CN103605108A - High-precision remote direction estimation method of acoustic vector array - Google Patents

Abstract

The invention provides a high-precision remote direction estimation method of an acoustic vector array. (1) Narrow-band filtering is performed on a received signal of a two-dimensional vector hydrophone array, and a narrow-band output signal on a frequency point to be processed is obtained; (2) two vibration velocity components of the two-dimensional vector hydrophone array that are orthogonal to each other are linearly combined in a complex domain, and are converted into two new vibration velocity output components; (3) maximum likelihood estimation is used to determine sound pressure snapshoted by a certain number of samples and a cross covariance matrix pair of the double vibration velocity components in the complex domain; and (4) applying an ESPRIT algorithm of a matrix pencil to calculate a rotation invariant factor between the sound pressure vibration velocity cross covariance matrix pair based on the generalized acoustic energy flow, thereby performing direction estimation. The high-precision remote direction estimation method of the acoustic vector array can perform high-precision direction estimation under the circumstance of any array type and even an unknown array type, and the phenomenon that signals in certain directions are severely weakened and even completely shielded does not occur. In addition, the high-precision remote direction estimation method has a relatively low processable signal to noise ratio threshold.

Description

Acoustic vector sensor array High Precision Long-distance direction estimation method

Technical field

What the present invention relates to is a kind of Underwater Detection and hydrolocation method, is specifically related to a kind of direction estimation method of acoustic vector sensor array.

Background technology

Sound field is with scalar field and vector field, and wherein acoustic pressure provides scalar information, and vibration velocity provides Vector Message, and this direction vector is consistent with sound direction of wave travel.Vector hydrophone is combined by pressure hydrophone and velocity hydrophone, can pick up acoustic pressure and vibration velocity information in sound field simultaneously, compared to traditional pressure hydrophone array, vector hydrophone arrays has obtained more information, thereby larger processing space can be provided.

Chinese scholars has been carried out a large amount of research to the orientation estimation application based on acoustic vector sensor array.Document [1], Malcolm Hawkes and Arye Nehorai (Acoustic Vector-sensor Corrections in Ambient Noise.IEEE J.Oceanic Eng, 2001,26 (3): 337-347.) correlation properties of the acoustic pressure in isotropic noise field and vibration velocity are inquired into, and provided the computing formula of acoustic pressure and vibration velocity related coefficient.Document [2] Bai Xingyu etc. (the long-range ESPRIT of a kind of new acoustic vector sensor array orientation algorithm for estimating. Harbin Engineering University's journal, 2006,27 (6): 891～895.), document [3] (acoustic vector sensor array information source number based on acoustic pressure vibration velocity Combined Treatment detects with orientation and estimates. acoustic journal, 2008,33 (1): 56～61.) propose the acoustic pressure vibration velocity Cross-covariance of acoustic vector sensor array to be applied in orientation estimation, the anti-isotropic noise ability of acoustic vector sensor array and high-resolution orientation algorithm for estimating are combined, obtained good effect.But the vibration velocity component of the method has spatial filtering effect, may cause interested signal conductively-closed to be fallen.Document [4] moon etc. (the non-space ESPRIT algorithm of vector array. Harbin Engineering University's journal, 2009,30 (4): 406～410.) utilize the acoustic pressure of acoustic vector sensor array and the rotational invariance between vibration velocity sensor array, can the in the situation that of unknown array manifold, to target, carry out orientation estimation, but the method is not utilized the uncorrelated characteristic of acoustic pressure and vibration velocity channel noise, estimated accuracy need to improve.

Summary of the invention

The object of the invention is to propose a kind ofly can be used in the even unknown formation acoustic vector of any formation array, and combine the anti-isotropic noise ability of acoustic vector sensor array, the acoustic vector sensor array High Precision Long-distance direction estimation method that precision is high.

The object of the present invention is achieved like this:

(1) the reception signal of two-dimensional vector hydrophone array is carried out to narrow-band filtering, obtain the arrowband output signal on pending frequency;

(2) two mutually orthogonal vibration velocity components of two-dimensional vector hydrophone array are carried out to linear combination in complex field, be converted to two new vibration velocity output components, construct the two vibration velocity components of complex field;

(3) adopt maximal possibility estimation to obtain acoustic pressure that some samplings take soon and the Cross-covariance pair of the two vibration velocity components of complex field, construct the acoustic pressure vibration velocity Cross-covariance pair based on broad sense acoustic energy flow;

(4) the ESPRIT algorithm of application matrix bundle calculate acoustic pressure vibration velocity Cross-covariance based on broad sense acoustic energy flow between the invariable rotary factor, thereby carry out orientation estimation.

The present invention can also comprise:

1. the two vibration velocity components of structure complex field described in specifically comprise: by the vibration velocity output component v of y direction _yj multiplies each other with complex unit, then with the vibration velocity output component v of x direction _xbe added and obtain new vibration velocity output component v ₁; By the vibration velocity output component v of y direction _ymultiply each other with negative complex unit-j, then with the vibration velocity output component v of x direction _xbe added and obtain new vibration velocity output component v ₂.V ₁and v ₂formed the two vibration velocity components of complex field, they have still carried the directional information of vibration velocity field, and have the receiving ability of omnidirectional, are a kind of improvement to conventional vibration velocity component.

2. described in, construct acoustic pressure vibration velocity Cross-covariance based on broad sense acoustic energy flow to specifically comprising: the output of the sound pressure sensor array of the snap quantity of necessarily sampling is multiplied each other with the conjugate transpose of two new complex field vibration velocity output components respectively, and ask its mean value, obtain Cross-covariance based on broad sense acoustic energy flow to R ₁and R ₂.

3. the ESPRIT algorithm based on broad sense acoustic energy flow described in, its thought is: solve the right non-zero generalized eigenvalue of acoustic pressure vibration velocity covariance matrix of above-mentioned gained, thereby provide the orientation estimated result of target.For fear of occurring potential numerical value difficulty in the process solving non-zero generalized eigenvalue, the ESPRIT algorithm of described application matrix bundle calculate acoustic pressure vibration velocity Cross-covariance based on broad sense acoustic energy flow between the invariable rotary factor in adopt svd (SVD) and total least square method (TLS) Ill-conditioned Least-squares Problems of a larger dimension to be converted into the non-ill total least square problem of a less dimension.

Advantage body of the present invention is mainly reflected in: the ESPRIT algorithm of (1) conventional acoustic vector sensor array adopts the autocorrelation matrix of data to carry out feature decomposition, even in the ideal case, still has noise in its signal subspace.For acoustic vector sensor array, when array element distance meets the noise correlation radius of acoustic pressure and vibration velocity, noise between acoustic pressure and vibration velocity component is incoherent, the present invention adopts the Cross-covariance of acoustic pressure and vibration velocity, take full advantage of the uncorrelated characteristic of acoustic pressure and vibration velocity noise field, there is good anti-isotropic noise ability; (2) conventional vibration velocity output component and the combination vibration velocity output based on Givess rotation all have space directivity, the incoming signal of different directions is had to different amplitude weightings, especially from the signal of the orthogonal directions incident of observed ray, can be masked completely.The present invention improves vibration velocity output component, and the incoming signal of all directions is all had to identical amplitude response, can completely receive from the signal of all directions incident; (3) between the vibration velocity output component that the present invention constructs, there is rotational invariance, its corresponding twiddle factor is only relevant with the incident direction of signal, and irrelevant with the locus of array, application ESPRIT algorithm can be realized in formation arbitrarily and even carry out the estimation of high-precision orientation unknown formation in the situation that.

Accompanying drawing explanation

Fig. 1 is the spatial correlation function curve of acoustic pressure vibration velocity in isotropic noise field;

Fig. 2 is the one group of irregular formation acoustic vector sensor array position view providing;

Fig. 3 is the block diagram of realizing of the present invention;

Fig. 4 (a)-Fig. 4 (e) is single goal signal two vibration velocity channel output spectrums of acoustic pressure, vibration velocity and complex field during from 30 ° of direction incident, wherein: Fig. 4 (a) sound pressure channel frequency spectrum, Fig. 4 (b) vibration velocity channel v _xfrequency spectrum, Fig. 4 (c) vibration velocity channel v _yfrequency spectrum, Fig. 4 (d) vibration velocity channel v ₁frequency spectrum, Fig. 4 (e) vibration velocity channel v ₂frequency spectrum;

Fig. 5 (a)-Fig. 5 (e) is that Bi-objective signal is respectively from 30 °, two vibration velocity channel output spectrums of acoustic pressure, vibration velocity and complex field during 40 ° of direction incident, wherein: Fig. 5 (a) sound pressure channel frequency spectrum, Fig. 5 (b) vibration velocity channel v _xfrequency spectrum, Fig. 5 (c) vibration velocity channel v _yfrequency spectrum, Fig. 5 (d) vibration velocity channel v ₁frequency spectrum, Fig. 5 (e) vibration velocity channel v ₂frequency spectrum;

Fig. 6 is that root-mean-square error curve is estimated in single goal orientation;

Fig. 7 is that root-mean-square error curve is estimated in Bi-objective orientation;

Fig. 8 (a)-Fig. 8 (e) is signal two vibration velocity channel output spectrums of acoustic pressure, vibration velocity and complex field during from the incident of y direction of principal axis, wherein: Fig. 8 (a) sound pressure channel frequency spectrum, Fig. 8 (b) vibration velocity channel v _xfrequency spectrum, Fig. 8 (c) vibration velocity channel v _yfrequency spectrum, Fig. 8 (d) vibration velocity channel v ₁frequency spectrum, Fig. 8 (e) vibration velocity channel v ₂frequency spectrum;

Fig. 9 (a)-Fig. 9 (b) is the orientation estimated result near the signal of direction incident y axle, Fig. 9 (a) is the orientation estimated result that adopts conventional vibration velocity output component, and Fig. 9 (b) is the orientation estimated result that adopts the two vibration velocity components of complex field.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is further described in detail:

1. calculation procedure

1.1. array output model

The polar coordinates of supposing i element position of the two-dimentional acoustic vector sensor array of N unit are [r _i, α _i], i=1,2 ..., N, far field, the arrowband plane wave that D centre frequency is ω incides on basic matrix, and wherein the incident direction of k signal is θ _k, the output signal of array can be designated as:

$\left\{\begin{array}{c}p\left(t\right)=A_{p}s\left(t\right)+n_p\left(t\right)\\ v_x\left(t\right)=A_{x}s\left(t\right)+n_x\left(t\right)\\ v_y\left(t\right)=A_{y}s\left(t\right)+n_y\left(t\right)\end{array}\right.---\left(1\right)$

P in formula (t), v _x(t), v _y(t) be respectively the N * 1 dimension reception snap vector of sound pressure channel and x, y vibration velocity channel; S (t) is that D far-field signal source is at D * 1 of reference position dimension acoustic pressure vector; n _p(t), n _x(t), n _y(t) be the local noise vector of N * 1 dimension that each passage is corresponding.

A _p=[a _p1..., a _pD] be the sound pressure channel stream shape matrix that signal vector is corresponding, the stream shape vector of k signal can be written as:

a _pk=[exp(jωr _Ncos(θ _k-α ₁)/c),…,exp(jωr _Ncos(θ _k-α _N)/c)] ^T (2)

Operational symbol () ^trepresenting matrix transposition, A _x=[a _x1..., a _xD] be the x direction vibration velocity channel stream shape matrix that signal vector is corresponding, the stream shape vector of k signal can be written as:

a _xk=[exp(jωr _Ncos(θ _k-α ₁)/c),…,exp(jωr _Ncos(θ _k-α _N)/c)] ^Tcosθ _k (3)

A _y=[a _y1..., a _yD] be the y direction vibration velocity channel stream shape matrix that signal vector is corresponding, the stream shape vector of k signal can be written as:

a _yk=[exp(jωr _Ncos(θ _k-α ₁)/c),…,exp(jωr _Ncos(θ _k-α _N)/c)] ^Tsinθ _k (4)

From formula (2)-(4), A _p, A _x, A _ybetween there is following relation:

A _x=A _pΦ _x,A _y=A _pΦ _y,A _y=A _xΦ _xy (5)

In formula, Φ _x, Φ _y, Φ _xyare all diagonal matrix, and have:

$\left\{\begin{array}{c}{\mathrm{\Φ}}_x=\mathrm{diag}[{\cos \mathrm{\θ}}_1,\·\·\·,{\cos \mathrm{\θ}}_D]\\ {\mathrm{\Φ}}_y=\mathrm{diag}[{\sin \mathrm{\θ}}_1,\·\·\·,{\sin \mathrm{\θ}}_D]\\ {\mathrm{\Φ}}_{\mathrm{xy}}=\mathrm{diag}[{\tan \mathrm{\θ}}_1,\·\·\·,{\tan \mathrm{\θ}}_D]\end{array}\right.---\left(6\right)$

From formula (5), can find out, for far field plane wave incoming signal, between the sound pressure sensor array of acoustic vector sensor array and each vibration velocity component sensor array, all there is invariable rotary relation, and this twiddle factor is only relevant with signal incident direction, and irrelevant with the locus of array, this is that acoustic vector sensor array can even be applied the physical basis that ESPRIT algorithm carries out orientation estimation unknown formation in the situation that in formation arbitrarily.

1.2. two vibration velocity output components of new complex field

In isotropic noise field, the spatial correlation function curve of acoustic pressure vibration velocity as shown in Figure 1.1. curve has shown the autocorrelation function curve of acoustic pressure, at d=0 (d is space length), locate correlativity the strongest, and be 0 at the auto-correlation function value that d=0.5 λ (λ is the wavelength of signal) locates acoustic pressure, therefore the algorithm based on acoustic pressure battle array adopts half-wavelength spacing conventionally.2. curve has shown the spatial cross correlation function of acoustic pressure and vibration velocity, and from figure, the acoustic pressure of space concurrent and vibration velocity are incoherent, and at d=0.715 λ place, its correlation function value is also 0, and the noise correlation radius of known acoustic pressure and vibration velocity is about 0.715 λ.Different from traditional acoustic pressure battle array, when if consider, the array element distance of acoustic vector sensor array meets the noise correlation radius of acoustic pressure and vibration velocity, now have:

$E\left(n_p{n}_x^H\right)=E\left(n_p{n}_y^H\right)=0---\left(7\right)$

In formula, operational symbol () ^hthe conjugate transpose of representing matrix.From formula (7), adopt the Cross-covariance of acoustic pressure and vibration velocity, will there is the ability of good anti-isotropic noise.It should be noted that the dipole directive property due to vibration velocity sensor self, conventional acoustic pressure vibration velocity Cross-covariance has the effect of spatial filtering, likely can cause interested parties to signal weakenedly even masked completely.The Cross-covariance of vibration velocity component of acoustic pressure and x direction of take is example, from the signal of y direction of principal axis incident, will be masked completely.In order to address this problem, the present invention proposes, in complex field, vibration velocity component is carried out to linear combination, the two vibration velocity output component v of complex field that structure makes new advances ₁, v ₂:

$\left\{\begin{array}{c}{v}_1=v_x+{\mathrm{jv}}_y=A_p{\mathrm{\&Phi;}}_{1}s+n_1\\ {\mathrm{<figure-callout\; id="2"\; label="v"\; filenames="1.png,HDA00003583689700051.png"\; state="\{\{state\}\}">v</figure-callout>}}_2=v_x-{\mathrm{jv}}_y=A_p{\mathrm{\&Phi;}}_{2}s+n_2\end{array}\right.---\left(8\right)$

In formula, n ₁, n ₂it is corresponding ground unrest vector; Φ _x, Φ _yare all diagonal matrix, and have:

$\left\{\begin{array}{c}{\mathrm{\&Phi;}}_1=\mathrm{diag}[e^{{\mathrm{j\&theta;}}_1},\&CenterDot;\&CenterDot;\&CenterDot;,e^{{\mathrm{j\&theta;}}_D}]\\ {\mathrm{\&Phi;}}_2=\mathrm{diag}[e^{{-\mathrm{j\&theta;}}_1},\&CenterDot;\&CenterDot;\&CenterDot;,e^{{-\mathrm{j\&theta;}}_D}]\end{array}\right.---\left(9\right)$

1.3. the acoustic pressure vibration velocity Cross-covariance based on broad sense acoustic energy flow

The two vibration velocity output components of new complex field are all identical to the amplitude response of the incoming signal of all directions, and its vector feature is only embodied in phase place, therefore can completely receive the signal of in any direction incident.Another feature is that the new two ground unrests of vibration velocity output component of complex field and the ground unrest of sound pressure channel is still incoherent, can find out from following formula:

$\left\{\begin{array}{c}E\left(n_p{n}_1^H\right)=E\left(n_p{n}_x^H\right)+\mathrm{jE}\left(n_p{n}_y^H\right)=0\\ E\left(n_p{n}_2^H\right)=E\left(n_p{n}_x^H\right)-\mathrm{jE}\left(n_p{n}_y^H\right)=0\end{array}\right.---\left(10\right)$

Consider that acoustic pressure is as follows with the covariance matrix of new complex field pair vibration velocity components:

$\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="R"\; filenames="HDA00003583689700022.png,HDA00003583689700053.png"\; state="\{\{state\}\}">R</figure-callout>}}_1=E\left({\mathrm{pv}}_1^H\right)=A_p{R}_s{\mathrm{\&Phi;}}_1^H{A}_{\mathrm{<figure-callout\; id="2"\; label="p\; H\; R"\; filenames="1.png,HDA00003583689700051.png"\; state="\{\{state\}\}">p</figure-callout>}}^H& \\ R_{}{}_2=E\left(p_2^H\right)=A_p{R}_s{\mathrm{\&Phi;}}_2^H{A}_p^H\end{array}---\left(11\right)\right.$

In formula: R _s=E (ss ^h) be the incoming signal covariance matrix of D * D dimension.When incoming signal mutual when uncorrelated, R _sa diagonal matrix, the power that diagonal element is each signal.Because the two vibration velocity output components of new complex field are the expansions to former acoustic vector sensor array vibration velocity component, therefore new acoustic pressure vibration velocity Cross-covariance is called to the Cross-covariance based on broad sense acoustic energy flow.

R in practical application ₁, R ₂employing maximum likelihood estimator replaces:

$\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="R"\; filenames="HDA00003583689700022.png,HDA00003583689700053.png"\; state="\{\{state\}\}">R</figure-callout>}}_1=\frac{1}{M}\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{pv}}_1^{\mathrm{<figure-callout\; id="2"\; label="H\; R"\; filenames="1.png,HDA00003583689700051.png"\; state="\{\{state\}\}">H</figure-callout>}}& \\ R_{}{}_2=\frac{1}{M}\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{pv}}_2^H\end{array}\right.---\left(12\right)$

In formula, M is the fast umber of beats of sampling.

1.4. the orientation based on pencil of matrix ESPRIT algorithm is estimated

Making γ is a constant, examination pencil of matrix (R ₁, R ₂):

$R_1-\mathrm{\&gamma;}{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="1.png,HDA00003583689700051.png"\; state="\{\{state\}\}">R</figure-callout>}}_2=A_p{R}_s({\mathrm{\&Phi;}}_1^H-\mathrm{\&gamma;}{\mathrm{\&Phi;}}_2^H)A_p^H---\left(13\right)$

Due to A _prow full ranks, and R _snonsingular, rank of matrix is asked in (13) formula both sides:

$\mathrm{rank}(R_1-\mathrm{\&gamma;}{R}_2)=\mathrm{rank}({\mathrm{\&Phi;}}_1^H-\mathrm{\&gamma;}{\mathrm{\&Phi;}}_2^H)---\left(14\right)$

When

time, matrix

be nonsingular matrix, and work as

time, matrix order wanes becomes singular matrix.

all pencil of matrix (R ₁, R ₂) non-zero generalized eigenvalue.

Generally, covariance matrix R ₁, R ₂be full rank not, in solving the process of non-zero generalized eigenvalue, likely cause potential numerical value difficulty.The present invention adopts svd (SVD) and total least square method (TLS), can avoid this problem.

To R ₁svd:

${\mathrm{<figure-callout\; id="1"\; label="R"\; filenames="HDA00003583689700022.png,HDA00003583689700053.png"\; state="\{\{state\}\}">R</figure-callout>}}_1={\mathrm{U\&Sigma;V}}^H=\left[\begin{array}{cc}{U}_1& {\mathrm{<figure-callout\; id="2"\; label="U"\; filenames="1.png,HDA00003583689700051.png"\; state="\{\{state\}\}">U</figure-callout>}}_2\end{array}\right]\left[\begin{array}{cc}{\mathrm{\&Sigma;}}_1& 0\\ 0& 0\end{array}\right]\left[\begin{array}{c}{V_1}^H\\ {V_2}^H\end{array}\right]---\left(15\right)$

As mentioned before, Σ in formula ₁by R ₁the diagonal matrix that forms of non-zero singular value; U ₁, V ₁respectively left and right singular vector corresponding to non-zero singular value; U ₂, V ₂zero left and right singular vector corresponding to singular value.So can obtain:

$U_1^H(R_1-\mathrm{\&gamma;}{R}_2){\mathrm{<figure-callout\; id="1"\; label="V"\; filenames="HDA00003583689700022.png,HDA00003583689700053.png"\; state="\{\{state\}\}">V</figure-callout>}}_1={\mathrm{\&Sigma;}}_1-\mathrm{\&gamma;}{U}_1^H{R}_2{V}_1---\left(16\right)$

Visible, N * N dimension pencil of matrix (R ₁, R ₂) D non-zero generalized eigenvalue be that D * D ties up pencil of matrix

whole generalized eigenvalues.

This direction estimation method of the present invention, as shown in Figure 3, concrete calculation procedure is summarized as follows implementation method:

(1) the reception signal of pair array carries out narrow-band filtering, obtains the arrowband output signal on pending frequency.

(2) according to formula (8), calculate the two vibration velocity output components of new complex field.

(3) according to formula (12), estimate the acoustic pressure vibration velocity Cross-covariance based on broad sense acoustic energy flow.

(4) covariance matrix is carried out to svd, and estimate non-zero singular value matrix Σ ₁left and right singular vector matrix U with correspondence ₁, V ₁.

(5) to pencil of matrix

carry out generalized eigenvalue decomposition, thereby obtain the azimuth estimation value of incoming signal.

2. the concrete calculated examples that orientation is estimated

As shown in Figure 2, the centre frequency of arrowband, incident far field plane wave is 1kHz to 5 yuan of acoustic vector sensor array element positions.

2.1. the root-mean-square error simulation example of target Bearing Estimation

Adopt the method in direction estimation method proposed by the invention and document [4] to carry out emulation experiment 500 times simultaneously, having carried out respectively single goal and Bi-objective orientation estimates, the fast umber of beats of sampling of each experiment is 100, and in single goal situation, target level position angle is 30 °; In Bi-objective situation, target level position angle is respectively 30 °, 40 °.

First, the reception data of pair array are carried out narrow-band filtering, and calculate the two vibration velocity output component v of new complex field according to formula (8) ₁, v ₂, when Fig. 4, Fig. 5 have shown respectively single goal and Bi-objective situation, the output spectrum of each passage when signal to noise ratio (S/N ratio) (SNR) is 0dB.As seen from the figure, due to the filtering characteristic of self, the spectrum peak of conventional vibration velocity output component is less than sound pressure channel; And the two vibration velocity components of complex field have the receiving ability of omnidirectional, its spectrum peak and sound pressure channel are basically identical.

Then, according to formula (12), estimate covariance matrix based on broad sense acoustic energy flow to R ₁, R ₂, last application matrix bundle ESPRIT method calculates matrix to R ₁, R ₂between the invariable rotary factor, obtain azimuth estimation value.Fig. 6, Fig. 7 have provided respectively root mean square (RMS) error of target level position angle estimation in two kinds of situations and the relation curve of SNR.In figure, horizontal ordinate is SNR, and the ordinate of Fig. 6 is the RMS error that single goal is estimated, the ordinate of Fig. 7 is the RMS error sum that Bi-objective is estimated.

By Fig. 6 and Fig. 7, can be found out, under logarithmic coordinate, the root-mean-square error of two kinds of methods of estimation and signal to noise ratio (S/N ratio) are roughly linear, and method proposed by the invention has lower evaluated error all the time compared to the method for document [4].In the single goal situation shown in Fig. 6, when SNR=5dB, the method root-mean-square error of document [4] is down to 1 ° of left and right, and method proposed by the invention is when SNR=0dB, is just down to 1 °; In the Bi-objective situation shown in Fig. 7, when SNR=17dB, the method root-mean-square error of document [4] is down to 1 ° of left and right, and method proposed by the invention is when SNR=10dB, just can be down to 1 °.Visible, method proposed by the invention effectively combines the ability of the anti-isotropic noise of vector hydrophone, thereby has the lower signal-noise ratio threshold processed.

2.2.y near axle, simulation example is estimated in the orientation of direction incoming signal

Consider the situation of single goal incident, the incident direction of signal changes to 93 ° with the step-length of 0.5 ° from 87 °, signal to noise ratio snr=0dB, and the fast umber of beats of sampling is 100.Adopt respectively the two vibration velocity output component structure of conventional vibration velocity output component and complex field acoustic pressure vibration velocity Cross-covariance, and carry out orientation estimation.When Fig. 8 has shown target from the incident of y direction of principal axis, the output signal spectrum of each passage, Fig. 9 has shown the orientation estimated result of this two schemes.

By Fig. 8 and Fig. 9, can be found out, when signal is during from the incident of y direction of principal axis, vibration velocity channel v _xthoroughly do not receive signal, therefore adopt the ESPRIT orientation estimated result of acoustic pressure and conventional vibration velocity component to lose efficacy, and the two vibration velocity channel v of complex field ₁, v ₂the complete signal that receives of energy, still can provide high-precision orientation estimated result.

Claims (5)

1. an acoustic vector sensor array High Precision Long-distance direction estimation method, is characterized in that:

(1) the reception signal of two-dimensional vector hydrophone array is carried out to narrow-band filtering, obtain the arrowband output signal on pending frequency;

(2) two mutually orthogonal vibration velocity components of two-dimensional vector hydrophone array are carried out to linear combination in complex field, be converted to two new vibration velocity output components, construct the two vibration velocity components of complex field;

(3) adopt maximal possibility estimation to obtain acoustic pressure that some samplings take soon and the Cross-covariance pair of the two vibration velocity components of complex field, construct the acoustic pressure vibration velocity Cross-covariance pair based on broad sense acoustic energy flow;

(4) the ESPRIT algorithm of application matrix bundle calculate acoustic pressure vibration velocity Cross-covariance based on broad sense acoustic energy flow between the invariable rotary factor, thereby carry out orientation estimation.

2. acoustic vector sensor array High Precision Long-distance direction estimation method according to claim 1, is characterized in that the two vibration velocity components of described structure complex field specifically comprise: by the vibration velocity output component v of y direction _yj multiplies each other with complex unit, then with the vibration velocity output component v of x direction _xbe added and obtain new vibration velocity output component v ₁; By the vibration velocity output component v of y direction _ymultiply each other with negative complex unit-j, then with the vibration velocity output component v of x direction _xbe added and obtain new vibration velocity output component v ₂.

3. acoustic vector sensor array High Precision Long-distance direction estimation method according to claim 1, it is characterized in that described constructing acoustic pressure vibration velocity Cross-covariance based on broad sense acoustic energy flow to specifically comprising: the output of the sound pressure sensor array of the snap quantity of necessarily sampling is multiplied each other with the conjugate transpose of two new complex field vibration velocity output components respectively, and ask its mean value, obtain Cross-covariance based on broad sense acoustic energy flow to R ₁and R ₂.

4. acoustic vector sensor array High Precision Long-distance direction estimation method according to claim 1 and 2, is characterized in that: the ESPRIT algorithm of described application matrix bundle calculate acoustic pressure vibration velocity Cross-covariance based on broad sense acoustic energy flow between the invariable rotary factor in adopt svd and total least square method the Ill-conditioned Least-squares Problems of a larger dimension to be converted into the non-ill total least square problem of a less dimension.

5. acoustic vector sensor array High Precision Long-distance direction estimation method according to claim 3, is characterized in that: the ESPRIT algorithm of described application matrix bundle calculate acoustic pressure vibration velocity Cross-covariance based on broad sense acoustic energy flow between the invariable rotary factor in adopt svd and total least square method the Ill-conditioned Least-squares Problems of a larger dimension to be converted into the non-ill total least square problem of a less dimension.

Images