cost function, the first breakdown is obtained

${\tilde{C}}_{eig}=\stackrel{\‾}{V}(l-1)\stackrel{\‾}{D}(l-1){\stackrel{\‾}{V}}^H(l-1),$

Wherein,It is characterized value matrix,It is characterized vector matrix, order matrixFor matrix The corresponding characteristic vector formation of the big characteristic value of preceding P matrix, set up the second cost function

Wherein,Cycle-index is represented, Eigenvalues Decomposition is carried out to second cost function, the second breakdown is obtained

Wherein,It is characterized value matrix,Vector matrix is characterized, first cost function, the second cost function areArriveOne cycle, to describedCycle calculations, until meeting

$\left|\right|U\left(l\right)U^H\left(l\right)-U(l-1)U^H(l-1)|{|}_F<\mathrm{\&epsiv;}$

When stop circulation, obtain stop circulation when

Step 5, when being circulated according to the stoppingDetermine signal subspaceOrderObtain projection matrixAngle estimation is obtained according to the projection matrix

$P({\widehat{\mathrm{\&alpha;}}}_p,{\widehat{\mathrm{\&beta;}}}_p)=\frac{1}{|a^H({\widehat{\mathrm{\&alpha;}}}_p,{\widehat{\mathrm{\&beta;}}}_p)\&lsqb;I-{\tilde{U}}_s{\tilde{U}}_s^H\&rsqb;a({\widehat{\mathrm{\&alpha;}}}_p,{\widehat{\mathrm{\&beta;}}}_p)|}.$