CN101387701B - Passive time reversal reverberation suppression method based on forward prediction - Google Patents

Abstract

The invention discloses a passive time reversion reverberation inhibition method based on forward predication, comprising: first, processing submarine reverberation and target echo via modeling and multiplying window operations, obtaining a signal sub space corresponding to the target distance, using forward predication to evaluate a submarine reverberation sub space corresponding to the target distance, deriving a group of optimal weighted vectors w of SRA, using the optimal weighted vectors w to process weighted sum on the echo signal x(r, omega) received by SRA, thereby restraining the reverberation component of the echo signal and improving the target echo signal component of the echo signal most, to further improve the echo signal mixed ratio. The invention can effectively restrain reverberation, keep or enhance target echo, to effectively improve echo signal mixed ratio, improve the target detection property of active sonar and improve practicality.

Description

A kind of passive time reversal reverberation suppression method based on forward prediction

Technical field

The present invention relates to a kind of underwater sound passive time reversal reverberation suppression method.

Background technology

Along with development of technology, active sonar constantly develops to low frequency, high-power, large aperture direction.Because the reduction of frequency of operation, bottom reverberation more and more becomes the key constraints of shallow sea Low-Frequency Active Sonar serviceability.Because its a series of special nature, Reverberation Rejection are the difficult problems that the puzzlement sonar signal is handled always.Handle time reversal (the anti-processing when also being called for short usually) be a kind of with marine environment itself the new adaptive processing method of matched filter during as sky, can be under the condition that does not have prioris such as environment and transducer array array structure, focus on when realizing the adaptive space of sound field.In recent years, Reverberation Rejection and echo letter that this time reversal, disposal route was applied in the waveguide of shallow sea gradually mix than strengthening, and Chinese scholars has proposed multiple implementation method:

1) anti-disposal route directly the time;

2) dereverberation groove method to set up initiatively the time;

Dereverberation inhibition method when 3) passive;

4) before and after under the reverberation constraint condition at zero point based on the time inverse operator letter that decomposes mix than this method of Enhancement Method.

These methods all have certain limitation in actual applications.Anti-disposal route requires to place one at target proximity and surveys sound source in the time of directly, and this is to be impossible in the reality.Dereverberation groove method to set up has been utilized pure bottom reverberation subspace in the time of initiatively, when comprising target and bottom reverberation signal in the echoed signal simultaneously, do not having under the condition of other prior imformation, it is very difficult wanting separate targets echo subspace and bottom reverberation subspace, this moment is if use this reverberation suppression method, target echo and reverberation are suppressed simultaneously, and, this is a kind of active process mode, needs in the reality may exist the last active send-receive that carries out successively of distance of target to handle to each.The popularization of dereverberation groove method to set up in passive situation when dereverberation inhibition method is active when passive, directly the reverb signal that receives is suppressed to handle, do not carry out active emission-reception processing on the distance and need not detect, may make target echo and reverberation repressed problem simultaneously but exist equally at each.Before and after under the reverberation constraint condition at zero point based on the time inverse operator letter that decomposes mix and solved the problems referred to above than Enhancement Method, but this is a kind of active process method, need may exist the last active send-receive that carries out successively of distance of target to handle to each, this needs to expend the long time in actual applications.

Summary of the invention

In order to overcome the deficiencies in the prior art, the invention provides a kind of during based on forward prediction passive dereverberation suppress new method, can solve that dereverberation inhibition method may make target echo and reverberation repressed problem simultaneously when current, dereverberation inhibition method need may exist the distance of target to carry out the initiatively repetitive operation of send-receive processing successively to each when avoiding initiatively simultaneously, for the practical engineering application of handling time reversal in Reverberation Rejection provides a kind of method effectively reliably.

The technical solution adopted for the present invention to solve the technical problems may further comprise the steps:

1. the modeling of bottom reverberation and target echo and windowing process

Array time reversal (SRA-Source Receiver Array) is taken as a transmitting-receiving of being made up of N array element to be closed and puts orthogonal array.SRA is at t=t ₀The pulse signal that it is τ that the moment is launched a pulsewidth, the echoed signal that this process produces is recorded on each array element.Ignore repeatedly scattering, with the SRA horizontal range be that near the r scatterer (comprising target and bottom scattering body) can be represented by a time window the effect of the echo that receives:

$w(t,r,\mathrm{\&Delta;})=\left\{\begin{array}{cc}1& |t-t_c-{\mathrm{<figure-callout\; id="0"\; label="t"\; filenames="H2008102318742E00021.png,H2008102318742E00022.png"\; state="\{\{state\}\}">t</figure-callout>}}_0|\&le;\mathrm{\&Delta;}/2\\ 0& |t-t_c-{\mathrm{<figure-callout\; id="0"\; label="t"\; filenames="H2008102318742E00021.png,H2008102318742E00022.png"\; state="\{\{state\}\}">t</figure-callout>}}_0|>\mathrm{\&Delta;}/2\end{array}\right.---\left(1\right)$

Wherein, w express time window, t express time, t _c=2r/c ₀Be the time proximity of propagating back and forth, r is the horizontal range of scatterer and SRA, c ₀It is average mode group velocity.Δ is the width of the time window that is used for handling, for a narrow band signal, typically, gets Δ=τ.

By turning back to the reverb signal windowing process of SRA, certain reverb signal constantly can be similar to and think to produce by single bottom scattering volume scattering.Fig. 1 has provided target and the bottom scattering body scattering model to SRA.From target S ₁Transition function vector to SRA can be written as h ₁=[h ₁₁, h ₁₂..., h _1N] ^T, wherein, h _1i(subscript i=1,2 ..., N) expression target S ₁Arrive the transition function of the i array element of SRA.Similarly, bottom scattering body one S ₂, bottom scattering body two S ₃With bottom scattering body Three S's ₄Can remember respectively to the transition function vector of SRA and to make h ₂, h ₃And h ₄Wherein, S ₂With target S ₁Be positioned at same distance, S ₃And S ₄Lay respectively at S ₂The both sides, front and back, with S ₂Spacing all be d=Tc ₀/ 2.

At detecting apart from r, the echoed signal that SRA receives can be write as following frequency domain form so:

Wherein, ω is the angular frequency of signal, and (s (ω) represents target echo signal to x for r, the ω) echoed signal apart from the r correspondence of expression SRA reception, and (r, the ω) reverb signal apart from the r correspondence of expression SRA reception is promptly by the bottom scattering body S on the target range for r ₂The reverb signal that scattering produces, n (ω) is a neighbourhood noise, can think additive white noise, and is independent with target echo and bottom reverberation.Shallow sea ray-normal mode the reverberation model that utilizes Ellis to propose can obtain certain bottom reverberation signal apart from correspondence, utilizes the scattering pressure and the travel-time (being equivalent to target echo signal) that can obtain target based on the KRAKEN model of normal mode.

2. obtain signal (target echo+bottom reverberation) subspace of target range correspondence

Inverse operator was decomposed signal (target echo+bottom reverberation) subspace that obtains the target range correspondence when this step will carry out the echoed signal after the previous step windowing process.

When on detecting, having target apart from r, this echoed signal x that receives apart from the SRA of correspondence (r ω) can be written as:

x(r，ω)＝s(ω)+r(r，ω)+n(ω) (3)

Wherein, ω is a frequency, and s (ω) is a target echo signal, and (r is by the bottom scattering body S on the target range ω) to r ₂The reverb signal that scattering produces, n (ω) is a neighbourhood noise, can think additive white noise, with s (ω) and r (r, ω) independence.Here mainly consider the reverberation background, ignore influence of environmental noise.Utilize the transition function vector, target echo s (ω) and bottom reverberation r (r ω) also can be written as:

$s\left(\mathrm{\&omega;}\right)=c_1{h}_1{h}_1^{T}e={\hat{c}}_1{h}_1---\left(4\right)$

$r(r,\mathrm{\&omega;})=c_2{h}_2{h}_2^{T}e={\hat{c}}_2{h}_2---\left(5\right)$

Wherein, e is the excitation weight vector of SRA, c ₁And c ₂Be respectively target S ₁With scatterer S ₂Dispersion factor.Get a plurality of different excitation e _i, can obtain a plurality of different received signal x _i(r, ω).Can be write as matrix form (slightly writing frequencies omega):

X＝H ^TCHE (6)

Wherein, subscript T represents transposition, X=[x ₁(r, ω), x ₂(r, ω) ..., x _M(r, ω)],, H=[h ₁, h ₂] ^T, C=diag (c ₁, c ₂), E=[e ₁, e ₂..., e _M] be excitation matrix, each column vector e _iRepresent a kind of excitation of SRA.When excitation matrix is N rank orthogonal matrix, (get standard quadrature Hadamard matrix here), utilize received signal matrix X can make up a covariance matrix R=XX ^H, then operator time reversal (TRO) can be write as:

TRO＝(XX ^H) ^*＝R ^* (7)

Can obtain by characteristic value decomposition:

TRO=U Λ U ^H+ V Ω V ^H(8) wherein, right formula two parts are corresponding respectively signal subspace and noise subspace, Λ is the diagonal matrix of a K * K, comprised K eigenwert greater than noise variance, Ω is the diagonal matrix of (N-K) * (N-K), comprised N-K eigenwert that equals noise variance, U is the matrix of a N * K, has comprised K proper vector (u of signal subspace _i, i=1 ...., K), V is the matrix of a N * (N-K), has comprised N-K proper vector (v of noise subspace _i, i=1 ...., N-K).Because comprised target echo and bottom reverberation in the received signal this moment, this signal subspace has comprised target echo subspace and reverberation subspace two parts simultaneously.Gou Zao rectangular projection operator may suppress bottom reverberation and target echo simultaneously thus.

3. utilize forward prediction to estimate the bottom reverberation subspace of target range correspondence

We have obtained signal (target echo+bottom reverberation) subspace of a target range correspondence to utilize step 2, owing to there is not other prior imformation, we can't therefrom extract a pure bottom reverberation subspace.This joint we will utilize the forward prediction method to estimate the bottom reverberation subspace of target range correspondence.

Generally speaking, can think that the echoed signal that the SRA of target echo previous moment receives is made of bottom reverberation fully:

x(r-d，ω)＝r(r-d，ω)+n(ω) (9)

Utilize the reverberation data x that repeatedly receives _j(r-d, ω) (bat of also expressing one's gratification, subscript j represents snap the j time) can construct a data covariance matrix

, it is carried out characteristic value decomposition can obtain:

$\hat{R}=\frac{1}{L}\underset{j=1}{\overset{L}{\mathrm{\&Sigma;}}}{x}_j(r-d,\mathrm{\&omega;})x_j^H(r-d,\mathrm{\&omega;})=\hat{U}\widehat{\mathrm{\&Lambda;}}{\hat{U}}^H+\hat{V}\widehat{\mathrm{\&Omega;}}{\hat{V}}^H---\left(10\right)$

Wherein, L is fast umber of beats, is taken as the array number of SRA here, L=N, matrix With Definition and formula (8) in Λ, Ω, U the same with V.At this moment, owing to utilized pure bottom reverberation signal, we have obtained a pure reverberation subspace

(proper vector is

, i=1 ...., K ').Because two signals are very approaching on time of reception, two reverb signals are steadily approximate, can think that the reverberation subspace and the reverberation subspace of target range correspondence of this moment are close, and reverberation subspace that previous moment obtains goes out the reverberation subspace of now as target echo a estimation is appearred in target echo.So, can obtain a rectangular projection operator of reverberation subspace:

$\hat{P}=I-\hat{U}{\hat{U}}^H=I-\underset{i=1}{\overset{K}{\mathrm{\&Sigma;}}}{\hat{u}}_i{\hat{u}}_i^H---\left(11\right)$

4. one of derivation SRA group of optimum weighing vector

A rectangular projection operator is constructed in the reverberation subspace that utilizes previous step to estimate

, utilize the echo that this rectangular projection operator receives SRA signal (wherein, mainly comprised target echo and bottom reverberation, x (r, ω)=s (ω)+r (r, ω)) carries out projection process $(\hat{P}x(r,\mathrm{\&omega;})=\hat{P}s\left(\mathrm{\&omega;}\right)+\hat{P}r(r,\mathrm{\&omega;})),$ The reverberation component that can suppress echoed signal to a certain extent $(\left|\hat{P}r(r,\mathrm{\&omega;})\right|<<\left|\hat{P}s\left(\mathrm{\&omega;}\right)\right|),$ Strengthen the echo letter and mix ratio.

As can be seen, this traditional rectangular projection disposal route has only been utilized the reverberation subspace information that previous moment is estimated

, and the signal subspace information U that does not utilize current time to obtain.Owing to comprised the transition function information of target in this signal subspace to SRA, utilize this information to improve: according to two signal subspaces that obtain constantly to this rectangular projection disposal route, seek one group of weight coefficient w of SRA, the echoed signal that SRA receives is carried out Filtering Processing (w ^TX (r, ω)) makes the reverberation of output signal be suppressed (w ^TR (r, ω)=0), and the output energy maximum (max|w of target echo ^TS (ω) | ²), mix ratio thereby further strengthen the echo letter.Can construct equation of constraint thus:

$\underset{w}{\max }\left(I\left(w\right)\right)=\underset{w}{\max }\left(w^H{s}^{*}\left(\mathrm{\&omega;}\right)s^T\left(\mathrm{\&omega;}\right)w\right)$ s.t.w ^Tr(r，ω)＝0，w ^Hw＝1 (12)

Since can't obtain pure target echo information s (ω) and bottom reverberation information r (r, ω), the Reverberation Rejection processing in the formula (12) can be written as ( $w^T{\hat{u}}_i=0,$ I=1 ..., K '), target echo s ^*(ω) can use ${\mathrm{<figure-callout\; id="0"\; label="u"\; filenames="H2008102318742E00021.png,H2008102318742E00022.png"\; state="\{\{state\}\}">u</figure-callout>}}_0={\mathrm{\&Sigma;}}_{i=1}^K{u}_i$ Replace.

According to the time inverse operator decomposition principle, can know u ₀=α s ^*(ω)+β r ^*(r, ω), so I (w) is launched and can obtain:

$I\left(w\right)=w^H{u}_0{u}_0^{H}w=w^H({\mathrm{\&alpha;s}}^{*}\left(\mathrm{\&omega;}\right)+{\mathrm{\&beta;r}}^{*}(r,\mathrm{\&omega;}))({\mathrm{\&alpha;s}}^{*}\left(\mathrm{\&omega;}\right)+{\mathrm{\&beta;r}}^{*}(r,\mathrm{\&omega;}){)}^{H}w$

$={\mathrm{\&alpha;}}^2{w}^H{s}^{*}\left(\mathrm{\&omega;}\right)s^T\left(\mathrm{\&omega;}\right)w+{\mathrm{\&alpha;\&beta;w}}^H{s}^{*}\left(\text{\&omega;}\right)r^T(r,\mathrm{\&omega;})w+{\mathrm{\&alpha;\&beta;w}}^H{r}^{*}(r,\mathrm{\&omega;})s^T\left(\mathrm{\&omega;}\right)w+$

$={\mathrm{\&alpha;}}^2{w}^H{s}^{*}\left(\mathrm{\&omega;}\right)s^T\left(\mathrm{\&omega;}\right)w+{\mathrm{\&alpha;\&beta;w}}^H{s}^{*}\left(\mathrm{\&omega;}\right)r^T(r,\mathrm{\&omega;})w+{\mathrm{\&alpha;\&beta;w}}^H{r}^{*}(r,\mathrm{\&omega;})s^T\left(\mathrm{\&omega;}\right)w+---\left(13\right)$

${\mathrm{\&beta;}}^2{w}^H{r}^{*}(r,\mathrm{\&omega;})r^T(r,\mathrm{\&omega;})w$

Because Reverberation Rejection is handled (w ^TR (r, ω) ≈ 0), back three that formula (13) is launched is zero substantially, the output energy mainly concentrates on first of expansion, the energy maximization of expression target echo.This shows, utilize this method also can guarantee target echo signal energy maximum simultaneously, mix ratio, improve target detection performance thereby strengthen the echo letter so that the reverb signal that SRA receives is suppressed.

Formula (12) can be written as so:

$\underset{w}{\max }\left(I\left(w\right)\right)=\underset{w}{\max }\left(w^H{u}_0{u}_0^{H}w\right)$ $s.t.w^T\hat{\text{U}}=0,$ w ^Hw＝1 (14)

Wherein, $\hat{U}=[{\hat{u}}_1,{\hat{u}}_2,\&CenterDot;\&CenterDot;\&CenterDot;,{\hat{u}}_{K^{\&prime;}}]\&CenterDot;$

Separate formula (14), can get the optimum weighing vector of SRA:

$w=\hat{P}u=\frac{\hat{P}{u}_0}{\sqrt{u_0^H\hat{P}{u}_0}}---\left(15\right)$

5. utilizing optimum weighing vector to carry out Reverberation Rejection handles

(r ω) is weighted summation and handles (w the echoed signal x that utilizes one group of formula (15) gained optimum weighing vector w that SRA is received ^TX (r, ω)) can suppress the reverberation component in the echoed signal, makes the target echo signal component maximum in the echoed signal simultaneously, mixes ratio thereby further strengthen the echo letter, improves target detection performance.

Because target range the unknown, we can't know that target echo goes out now, this method can be handled the received signal of each period successively by sliding window, with the return signal of a certain period to start with, utilize step 2 that it is handled and obtain a reverberation subspace, with of the prediction (estimation) of this reverberation subspace as the reverberation subspace of next one moment received signal, thereby suppress the reverberation component in the next received signal constantly, successively each return signal is constantly carried out the inhibition that same processing can be finished all reverb signals in the whole time of reception, strengthen the echo letter and mix ratio.

The invention has the beneficial effects as follows: ultimate principle of the present invention and specific implementation have obtained the checking of simulated experiment in the typical shallow sea waveguide environment, and the result shows:

1. the present invention reverberation subspace that previous moment obtains goes out the reverberation subspace of now as target echo a estimation occurs with target echo.Though going out the real reverberation subspace of now, this estimation reverberation subspace and target echo exist certain difference certainly, the thus obtained Reverberation Rejection disposal route (this method can't realize in practice) that is not so good as to go out the reverberation subspace of now based on the Reverberation Rejection ability of forward prediction method based on target echo, but with handle before compare, the present invention still is effectively, and the letter that can obtain nearly 60dB under simulated conditions mixes than gain;

2. the present invention has adopted passive processing mode, directly the echo to each SRA reception constantly carries out the Reverberation Rejection processing, and anti-letter mixes than need existing each the Enhancement Method that the distance of target is last carries out the processing of active send-receive successively when not needing as tradition initiatively, thereby this method has stronger practicality;

3. owing to utilized the reverberation subspace of forward prediction, the present invention also solved current when passive dereverberation inhibition method may make target echo and reverberation repressed problem simultaneously, can suppress reverberation effectively, keep simultaneously or the enhancing target echo, mix ratio thereby strengthen the echo letter effectively, improve the target detection performance of active sonar;

4. as long as utilize sub-band division, and a plurality of subbands are similarly handled, the present invention just can be applied to the inhibition of broadband signal reverb signal and handle.

The present invention is further described below in conjunction with drawings and Examples.

Description of drawings

Fig. 1 is the mode synoptic diagram between target of the present invention, bottom scattering body and the SRA.

Among the figure, anti-array when SRA represents is a vertical battle array of being made up of N array element, 1, and .., j ..., N represents each number array element, S ₁The expression target, S ₂Expression bottom scattering body 1, S ₃Expression bottom scattering body 2, S ₄Expression bottom scattering body 3, h _{1, j}Expression target S ₁Arrive the transition function of the j array element of SRA, r represents target S ₁To the horizontal range of SRA, bottom scattering body S ₂With target S ₁Be positioned at same distance, bottom scattering body S ₃And S ₄Lay respectively at S ₂The both sides, front and back, with S ₂Spacing all be d.

Fig. 2 is a simulated environment synoptic diagram of the present invention.

Among the figure, z ₀Expression ocean water laminar surface, z ₁Expression sedimentary deposit surface, z ₂Expression subsea level surface.The degree of depth of water layer is 120m, curve c ₁The velocity of sound section of expression water layer; The degree of depth of sedimentary deposit is 2.5m, curve c ₂Expression deposition velocity of sound section layer by layer, c ₃The velocity of sound of representing even seabed semispace.Anti-array when SRA represents is that an array number N=32, array element distance are the vertical battle array of 3m.S ₁A target that exists in the expression waveguide, r represents target S ₁With the horizontal range of SRA, z _sExpression target S ₁The degree of depth.ρ represents Media density, and α represents the acoustical absorptivity of medium.

Fig. 3 (a) is the average attenuation curve of the bottom reverberation echoed signal of SRA reception; Fig. 3 (b) is the average attenuation curve that the bottom reverberation of SRA reception adds target echo signal.

The average attenuation curve of the bottom reverberation signal after dereverberation suppressed to handle when Fig. 4 (a) was passive; Bottom reverberation after dereverberation suppressed to handle when Fig. 4 (b) was passive adds the average attenuation curve of target echo signal.

Fig. 5 (a) utilizes traditional rectangular projection method to handle the average attenuation curve of the signal that obtains; Fig. 5 (b) utilizes method that the present invention provides to handle the average attenuation curve of the signal that obtains.

Fig. 6 is under the condition of different distance interval, the performance evaluation curve map that the different disposal method is handled.

Fig. 7 (a) is the average attenuation curve of the target depth signal that utilizes the present invention to handle during for 90m to obtain; Fig. 7 (b) is the average attenuation curve of the target depth signal that utilizes the present invention to handle during for 114m to obtain.

Fig. 8 is when having a plurality of target in the waveguide, the average attenuation curve of the signal that utilizes the present invention to handle to obtain.

Embodiment

With typical shallow sea waveguide environment is example, has provided embodiment of the present invention.

The waveguide environment that adopts is the irrelevant layering shallow sea waveguide environment of typical range, as shown in Figure 2.Form by water layer, sedimentary deposit and subsea level.Z among the figure ₀The expression seawater surface, z ₁Expression sedimentary deposit surface, z ₂Expression subsea level surface.The degree of depth of water layer is 120m, curve c among velocity of sound section such as the figure ₁Shown in; The degree of depth of sedimentary deposit is 2.5m, curve c among velocity of sound section such as the figure ₂Shown in, the seabed is a homogeneous half space, velocity of sound c ₃=1800m/s.SRA is the vertical battle array that an array number is 32, array element distance is 3m.Environment is constant in the process of the test.L-G simulation test has only been considered bottom reverberation, ignores the reverberation that influences of sea and other scattering.Sound field is calculated the KRAKEN model that has adopted based on normal mode.Signal adopts the simple signal of 500Hz, and pulsewidth is 70ms.

At this typical case shallow sea waveguide environment, the key step that the present invention realizes is as follows:

1. bottom reverberation and target echo modeling and windowing process

The present invention need handle the echoed signal that SRA receives, and wherein the echoed signal of Jie Shouing comprises bottom reverberation and target echo two parts.Bottom reverberation exists always, and target echo only appears at its travel-time and engraves when corresponding.Shallow sea ray-normal mode the reverberation model that utilizes Ellis to propose can obtain certain bottom reverberation signal constantly, utilizes KRAKEN model based on normal mode can obtain the bottom scattering body on the respective distance and the scattering pressure (being equivalent to target echo signal and travel-time thereof) of target.So, the echoed signal that SRA receives in the time of before target echo does not occur is exactly the seabed bottom reverberation, and the echoed signal when target echo occurs is emulation to be obtained the bottom reverberation signal plus in the target echo signal and the corresponding moment obtain.Can realize bottom reverberation and target echo modeling and windowing process by this processing.Fig. 3 (a) has provided the die-away curve of the average reverberation of SRA reception when not having target in the waveguide with distance (time).Fig. 3 (b) has provided the average attenuation curve of SRA reception echoed signal when having target in the waveguide.As can be seen, target echo signal has appearred in 3.7s constantly, but is submerged in basically in the bottom reverberation, and letter mixes lower.

2. obtain signal (target echo+bottom reverberation) subspace of target range correspondence

At first the echoed signal that SRA is received is carried out segmentation.Because bottom reverberation signal is sometime produced by the scatterer scattering on contributive zone on the respective distance (d=c τ/2), so, in Simulating Test Study, can carry out segmentation to the seabed metric space, be taken as c τ/2 ≈ 50m at interval.At target range, according to formula (6), (7) and formula (8), the echoed signal of corresponding period is carried out characteristic value decomposition, (proper vector is u can to obtain a signal subspace U _i, i=1 ...., K).Because comprised target echo and bottom reverberation in the received signal this moment, this signal subspace has comprised target echo subspace and reverberation subspace two parts simultaneously.Gou Zao rectangular projection operator may suppress bottom reverberation and target echo simultaneously thus, as shown in Figure 4.

3. utilize forward prediction to estimate the bottom reverberation subspace of target range correspondence

Utilize formula (10), the echoed signal (being made of pure bottom reverberation) of previous moment is appearred in target echo, carry out characteristic value decomposition and can obtain a reverberation subspace

(proper vector is I=1 ...., K '), and it is gone out a prediction (estimation) of the reverberation subspace of now as target echo.According to formula (11), utilize the reverberation subspace of this estimation Can construct a rectangular projection operator , the echoed signal x (ω) that utilizes this projection operator that SRA is received carry out projection process ( ), can suppress the bottom reverberation component in the echoed signal, the basic simultaneously component of signal that keeps in the echoed signal is mixed ratio thereby strengthen the echo letter, shown in Fig. 5 (a).

4. one of derivation SRA group of optimum weighing vector

Two different signal subspace U and reverberation subspace information that obtain constantly according to top step (2) and step (3) acquisition

Utilize formula (15) can derive one group of optimal weight vector w of SRA.The echoed signal of utilizing this optimum weighing vector that SRA is received is weighted summation and handles, can be so that the reverberation of output signal be suppressed, and the output energy maximum of target echo, thereby improve traditional rectangular projection disposal route.

5. utilizing optimum weighing vector to carry out Reverberation Rejection handles

Utilize one group of optimum weighing vector w of the SRA of formula (15) derivation that SRA reception echoed signal x (ω) is weighted summation (w ^TX (ω)) handles.Fig. 5 (b) has provided and has utilized this optimal weighting coefficients to be weighted summation processing (w ^TX (ω)) the average attenuation curve of the signal that obtains.Comparison diagram 5 (a) and (b), directly perceived on as can be seen, two kinds of target-echo intensity basically identicals that method produces, and the weighted filtering disposal route has lower bottom reverberation level, this is because weight coefficient normalized (w in the formula (14) ^HW=1) cause that in fact, the Reverberation Rejection ability of two kinds of methods is the same, and the weighted filtering disposal route can obtain stronger target echo.

The present invention reverberation subspace that previous moment obtains goes out the reverberation subspace of now as target echo a estimation occurs with target echo.The real reverberation subspace that reverberation subspace that is obtained by forward prediction and target echo go out now exists certain difference certainly, and this species diversity can along with two reverb signal time of receptions at interval (corresponding seabed apart from interval) increase and increase.Fig. 6 has provided the letter that different distance utilizes the present invention to handle under the condition to obtain at interval and has mixed ratio.Generally speaking, we get the two neighboring sections signal and handle, time of reception is very near (time of reception equals the pulsewidth that transmits at interval), though be not so good as to go out the Reverberation Rejection disposal route of the reverberation subspace of now this moment based on the Reverberation Rejection ability of forward prediction method based on target echo, but compare with (seeing that Fig. 6 deceives solid line) before the processing, the present invention still is effectively, and the letter that can obtain nearly 60dB under simulated conditions mixes than gain.

When Fig. 7 (a) and (b) have provided target depth respectively and have been 90m and 114m, the average attenuation curve of the signal that utilizes the present invention to handle to obtain.As can be seen, during the gtoal setting seabed, because the correlativity between the bottom reverberation that seabed scatterer scattering produces on target echo that SRA receives and the respective distance strengthens, the present invention can suppress target echo to a certain extent, but still is effective.Consider to exist on a plurality of distances in the waveguide situation of target.Simulated environment is constant, and the distance of three targets is respectively 2.8km, 3.5km and 4.0km, and the degree of depth is respectively 60m, 30m and 90m.

Fig. 8 provided this during based on forward prediction passive dereverberation inhibition method handle the average attenuation curve of the signal that produces.As can be seen, utilize the present invention to suppress bottom reverberation effectively, the echo letter that strengthens each target mixes ratio, improves the target detection ability.

Claims (1)

1. the passive time reversal reverberation suppression method based on forward prediction is characterized in that comprising the steps:

(a) modeling of bottom reverberation and target echo and windowing process, the echoed signal that time reversal, array SRA received can be write as following frequency domain form

Wherein, ω is the angular frequency of signal, and (s (ω) represents target echo signal to x for r, the ω) echoed signal apart from the r correspondence of expression SRA reception, and (n (ω) is a neighbourhood noise to r for r, the ω) reverb signal apart from the r correspondence of expression SRA reception; Utilize shallow sea ray-normal mode reverberation model to obtain certain bottom reverberation signal, utilize the scattering pressure and the travel-time that obtain target based on the KRAKEN model of normal mode apart from correspondence;

(b) signal subspace of acquisition target range correspondence, when on detecting, having target apart from r, this echoed signal x that receives apart from the SRA of correspondence (r, ω)=s (ω)+r (r, ω)+n (ω),

Wherein, e is the excitation weight vector of SRA, c ₁And c ₂Be respectively target S ₁With scatterer S ₂Dispersion factor, h ₁Be from target S ₁To the transition function vector of SRA, h ₂Be bottom scattering body one S ₂Transition function vector to SRA; Get a plurality of different excitation e _i, can obtain a plurality of different received signal x _i(r, ω); Can be write as matrix form X=H ^TCHE, wherein, subscript T represents transposition, X=[x ₁(r, ω), x ₂(r, ω) ..., x _M(r, ω)], H=[h ₁, h ₂] ^T, C=diag (c ₁, c ₂), E=[e ₁, e ₂..., e _M] be excitation matrix, each column vector e _iRepresent a kind of excitation of SRA; When excitation matrix is N rank orthogonal matrix, utilize received signal matrix X can make up a covariance matrix R=XX ^H, then time reversal, operator TRO can obtain TRO=U Λ U by characteristic value decomposition ^H+ V Ω V ^HWherein, right formula two parts are corresponding respectively signal subspace and noise subspace, Λ is the diagonal matrix of a K * K, comprised K the eigenwert greater than noise variance, Ω is the diagonal matrix of (N-K) * (N-K), has comprised N-K eigenwert that equals noise variance, U is the matrix of a N * K, has comprised K proper vector (u of signal subspace _i, i=1 ...., K), V is the matrix of a N * (N-K), has comprised N-K proper vector (v of noise subspace _i, i=1 ...., N-K);

(c) utilize forward prediction to estimate the bottom reverberation subspace of target range correspondence, think echoed signal that the SRA of target echo previous moment receives fully by bottom reverberation constitute x (r-d, ω)=r (r-d, ω)+n (ω); Utilize the reverberation data x that repeatedly receives _j(r-d ω) constructs a data covariance matrix

, it is carried out characteristic value decomposition can obtain

Wherein, L is fast umber of beats, is taken as the array number of SRA here, has obtained a pure reverberation subspace thus

A rectangular projection operator with the reverberation subspace

(d) one of derivation SRA group of optimum weighing vector

u ₀=α _s* (ω)+β r* (r, ω);

(e) (r ω) is weighted summation and handles the echoed signal x that utilizes optimum weighing vector w that SRA is received, and can suppress the reverberation component in the echoed signal, makes the target echo signal component maximum in the echoed signal simultaneously, mixes ratio thereby further strengthen the echo letter.

Images