CN104730513A - Multistage sub-array focusing MVDR wave beam forming method - Google Patents

Abstract

The invention relates to a multistage sub-array focusing MVDR wave beam forming method. According to the multistage sub-array focusing MVDR wave beam forming method, a common rapid frequency-domain broadband wave beam forming algorithm and a sub-array focusing MVDR algorithm are combined, a target orientation is firstly and roughly measured through the rapid frequency-domain broadband wave beam forming algorithm, then sub-arrays are divided, finally the target orientation is accurately measured through the sub-array focusing MVDR algorithm, and rapid and high-accuracy wave beam forming is achieved. The multistage sub-array focusing MVDR wave beam forming method includes the steps of 1, firstly completing rapid frequency-domain broadband wave beam forming on data received by a linear array sonar device through one-time time-domain FFT and airspace phase compensation, and completing initial estimation of the target orientation; 2, then carrying out peak detection on a common wave beam forming curve, after a target is found, carrying out sub-array dividing on a frequency-domain array signal, carrying out sub-array focusing MVDR wave beam forming on the orientation nearby the target, carrying out secondary orientation detection, and obtaining the more accurate target orientation. By means of the multistage sub-array focusing MVDR wave beam forming method, the performance of a whole sonar positioning system is improved, and meanwhile the requirements for the real-time performance, the high resolution and the tolerance are met.

Description

A kind of classification submatrix focuses on MVDR Beamforming Method

Technical field

The invention belongs to sonar digital processing field, MVDR (MinimumVarianceDistortionlessResponse is focused in particular to a kind of classification submatrix, response algorithm that minimum variance is undistorted) Beamforming Method, various towed array sonar system can be applied to, realize the Wave beam forming of quick high accuracy.

Background technology

By the restriction of Rayleigh limit, for small-bore basic matrix, adopt traditional ripple to reach algorithm for estimating and be difficult to differentiate 2 or multiple target close in orientation.The MVDR algorithm of Capon proposition in document " High-resolutionfrequency-wavenumberspectrumanalysis " in 1969 is that a kind of ripple in theory with high-resolution performance reaches algorithm for estimating, but higher dimensional matrix can be related in calculating process invert, complexity is very high, and algorithm requires higher to signal to noise ratio (S/N ratio), stability is also poor.In order to address these problems, Swingler.D.N proposes a kind of not overlapping Length discrepancy submatrix Beamforming Method continuously in document " Alow-complexityMVDRbeamformerforusewithshortobservationt imes ", Tian Biao etc. propose a kind of similar MVDR algorithm based on even adjacent subarray disposal route in document " the real-time ripple of multiple submatrixes high-resolution reaches algorithm for estimating research ", these methods can also keep the high-resolution performance of MVDR algorithm while reducing calculated amount, and real-time is better.In addition, improve the signal to noise ratio (S/N ratio) of single array element by submatrix process, also better to the tolerance of signal to noise ratio (S/N ratio).

Although submatrix focuses on MVDR algorithm and effectively reduces computational complexity to a certain extent, also have certain gap apart from the real-time required by general sonar system.

Summary of the invention

The object of the invention is, overcome the limited resolution that conventional beamformer faces, the high-precision direction finding to target cannot be realized, and the computational complexity that the self-adaptation high-resolution Beamforming Method such as MVDR faces is high, the problem of poor stability, thus propose a kind of classification submatrix focusing MVDR Beamforming Method, to meet the demand of sonar system Wave beam forming in real-time, high resolution capacity and tolerance simultaneously.

The present invention considers the huge advantage of conventional beams frequency domain Broadband Beamforming Method in real-time and tolerance, in order to solve the problem of the passive direction finding of small-bore horizontal basic matrix high precision, the thinking that the present invention proposes a kind of classification Wave beam forming is: common fast frequency-domain broad-band EDFA and submatrix are focused on MVDR algorithm and combines, complete target azimuth bigness scale by fast frequency-domain broad-band EDFA algorithm, the accurate measurement to target azimuth is completed again by submatrix focusing MVDR algorithm, thus improve the performance of whole sonar orientation system, meet real-time simultaneously, the needs of high-resolution and tolerance.A kind of towed array sonar device, can be a towed array, also can be shell side cooler, described linear array be made up of multiple nautical receiving set.Here establish array element number N, array element distance d, target incident direction θ, array element Received signal strength is expressed as x (t), and signal band scope is f _min~ f _max; The velocity of sound is c, and data snap length is L.

Method detailed process of the present invention as shown in Figure 1, first the data that linear array receives are converted and spatial domain phase compensation by a time domain FFT, complete rapid wideband frequency-domain beamforming, because conventional beams formation resolving power is limited, this step just completes to target azimuth according to a preliminary estimate, and what beam separation can be arranged is wider.Then, peakvalue's checking is carried out to conventional beams forming curves, after finding target, Subarray partition is carried out to frequency domain array signal, only submatrix is done to the orientation of target proximity and focus on MVDR Wave beam forming, then carry out secondary orientation detection, obtain more accurate target azimuth.

Concrete steps are described below:

11) linear array data input

Receive spacing wave by linear array, obtain the time-domain signal of M array element, the data of getting L sampled point length form a data snap, input as data.

X(t)=[x ₁(k) x ₂(k) … x _M(k)] ^T

Wherein, x _m(k)=[x _m(1) x _m(2) ... x _m(L)]

12) rapid wideband frequency-domain beamforming

To data snap X (t) at frequency band range f _min~ f _maxinside do fast frequency-domain broad-band EDFA, detailed process is as follows.

First each array element signals of pair array signal X (t) is time domain FFT, intercepts f _min~ f _maxthe frequency-region signal of scope is expressed as follows:

X(f)=[x ₁(f) x ₂(f) … x _M(f)] ^T

Wherein, x _m(f)=[x _m(f _min) ... x _m(f _max)]

To different frequency component f _istructure steering vector:

$w(f_i,{\mathrm{\θ}}_l)=\left[\begin{array}{cccc}1& e^{-j2\mathrm{\π}{f}_{i}d\cos \left({\mathrm{\θ}}_l\right)/c}& ...& e^{-j2\mathrm{\π}{f}_i\·(M-1)\·d\cos \left({\mathrm{\θ}}_l\right)/c}\end{array}\right]$

Then carry out phase compensation, the wave beam obtaining each frequency component exports:

y(f,θ _l)=w(f _i,θ _l)X(f)

Finally summed square is carried out in each wave beam of different frequency component output and obtains beam pattern:

$\begin{array}{c}{P}_{\mathrm{CBF}}\left({\mathrm{\θ}}_l\right)=y(f,{\mathrm{\θ}}_l)\·y*(f_i,{\mathrm{\θ}}_l)\\ =\underset{i=\min }{\overset{\max }{\mathrm{\Σ}}}y(f_i,{\mathrm{\θ}}_l)\·y*(f_i,{\mathrm{\θ}}_l),l=\mathrm{1,2},...,180\end{array}$

In fact whole process is two-dimensional Fourier transform, first does Fourier transform to array element signals, then does spatial fourier transform by frequency spectrum that array element extracts each conversion.

21) orientation detection

For wave beam curve P _cBF(θ _l) (l=1,2 ..., 180 °) and carry out orientation detection, first peakvalue's checking is carried out to wave beam curve, detection coefficient DT is calculated to each peak value:

$\mathrm{DT}=20\lg \frac{P_{\mathrm{CBF}}\left({\mathrm{\θ}}_{\mathrm{peak}}\right)-E\left[P_{\mathrm{CBF}}\left({\mathrm{\θ}}_l\right)\right]}{\mathrm{\δ}\left[P_{\mathrm{CBF}}\left({\mathrm{\θ}}_l\right)\right]}$

Wherein P _cBF(θ _peak) represent target peak, E [P _cBF(θ _l)] represent the average of wave beam curve, δ [P _cBF(θ _l)] represent the fluctuating value of wave beam curve.Sonar engineering field, generally gets detection coefficient DT=6dB, and at this moment detection probability is 95.4%.

Suppose to detect Q target, orientation is expressed as θ _q(k=1,2 ..., Q).

22) submatrix focuses on MVDR Wave beam forming

To each target θ detected _q(k=1,2 ..., Q), do submatrix to the orientation (such as 20 ° of scopes) near it and focus on MVDR Wave beam forming, detailed process is as follows:

According to the signal of accompanying drawing 2, linear array is divided into N number of submatrix, each submatrix contains Ms array element, then have

Last submatrix length is:

X' is shown in definition _nt () represents the time domain array signal of the n-th submatrix, then the frequency domain vectors X' of the n-th submatrix Received signal strength _nbe expressed as:

${X^{\′}}_n=\left[\begin{array}{cccc}{X}_{(n-1)M_s+1}\left(f\right);& X_{(n-1)M_s+2}\left(f\right);& ...;& X_{(n-1)M_s+M_s}\left(f\right)\end{array}\right]$

Wherein, X _m(f)=[X _m(f ₁) X _m(f ₂) ... X _m(f _k)], K represents the number of frequency component.

To observed ray θ, structure frequency f _kcorresponding submatrix direction vector is

$a(\mathrm{\θ},f_k)={\left[\begin{array}{cccc}1& e^{-j2\mathrm{\π}{f}_{k}d\cos \left(\mathrm{\θ}\right)/c}& ...& e^{-j2\mathrm{\π}{f}_k(M_s-1)\·d\cos \left(\mathrm{\θ}\right)}\end{array}/c\right]}^T$

Do beam direction phase compensation to each submatrix frequency domain array signal, obtaining submatrix beamformer output is:

Y _n(θ)=[Y _n(θ,f ₁) Y _n(θ,f ₂) … Y _n(θ,f _K)]

Wherein, Y _n(θ, f _k)=a ^h(θ, f _k) X' _n(f _k)

By Y _n(θ) be considered as the array element signals of the n-th Virtual array, then the array signal of virtual array is expressed as:

Y(θ)=[Y ₁(θ); Y ₂(θ); …; Y _N(θ)]

According to plane wave approximation, with first submatrix for benchmark, delay and focusing is carried out to each submatrix:

${\hat{Y}}_n\left(\mathrm{\θ}\right)=Y_n\left(\mathrm{\θ}\right)\·\mathrm{eye}\left[{\hat{a}}_n\right]$

Wherein ${\hat{a}}_n=\left[\begin{array}{cccc}{e}^{-j2\mathrm{\π}{f}_1(n-1)M_{s}d\cos \left(\mathrm{\θ}\right)/c}& e^{-j2\mathrm{\π}{f}_2(n-1)M_{s}d\cos \left(\mathrm{\θ}\right)/c}& ...& e^{-j2\mathrm{\π}{f}_K(n-1)M_{s}d\cos \left(\mathrm{\θ}\right)}\end{array}/c\right]$

After obtaining submatrix focusing alignment like this, the covariance matrix of virtual array is:

$\hat{R}\left(\mathrm{\θ}\right)=\frac{1}{M}\hat{Y}\left(\mathrm{\θ}\right)\·{\hat{Y}}^H\left(\mathrm{\θ}\right)$

Wherein, $\hat{Y}\left(\mathrm{\θ}\right)=\left[\begin{array}{cccc}{\hat{Y}}_1\left(\mathrm{\θ}\right);& {\hat{Y}}_2\left(\mathrm{\θ}\right);& ...;& {\hat{Y}}_M\left(\mathrm{\θ}\right)\end{array}\right]$

Then the output power spectrum in θ direction is estimated as:

$P_{\mathrm{SA}-\mathrm{MVDR}}\left(\mathrm{\θ}\right)=w^H\left(\mathrm{\θ}\right){\hat{R}}^{-1}\left(\mathrm{\θ}\right)w\left(\mathrm{\θ}\right)=\frac{1}{1^H\·{\hat{R}}^{-1}\left(\mathrm{\θ}\right)\·1}$

Wherein, optimal weight vector $w\left(\mathrm{\θ}\right)=\frac{{\hat{R}}^{-1}\left(\mathrm{\θ}\right)\·1}{1^H\·{\hat{R}}^{-1}\left(\mathrm{\θ}\right)\·1}.$

23) to the wave beam curve P that submatrix focusing MVDR obtains _sA-MVDR(θ) carry out secondary orientation detection, obtain more accurate target azimuth θ _q' (k=1,2 ..., Q), export target azimuth θ _q' (k=1,2 ..., Q).

Repeat above step.

On the basis of technique scheme, further, in step 4), generally linear array is divided into 3-5 submatrix comparatively suitable, significantly can reduces the dimension of MVDR matrix inversion on the one hand, reduce operand; Also the resolving power of algorithm can not be affected on the other hand.

On the basis of technique scheme, further, in step 4), submatrix focuses on the covariance matrix of the rear virtual array of alignment for the matrix of N*N, dimension is much smaller than the full battle array covariance matrix of M*M.Therefore, the calculated amount of corresponding matrix inversion will be reduced greatly, thus make real-time high resolution algorithm become possibility.

The invention has the advantages that:

(1) adopt classification Beamforming Method, take full advantage of the real-time high-efficiency performance of conventional beams formation and the high resolution capacity of adaptive M VDR algorithm, and the advantage of submatrix process, meet real-time and high-resolution requirement simultaneously.

(2) submatrix focuses on the correlativity that MVDR method can ensure signal in relatively little aperture, obtain higher signal to noise ratio (S/N ratio), strengthen the tolerance to environment, and the basic matrix more to large aperture, array element number, then MVDR Wave beam forming is carried out by dividing submatrix, greatly can reduce the dimension of matrix inversion, ensure the reversibility of matrix, reduce computational complexity.

(3) test figure proves that method of the present invention is effective.

The present invention may be used for various towed array sonar system, realizes the Wave beam forming of quick high accuracy.

Accompanying drawing explanation

Fig. 1 is the process flow diagram that classification submatrix of the present invention focuses on MVDR Beamforming Method;

Fig. 2 is Subarray partition strategy schematic diagram of the present invention;

Fig. 3 a and 3b is the bearing history result of lake of the present invention examination data processing; Wherein, Fig. 3 a is the result of orientation rough estimate, and Fig. 3 b carries out submatrix to focus on the result that MVDR beamforming algorithm accurately estimates;

Fig. 4 is the Wave beam forming curve map of lake of the present invention examination data processing.

Specific embodiment

Below in conjunction with certain Lake trial data and accompanying drawing, the specific embodiment of the present invention is described in further detail.

The Lake trial data of input data towed array sonar, basic matrix nautical receiving set number M=32, nautical receiving set spacing d=0.15m, array length 4.8m, sample rate f s=30KHz, velocity of sound c=1516m/s.Process bandwidth 2000 ~ 5000Hz, snap length L=2048, basic matrix is divided into 4 submatrixs, and each submatrix 8 array elements, Wave beam forming is set to 1 °, does the Wave beam forming of 1 ~ 180 ° of scope.

Adopt classification submatrix to focus on MVDR Beamforming Method, concrete steps are as follows:

Step 1: the step 101 in corresponding diagram 1, to the sampled signal of M=32 array element, the array data getting L=2048 sampled point length forms a snap, inputs as data.

X(t)=[x ₁(k) x ₂(k) … x _M(k)] ^T

Wherein, x _m(k)=[x _m(1) x _m(2) ... x _m(L)]

Step 2: step 102, step 103 in corresponding diagram 1, in frequency band range 2000 ~ 5000Hz, do fast frequency-domain broad-band EDFA to data snap X (t), detailed process is as follows.

First, the FFT calculating frequency band range corresponding counts:

f _min=2000/fs×2048=136

f _max=5000/fs×2048=342

Each array element signals of pair array signal X (t) is time domain 2048 FFT, intercepts f _min~ f _maxthe frequency-region signal of scope is expressed as follows:

X(f)=[x ₁(f) x ₂(f) … x _M(f)] ^T

Wherein, x _m(f)=[x _m(f _min) ... x _m(f _max)]

To different frequency component f _iarray element signals carry out phase compensation, obtain wave beam export be:

y(f,θ _l)=w(f _i,θ _l)X(f)

Wherein steering vector: $w(f_i,{\mathrm{\θ}}_l)=\left[\begin{array}{cccc}1& e^{-j2\mathrm{\π}{f}_{i}d\cos \left({\mathrm{\θ}}_l\right)/c}& ...& e^{-j2\mathrm{\π}{f}_i\·(M-1)\·d\cos \left({\mathrm{\θ}}_l\right)/c}\end{array}\right]$

Summed square is carried out in each wave beam of different frequency component output and obtains beam pattern:

$\begin{array}{c}{P}_{\mathrm{CBF}}\left({\mathrm{\θ}}_l\right)=y(f,{\mathrm{\θ}}_l)\·y*(f_i,{\mathrm{\θ}}_l)\\ =\underset{i=\min }{\overset{\max }{\mathrm{\Σ}}}y(f_i,{\mathrm{\θ}}_l)\·y*(f_i,{\mathrm{\θ}}_l),l=\mathrm{1,2},...,180\end{array}$

P _cBF(θ _l) (l=1,2 ..., 180) and be the wave beam curve of fast frequency-domain broad-band EDFA.

Step 3: the step 104 in corresponding diagram 1, to wave beam curve P _cBF(θ _l) (l=1,2 ..., 180) and carry out orientation detection.First the average E [P of compute beam curve _cBF(θ _l)] and variance δ [P _cBF(θ _l)], then calculate each orientation detection coefficient

$\mathrm{DT}=20\lg \frac{P_{\mathrm{CBF}}\left({\mathrm{\θ}}_{\mathrm{peak}}\right)-E\left[P_{\mathrm{CBF}}\left({\mathrm{\θ}}_l\right)\right]}{\mathrm{\δ}\left[P_{\mathrm{CBF}}\left({\mathrm{\θ}}_l\right)\right]}$

Get detection coefficient threshold value DT=6dB, if the detection DT [B in i direction _i(t _k)]>=6dB, then illustrate that there is target in i direction.According to said process, suppose to detect Q target, orientation is expressed as θ _q(k=1,2 ..., Q).

Step 4: step 105, step 106, step 107, step 108 in corresponding diagram 1, to each target θ detected _q(k=1,2 ..., Q), do submatrix to the orientation of its neighbouring 15 ° of scopes and focus on MVDR Wave beam forming, detailed process is as follows.

Linear array is divided into 4 submatrixs, each submatrix contains 8 array elements, the frequency domain vectors X' of the n-th submatrix Received signal strength _nbe expressed as:

${X^{\′}}_n=\left[\begin{array}{cccc}{X}_{(n-1)M_s+1}\left(f\right);& X_{(n-1)M_s+2}\left(f\right);& ...;& X_{(n-1)M_s+M_s}\left(f\right)\end{array}\right]$

Wherein, X _m(f)=[X _m(f _min) X _m(f _min+1) ... X _m(f _max)].

Do beam direction phase compensation to each submatrix frequency domain array signal, obtaining submatrix beamformer output is:

Y _n(θ)=[Y _n(θ,f ₁) Y _n(θ,f ₂) … Y _n(θ,f _K)]

Wherein, Y _n(θ, f _k)=a ^h(θ, f _k) X' _n(f _k)

Submatrix direction vector is: $a(\mathrm{\θ},f_k)={\left[\begin{array}{cccc}1& e^{-j2\mathrm{\π}{f}_{k}d\cos \left(\mathrm{\θ}\right)/c}& ...& e^{-j2\mathrm{\π}{f}_k(M_s-1)\·d\cos \left(\mathrm{\θ}\right)}\end{array}/c\right]}^T$

By Y _n(θ) be considered as the array element signals of the n-th Virtual array, the array signal obtaining virtual array is:

Y(θ)=[Y ₁(θ); Y ₂(θ); …; Y _N(θ)]

According to plane wave approximation, with first submatrix for benchmark, delay and focusing is carried out to each submatrix:

${\hat{Y}}_n\left(\mathrm{\θ}\right)=Y_n\left(\mathrm{\θ}\right)\·\mathrm{eye}\left[{\hat{a}}_n\right]$

Wherein ${\hat{a}}_n=\left[\begin{array}{cccc}{e}^{-j2\mathrm{\π}{f}_1(n-1)M_{s}d\cos \left(\mathrm{\θ}\right)/c}& e^{-j2\mathrm{\π}{f}_2(n-1)M_{s}d\cos \left(\mathrm{\θ}\right)/c}& ...& e^{-j2\mathrm{\π}{f}_K(n-1)M_{s}d\cos \left(\mathrm{\θ}\right)}\end{array}/c\right]$

Thus the covariance matrix of virtual array is after obtaining submatrix focusing alignment:

$\hat{R}\left(\mathrm{\θ}\right)=\frac{1}{M}\hat{Y}\left(\mathrm{\θ}\right)\·{\hat{Y}}^H\left(\mathrm{\θ}\right)$

Wherein, $\hat{Y}\left(\mathrm{\θ}\right)=\left[\begin{array}{cccc}{\hat{Y}}_1\left(\mathrm{\θ}\right);& {\hat{Y}}_2\left(\mathrm{\θ}\right);& ...;& {\hat{Y}}_M\left(\mathrm{\θ}\right)\end{array}\right]$

Then the output power spectrum in θ direction is estimated as:

$P_{\mathrm{SA}-\mathrm{MVDR}}\left(\mathrm{\θ}\right)=w^H\left(\mathrm{\θ}\right){\hat{R}}^{-1}\left(\mathrm{\θ}\right)w\left(\mathrm{\θ}\right)=\frac{1}{1^H\·{\hat{R}}^{-1}\left(\mathrm{\θ}\right)\·1}$

Wherein, optimal weight vector $w\left(\mathrm{\θ}\right)=\frac{{\hat{R}}^{-1}\left(\mathrm{\θ}\right)\·1}{1^H\·{\hat{R}}^{-1}\left(\mathrm{\θ}\right)\·1}.$

Obtain the wave beam output that submatrix focuses on MVDR like this: P _sA-MVDR(θ _l') (l'=q-7 ..., q ..., q+7).

Step 5: the step 109 in corresponding diagram 1, carries out orientation detection to the wave beam curve that submatrix focusing MVDR obtains, obtains more accurate target azimuth θ _q' (q=1,2 ..., Q).

Step 6: the step 110 in corresponding diagram 1, exports accurate target azimuth θ _q' (k=1,2 ..., Q).

Step 7: repeat above step.

Fig. 3 gives bearing history result, Fig. 3 a is the result of orientation rough estimate, and Fig. 3 b carries out submatrix to focus on the result that MVDR beamforming algorithm accurately estimates, can find out, carry out after secondary submatrix focuses on MVDR Wave beam forming, object beam is meticulousr.The wave beam section marked in bearing history figure amplified, as shown in Figure 4, the orientation of these two targets is separated by 4 ° to wave beam curve, and being formed by conventional beams can only rapid preliminary estimating target orientation, and can not be differentiated out.The region of 15 ° near selected target orientation, adopts submatrix to focus on MVDR Wave beam forming further, and two targets can accurately be differentiated be opened, and effect is fine.

In order to real-time process advantage of the present invention is described, add up the time span that matlab routine processes test figure needs specially.Signal sampling rate fs=30000Hz, data snap length is 2048 points, for 30 array element linear arrays, be divided into 4 submatrixs, conventional beamformer 47.6ms consuming time, submatrix focuses on MVDR algorithm 418.5ms consuming time, and the present invention expends time in as 57.1ms, algorithm real-time is suitable with conventional beamformer, and real-time is very high, in fact, length is the corresponding time span of fast beat of data of 2048 is 2048/fs=68ms.Can find out, adopt the present invention can process test figure in real time completely, and there is higher target resolution characteristic.

In a word, the present invention can realize obtaining high-precision Wave beam forming in real time, the accurate target azimuth of quick obtaining.

It should be noted last that, above embodiment is only in order to illustrate technical scheme of the present invention and unrestricted.Although with reference to embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that, modify to technical scheme of the present invention or equivalent replacement, do not depart from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of right of the present invention.

Claims (8)

1. a classification submatrix focuses on MVDR Beamforming Method, common fast frequency-domain broad-band EDFA algorithm is focused on MVDR algorithm with submatrix and combines by the method, first complete target azimuth bigness scale by fast frequency-domain broad-band EDFA, divide submatrix again, finally complete the accurate measurement to target azimuth by submatrix focusing MVDR algorithm, realize the Wave beam forming of quick high accuracy.

2. classification submatrix according to claim 1 focuses on MVDR Beamforming Method, it is characterized in that, said method comprising the steps of:

1) first, the data that towed array sonar device receives are converted and spatial domain phase compensation by a time domain FFT, completes rapid wideband frequency-domain beamforming, complete to target azimuth according to a preliminary estimate;

2) then, peakvalue's checking is carried out to conventional beams forming curves, after finding target, Subarray partition is carried out to frequency domain array signal, and only submatrix is done to the orientation of target proximity and focus on MVDR Wave beam forming, carry out secondary orientation detection, obtain more accurate target azimuth.

3. classification submatrix according to claim 2 focuses on MVDR Beamforming Method, and it is characterized in that, described towed array sonar device is made up of multiple hydrophone array unit, is a towed array or shell side cooler; If nautical receiving set array element number N, array element distance d, target incident direction θ, array element Received signal strength is expressed as x (t), and signal band scope is f _min~ f _max; The velocity of sound is c, and data snap length is L.

4. classification submatrix according to claim 3 focuses on MVDR Beamforming Method, and it is characterized in that, the concrete steps of described method are as follows:

11) linear array data input;

Towed array sonar device receives spacing wave, and obtain the time-domain signal of M array element, the data of getting L sampled point length form a data snap, input as data:

X(t)＝[x ₁(k) x ₂(k) … x _M(k)] ^T；

Wherein, x _m(k)=[x _m(1) x _m(2) ... x _m(L)];

12) rapid wideband frequency-domain beamforming;

To data snap X (t) at frequency band range f _min~ f _maxinside do fast frequency-domain broad-band EDFA:

First, each array element signals of pair array signal X (t) is time domain FFT, intercepts f _min~ f _maxthe frequency-region signal of scope is expressed as follows:

X(f)＝[x ₁(f) x ₂(f) … x _M(f)] ^T；

Wherein, x _m(f)=[x _m(f _min) ... x _m(f _max)];

To different frequency component f _istructure steering vector:

$w(f_i,{\mathrm{\θ}}_l)=\left[\begin{array}{cccc}1& e^{-j2\mathrm{\π}{f}_{i}d\cos \left({\mathrm{\θ}}_l\right)/c}& ...& e^{-j2\mathrm{\π}{f}_i\·(M-1)\·d\cos \left({\mathrm{\θ}}_l\right)/c}\end{array}\right];$

Then, carry out phase compensation, the wave beam obtaining each frequency component exports:

y(f,θ _l)＝w(f _i,θ _l)X(f)；

Finally, summed square is carried out in each wave beam of different frequency component output and obtains beam pattern:

$\begin{array}{c}{P}_{\mathrm{CBF}}\left({\mathrm{\θ}}_l\right)=y(f,{\mathrm{\θ}}_l)\·y*(f_i,{\mathrm{\θ}}_l)\\ =\underset{i=\min }{\overset{\max }{\mathrm{\Σ}}}y(f_i,{\mathrm{\θ}}_l)\·y*(f_i,{\mathrm{\θ}}_l),l=\mathrm{1,2},...,180\end{array};$

21) orientation detection;

For wave beam curve P _cBF(θ _l) (l=1,2 ..., 180 °) and carry out orientation detection, first, peakvalue's checking is carried out to wave beam curve, detection coefficient DT is calculated to each peak value:

$\mathrm{DT}=20\lg \frac{P_{\mathrm{CBF}}\left({\mathrm{\θ}}_{\mathrm{peak}}\right)-E\left[P_{\mathrm{CBF}}\left({\mathrm{\θ}}_l\right)\right]}{\mathrm{\δ}\left[P_{\mathrm{CBF}}\left({\mathrm{\θ}}_l\right)\right]};$

Wherein, P _cBF(θ _peak) represent target peak, E [P _cBF(θ _l)] represent the average of wave beam curve, δ [P _cBF(θ _l)] represent the fluctuating value of wave beam curve;

Suppose to detect Q target, orientation is expressed as θ _q(k=1,2 ..., Q);

22) submatrix focuses on MVDR Wave beam forming;

To each target θ detected _q(k=1,2 ..., Q), submatrix is done to the orientation near it and focuses on MVDR Wave beam forming:

Linear array is divided into N number of submatrix, and each submatrix contains M _sindividual array element, then have

Finally, a sub-array length degree is:

Definition x' _nt () represents the time domain array signal of the n-th submatrix, then the frequency domain vectors X' of the n-th submatrix Received signal strength _nbe expressed as:

${X^{\′}}_n=\left[\begin{array}{cccc}{X}_{(n-1)M_s+1}\left(f\right);& X_{(n-1)M_s+2}\left(f\right);& ...;& X_{(n-1)M_s+M_s}\left(f\right)\end{array}\right];$

Wherein, X _m(f)=[X _m(f ₁) X _m(f ₂) ... X _m(f _k)], K represents the number of frequency component;

To observed ray θ, structure frequency f _kcorresponding submatrix direction vector is:

$a(\mathrm{\θ},f_k)={\left[\begin{array}{cccc}1& e^{-j2\mathrm{\π}{f}_{k}d\cos \left(\mathrm{\θ}\right)/c}& ...& e^{-j2\mathrm{\π}{f}_k(M_s-1)\·d\cos \left(\mathrm{\θ}\right)}\end{array}/c\right]}^T;$

Do beam direction phase compensation to each submatrix frequency domain array signal, obtaining submatrix beamformer output is:

Y _n(θ)＝[Y _n(θ,f ₁) Y _n(θ,f ₂) … Y _n(θ,f _K)]；

Wherein, Y _n(θ, f _k)=a ^h(θ, f _k) X' _n(f _k);

By Y _n(θ) be considered as the array element signals of the n-th Virtual array, then the array signal of virtual array is expressed as:

Y(θ)=[Y ₁(θ); Y ₂(θ); …; Y _N(θ)]；

According to plane wave approximation, with first submatrix for benchmark, delay and focusing is carried out to each submatrix:

${\hat{Y}}_n\left(\mathrm{\θ}\right)=Y_n\left(\mathrm{\θ}\right)\·\mathrm{eye}\left[{\hat{a}}_n\right];$

Wherein, ${\hat{a}}_n=\left[\begin{array}{cccc}{e}^{-j2\mathrm{\π}{f}_1(n-1)M_{s}d\cos \left(\mathrm{\θ}\right)/c}& e^{-j2\mathrm{\π}{f}_2(n-1)M_{s}d\cos \left(\mathrm{\θ}\right)/c}& ...& e^{-j2\mathrm{\π}{f}_K(n-1)M_{s}d\cos \left(\mathrm{\θ}\right)}\end{array}/c\right];$

Like this, after obtaining submatrix focusing alignment, the covariance matrix of virtual array is:

$\hat{R}\left(\mathrm{\θ}\right)=\frac{1}{M}\hat{Y}\left(\mathrm{\θ}\right)\·{\hat{Y}}^H\left(\mathrm{\θ}\right);$

Wherein, $\hat{Y}\left(\mathrm{\θ}\right)=\left[\begin{array}{cccc}{\hat{Y}}_1\left(\mathrm{\θ}\right);& {\hat{Y}}_2\left(\mathrm{\θ}\right);& ...;& {\hat{Y}}_M\left(\mathrm{\θ}\right)\end{array}\right];$

Then the output power spectrum in θ direction is estimated as:

$P_{\mathrm{SA}-\mathrm{MVDR}}\left(\mathrm{\θ}\right)=w^H\left(\mathrm{\θ}\right){\hat{R}}^{-1}\left(\mathrm{\θ}\right)w\left(\mathrm{\θ}\right)=\frac{1}{1^H\·{\hat{R}}^{-1}\left(\mathrm{\θ}\right)\·1};$

Wherein, optimal weight vector $w\left(\mathrm{\θ}\right)=\frac{{\hat{R}}^{-1}\left(\mathrm{\θ}\right)\·1}{1^H\·{\hat{R}}^{-1}\left(\mathrm{\θ}\right)\·1};$

23) to the wave beam curve P that submatrix focusing MVDR obtains _sA-MVDR(θ) carry out secondary orientation detection, obtain more accurate target azimuth θ _q' (k=1,2 ..., Q) and export.

5. classification submatrix according to claim 4 focuses on MVDR Beamforming Method, it is characterized in that, described step 22) in, generally linear array is divided into 3 ~ 5 submatrixs.

6. classification submatrix according to claim 4 focuses on MVDR Beamforming Method, it is characterized in that, described step 21) in, generally get detection coefficient DT=6dB, at this moment detection probability is 95.4%.

7. classification submatrix according to claim 4 focuses on MVDR Beamforming Method, it is characterized in that, described step 22) in, to each target θ detected _q(k=1,2 ..., Q) near orientation do submatrix and focus on MVDR Wave beam forming, the orientation value of described vicinity is θ _q-Δ θ ~ θ _q+ Δ θ, Δ θ span is 5 ~ 10 °.

8. classification submatrix according to claim 4 focuses on MVDR Beamforming Method, it is characterized in that, described step 23) in the method that detects with first power position of secondary orientation detection completely the same.