CN105490974A - Doppler estimation method of MIMO-OFDM hydroacoustic communication system - Google Patents

Abstract

The invention discloses a Doppler estimation method of a MIMO-OFDM hydroacoustic communication system. The method comprises the following steps: sampling sparse signals received at a receiving end at a sampling interval to obtain a sparse vector y; through synchronization pilot signals s(t) already known to both two receiving sides and channel features, selecting proper precision to construct an over-complete dictionary; in one period, performing accumulation on all sampling values to obtain y=Cx+n, wherein a receiving vector is seen as a linear combination of a C series taking an x element as a coefficient; receiving the sparse vector y through the over-complete dictionary C, and reconstructing signals x by use of an SBL algorithm; and finding out a corresponding column in the over-complete dictionary C corresponding to a non-zero element in the x, and accordingly finding out an equivalent Doppler factor corresponding to each path. According to the invention, the estimation precision of a MIMO-OFDM hydroacoustic Doppler factor is greatly improved, and a new solution method is also brought forward for routine Doppler spread.

Description

A kind of Doppler estimation of MIMO-OFDM underwater sound communication system

Technical field

The invention belongs to technical field of underwater acoustic communication, especially a kind of Doppler estimation of MIMO-OFDM underwater sound communication system.

Background technology

Underwater sound communication network has important using value in the Long-distance Control of Offshore Oil Industry, marine environmental monitoring, Marine Sciences Data Collection, under water call, exploration of ocean resources etc.It is made up of submarine sensor, Autonomous Underwater Vehicle and surface station usually.All difficulties such as the characteristic of ocean transportation channel is as limited in available band, propagation delay time is long and ambient noise level is high seriously limit the service quality of underwater sound communication network.Therefore, the realization of high-performance, high reliability underwater sound communication network has great technological challenge, and its key is the communication network technology studied based on ocean channel condition and underwater sound communication network feature.Become a study hotspot in subsurface communication field with its high data rate and high reliability based on the underwater sound communication system of OFDM (MIMO-OFDM) technology of multiple-input and multiple-output.And the frequency displacement that the frequency selective fading that in underwater acoustic channel, Multipath Transmission causes and intersymbol interference and Doppler effect cause is the principal element affecting underwater sound communication system speed and data transmission credibility.The underwater sound communication system Doppler estimation that Doppler estimation in MIMO-OFDM system exports than single input is much complicated.

Mostly Doppler estimation in current underwater sound communication is the method based on ambiguity function, and its concrete methods of realizing is roughly divided into two classes:

First kind method mostly depends on one or more to the insensitive synchronizing signal of doppler spread (as linear FM signal LFM, pseudo random sequence PN etc.), one group of correlator (cross-correlation or auto-correlation) is set, adopts the relevant theories such as relevant, the ambiguity function in signal transacting to estimate doppler spread.

Equations of The Second Kind method is the feature according to underwater acoustic channel, adopts up-to-date signal processing theory, and Doppler shift problem is converted into other problems research.Mainly contain: Doppler's estimation problem is considered as a part for underwater sound signal Parameter Estimation Problem by (1); (2) Doppler shift is looked as a whole, be modeled as Sparse Signal Representation or other signal processing problems, find suitable signal processing algorithm to study.

These Doppler factor algorithm for estimating all can expand in mimo system, but now need the impact considering that the otherness of multiple receipts channel variation is estimated Doppler factor, notice the sparsity structure of underwater acoustic channel simultaneously, namely only have the stronger subchannel of small part energy to transfer signals in underwater sound mimo channel and reach receiving terminal by hydrophone reception process.For MIMO-OFDM system, the present invention is by designing the synchronous code of joint pilot and making full use of the feature of synchronization code signal, Doppler factor in MIMO-OFDM underwater sound communication system is accurately estimated, improve the deficiency of traditional single-input single-output Doppler factor method of estimation in multi-input multi-output system application, substantially increase estimated accuracy.

Summary of the invention

Goal of the invention: the object of the present invention is to provide a kind of underwater sound MIMO-OFDM Doppler factor precise Estimation Method, for the feature of designed synchronizing pilot, a kind of new signal processing technology is introduced underwater sound communication and carries out Doppler's estimation, the method of the accurate estimating Doppler factor of a kind of energy is proposed, solve traditional correlation technique estimated result to be forbidden, do not consider the shortcoming of the sparse characteristic of MIMO-OFDM communication system transmission.

Technical scheme: a kind of Doppler estimation of MIMO-OFDM underwater sound communication system, comprises the steps:

S1: the sparse signal that receiving terminal receives was sampled with the sampling interval, obtains sparse vector y;

S2: by receiving all known synchronizing pilot signal s (t) of both sides in conjunction with channel characteristic, choosing suitable accuracy, constructing super complete dictionary;

S3: accumulate all sample values in one-period: y=Cx+n, receives vector and regards the linear combination respectively arranged with the x element C that is coefficient as;

S4: by super complete dictionary C, receive sparse vector y, utilize SBL algorithm, reconstruction signal x;

S5: find the respective column in super complete dictionary C corresponding to nonzero element in x, then find the Doppler factor of equal value corresponding to each footpath.

Described step S1 is specially: considering one has N _tindividual transducer, N _rthe underwater sound MIMO-OFDM communication system of individual hydrophone, suppose in the MIMO multipath channel between i-th transducer and a jth hydrophone, in a symbol period, multipath fading is approximate constant, and each transducer is different to the equivalent Doppler factor of hydrophone; Definition Doppler factor ν _p, time delay factor τ _pparameter sets:

ν _p∈{ν ₁,ν ₂,...,ν _l,...,ν _L}τ _p∈{τ ₁,τ ₂,...,τ _k,...,τ _K}

Wherein ν _l=ν ₁+ (l-1) Δ ν, τ _k=τ ₁+ (k-1) Δ τ, Δ ν, Δ τ represent the discrete precision that Doppler and time delay are estimated, can adjust flexibly as required, L, K represent Doppler factor, time delay value number that mimo channel is possible respectively; Assuming that transmitting terminal synchronizing pilot signal is s (t), after the sparse time varying channel of MIMO, a jth hydrophone Received signal strength is expressed as y _j(t).

Described step S2 is specially: set the duration of s (t) as [0, T], at receiving terminal to y _jt () is sampled with interval of delta t, make τ in described S1 step ₁=0, sampling interval Δ t=Δ τ, and omit the Δ t in above formula, abbreviation is also write as vector form and can be obtained y [i]=c [i] ^tx+n [i].

Described step S3 is specially: accumulate all sample values in one-period: y=Cx+n, the vector form of Received signal strength shown in considering, receives vector and regards the linear combination respectively arranged with the x element C that is coefficient as.

Described step S4 is specially: adopt the MIMO-OFDM underwater sound Doppler estimation based on management loading, from the sparse characteristic of underwater sound mimo channel, based on super complete dictionary, MIMO Doppler estimation problem is modeled as rarefaction representation problem, Bayesian learning algorithm is utilized to solve, under the prerequisite of known y and C, SBL algorithm reconstructed amplitude fading matrix x at that time with self-adaptative adjustment.

Described step S5 is specially: the sparse matrix x reconstructed according to described step S4, finds corresponding element in the corresponding C matrix of its non-vanishing vector, then finds the Doppler factor of each corresponding mimo channel.

In described step S4, the specific implementation step of SBL algorithm is:

The first step, input: reception vector y, I × (L × K) of I dimension ties up super complete dictionary Matrix C, I=T/ Δ t

Second step, initialization: hyper parameter γ ⁽⁰⁾>=0, β ⁽⁰⁾> 0, number of iterations k=1, ε > 0;

3rd step, algorithm iteration:

A. remember Λ=diag (1/ γ), utilize

$p\left(x\right|y,\mathrm{\γ},\mathrm{\β})={\left(2\mathrm{\π}\right)}^{-\frac{N}{2}}{\left|\mathrm{\Σ}\right|}^{-\frac{1}{2}}\exp (-\frac{1}{2}{\left(x-\mathrm{\μ}\right)}^T{\mathrm{\Σ}}^{-1}(x-\mathrm{\μ}\left)\right)$

Σ＝(βC ^TC+Λ) ^-1

Estimate μ respectively ^(k-1), Σ ^(k-1);

B. A=β is remembered ^-1i+C Λ ^-1c ^t, utilize

L(γ|β)＝log|A|+y ^TA ^-1y

${\mathrm{\γ}}_i^{(k+1)}={\mathrm{\μ}}_i^2+{\mathrm{\Σ}}_{ii}$

Undated parameter γ, β, obtain γ respectively ^(k), β ^(k);

C. judge whether to meet || γ ^(k)-γ ^(k+1)||≤ε, if meet, goes to the 4th step; Discontented then k=k+1, repeats the 3rd step 1;

4th step, exports the estimated value of sparse vector:

Beneficial effect: the present invention, on the basis of traditional OFDM frame structure, devises the OFDM symbol structure of a kind of new pilot tone and synchronous multiplexing.By the amplitude of linear FM signal and phase place through certain setting, time domain is used for sign synchronization and Doppler and estimates, also can be used as pilot tone at frequency domain and carry out channel estimating, linear FM signal and the same length of OFDM symbol, same frequency band are set, after fast Fourier transform (FFT), in its frequency band, the frequency spectrum of each frequency is also corresponding with ofdm signal sub-carrier positions, makes frequency pilot sign.When carrying out fast Fourier transform inverse transformation, pilot tone can revert to again the linear FM signal of time domain, can be used for carrying out Doppler's estimation herein and sign synchronization when receiving.Achieve the multiplexing of signal, also improve the communication efficiency of underwater sound MIMO-OFDM communication system to a certain extent.

Compared with prior art, the invention has the beneficial effects as follows: (1) devises a kind of reasonably MIMO-OFDM frame structure, synchronizing pilot wherein is accurately estimated except carrying out Doppler factor, also can be used to realize Timing Synchronization and channel estimating, improves reusing degree; (2) make full use of the intellectual of the intrinsic signal characteristic of synchronizing pilot and receiving terminal, utilize the openness of MIMO underwater acoustic channel to carry out Doppler's estimation, reduce it and realize complexity, improve estimated accuracy; (3) the present invention not only substantially increases the estimated accuracy of MIMO-OFDM underwater sound Doppler factor, and it is also proposed new solution for conventional Doppler expansion.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of the inventive method;

Fig. 2 is the OFDM symbol structure chart of pilot tone of the present invention and synchronous multiplexing;

Fig. 3 is underwater sound MIMO-OFDM system block diagram of the present invention;

The SBL algorithm flow chart that Fig. 4 designs for the present invention.

Embodiment

Below in conjunction with accompanying drawing, case study on implementation of the present invention is described in detail;

For ease of understanding general principle of the present invention and concrete grammar, first provide system model of the present invention.

1.MIMO underwater acoustic channel model

The present invention is based on following underwater acoustic channel model.Consider one and have N _tindividual transducer, N _rthe underwater sound MIMO-OFDM communication system of individual hydrophone, suppose in the underwater sound condition of sparse channel between i-th transducer and a jth hydrophone, in a symbol period, multipath fading is approximate constant, suppose that the signal transmission path that each transducer sends experiences different amplitude fadings, doppler spread and delay spread successively, the characterising parameter of its correspondence is respectively amplitude factor α _p, Doppler factor ν ' _pwith time delay factor τ ' _p.ν ' _p=v/c represents Doppler factor, i.e. the signal parameter that will estimate of the present invention, and v represents the diametrically movement velocity between transceiver, and c represents the propagation velocity of sound wave in water (about 1500m/s).

Underwater acoustic channel impulse response then through discrete sampling can be expressed as:

$h_{ij}(\mathrm{\τ},t)=\underset{p=1}{\overset{P}{\Σ}}{a}_p^{\left(ij\right)}\mathrm{\δ}(\mathrm{\τ}-({{\mathrm{\τ}}^{\′}}_p^{\left(ij\right)}-{{\mathrm{\ν}}^{\′}}_p^{\left(ij\right)}t\left)\right)---\left(1\right)$

i＝1,2,…,N _t,j＝1,2,…,N _r

Because a frame data usual duration is shorter, period channel variation slow, and doppler spread is mainly because the relative motion between transceiver causes.Because underwater acoustic channel has openness, so be not the signal that each hydrophone just can receive each transducer and launches.The present invention put forward MIMO-OFDM method and suppose that each transducer is different to the equivalent Doppler factor of hydrophone.Therefore, assuming that the synchronizing pilot signal that transmitting terminal sends is s (t), after above-mentioned time varying channel, a jth hydrophone receives i-th transducer signal and can be expressed as

$\begin{array}{c}{y}_{ij}\left(t\right)=s\left(t\right)*h_{ij}(\mathrm{\τ},t)+n_{ij}\left(t\right)\\ ={\∫}_{-\mathrm{\∞}}^{+\mathrm{\∞}}{h}_{ij}(\mathrm{\τ},t)s(t-\mathrm{\τ})d\mathrm{\τ}+n_{ij}\left(t\right)\end{array}---\left(2\right)$

Namely within a frame data duration, in multipath channel, the fading factor of every paths is consistent, and the Doppler factor of every paths is identical.

In formula, n (t) is white Gaussian noise vector, because underwater acoustic channel has openness, so be not the signal that each hydrophone just can receive each transducer and launches.Due to the present invention put forward MIMO-OFDM method and suppose that each transducer is all different to the equivalent Doppler factor of hydrophone, the ν be defined as follows _p, τ _pparameter sets:

ν _p∈{ν ₁,ν ₂,...,ν _l,...,ν _L}(3)

τ _p∈{τ ₁,τ ₂,...,τ _k,...,τ _K}(4)

Wherein ν _l=ν ₁+ (l-1) Δ ν, τ _k=τ ₁+ (k-1) Δ τ, Δ ν, Δ τ represent the discrete precision that Doppler and time delay are estimated, can adjust flexibly as required, L, K represent Doppler factor, the time delay value number that underwater sound condition of sparse channel is possible respectively.Then a jth hydrophone receives all transducer signals and can utilize formula (2) abbreviation to be:

$y_j\left(t\right)\≈\underset{k=1}{\overset{K}{\Σ}}\underset{l=1}{\overset{L}{\Σ}}{\mathrm{\α}}_{kl}s\[{\mathrm{\ν}}_l(t-{\mathrm{\τ}}_k)\]+n_j\left(t\right)---\left(5\right)$

If the duration of s (t) is [0, T], sample to y (t) at receiving terminal with interval of delta t, m sample value is:

$y_j\left(m\mathrm{\Δ}t\right)=\underset{k=1}{\overset{K}{\Σ}}\underset{l=1}{\overset{L}{\Σ}}{\mathrm{\α}}_{lk}s\[{\mathrm{\ν}}_l(t-{\mathrm{\τ}}_k)\]+n_j\left(m\mathrm{\Δ}t\right)---\left(6\right)$

Make τ in (4) ₁=0, sampling interval Δ t=Δ τ, and omit the Δ t in above formula, obtaining its reduced form is:

$y_j\left(m\right)=\underset{k=1}{\overset{K}{\Σ}}\underset{l=1}{\overset{L}{\Σ}}{\mathrm{\α}}_{lk}s\[{\mathrm{\ν}}_l(i-k+1)\mathrm{\Δ}\mathrm{\τ}\]+n_j\[m\]---\left(7\right)$

Write above formula as vector form:

y _j[m]＝c[m] ^Tx+n _j[m](8)

Wherein

x＝[α _1,1,...,α _1,K,α _2,1,...,α _2,K,......,α _L,1,...,α _L,K] ^T(9)

c[m]＝[ξ[1],ξ[2],...,ξ[L]] ^T(10)

ξ[l]＝[s _l,1(m),s _l,2(m),...,s _l,K(m)]

s _l,k(m)＝s[ν _l(m-k+1)Δτ](11)

By all sample value accumulations of receiving terminal in [0, T], obtain the matrix notation of (8) formula:

y _j＝Cx+n _j(12)

2. principle is derived

Y _jfor the reception vector of I dimension, C is that I × (L × K) ties up matrix, and be referred to as super complete dictionary, be known for receiving terminal, and be the white Gaussian noise vector of I dimension, I=T/ Δ t is the number of samples of Received signal strength.

The vector form of Received signal strength shown in consideration formula (12), receives vector and can regard the linear combination respectively arranged with the x element C that is coefficient as.Analysis mode (10), (11) find, s in each row of C _l,km () has identical subscript (l, k), certain possible Doppler and time delay ν in lucky corresponding (9) _l, τ _k, and x is the corresponding amplitude fading factor.At known y _jwith under the prerequisite of C, if x can be obtained from formula (11), then and then can equivalent Doppler factor corresponding to this mimo channel and time delay, thus reach the object that MIMO-OFDM underwater sound Doppler estimates.

3. management loading is theoretical

Suppose that observation noise is independently Gaussian process, and average is 0, variance is σ ², then

$p\left(y\right|x,{\mathrm{\σ}}^2)={\left(2{\mathrm{\π\σ}}^2\right)}^{-\frac{M}{2}}\exp (-\frac{1}{2{\mathrm{\σ}}^2}||y-Cx|{|}_2^2)---\left(13\right)$

Suppose sparse signal x again _iobeying average is 0, and variance is γ _igaussian Profile, then have

$p\left(x\right|\mathrm{\γ})=\underset{i=1}{\overset{N}{\Π}}{\left(2{\mathrm{\π\γ}}_i\right)}^{-\frac{1}{2}}\exp (-\frac{x_i^2}{2{\mathrm{\γ}}_i})---\left(14\right)$

Make β=σ ^-2, can obtain posterior probability p (x|y, γ, β) is

$p\left(x\right|y,\mathrm{\γ},\mathrm{\β})={\left(2\mathrm{\π}\right)}^{-\frac{N}{2}}{\left|\mathrm{\Σ}\right|}^{-\frac{1}{2}}\exp (-\frac{1}{2}{\left(x-\mathrm{\μ}\right)}^T{\mathrm{\Σ}}^{-1}(x-\mathrm{\μ}\left)\right)---\left(15\right)$

Easily know that posterior probability p (x|y, γ, β) meets Gaussian Profile, average μ and covariance matrix Σ is respectively

Σ＝(βC ^TC+Λ) ^-1

(16)

μ＝βΣC ^Ty

(17)

Wherein Λ=diag (1/ γ).As long as estimate hyper parameter γ, β, the estimated value of sparse vector just can be obtained for the estimation of hyper parameter γ, β, then obtain by maximizing marginal probability p (y| γ, β), wherein p (y| γ, β) is expressed as

$p\left(y\right|\mathrm{\γ},\mathrm{\β})={\left(2\mathrm{\π}\right)}^{-\frac{N}{2}}{\left|A\right|}^{-\frac{1}{2}}\exp (-\frac{1}{2}{\left(x-\mathrm{\μ}\right)}^T{A}^{-1}(x-\mathrm{\μ}\left)\right)---\left(18\right)$

Wherein A=β ^-1i+C Λ ^-1c ^t.

Get-logp (y| γ, β) and remove constant term, then maximization marginal probability p (y| γ, β) is equivalent to and minimizes cost function

L(γ|β)＝log|A|+y ^TA ^-1y

(19)

By EM algorithm, can construct undated parameter γ, the Iteration of β is respectively

${\mathrm{\γ}}_i^{(k+1)}={\mathrm{\μ}}_i^2+{\mathrm{\Σ}}_{ii}---\left(20\right)$

${\mathrm{\β}}^{(k+1)}=\frac{N}{\left|\right|y-Au|{|}_2^2+{\left({\mathrm{\β}}^{-1}\right)}^{(k+1)}\underset{i=1}{\overset{N}{\Σ}}\[1-{\left({\mathrm{\γ}}_i^{\left(k\right)}\right)}^{-1}{\mathrm{\Σ}}_{ii}\]}---\left(21\right)$

Judge whether to meet || γ ^(k)-γ ^(k+1)||≤ε, then continuing iteration if do not met, then can calculate as met $\hat{x}={\mathrm{\μ}}^{(k-1)}.$

4. the MIMO-OFDM underwater sound Doppler based on SBL algorithm estimates performing step

4-1, by receiving known synchronizing pilot signal s (t) of both sides, in conjunction with channel characteristic, choosing suitable accuracy, constructing super complete dictionary C.

4-2, selects suitable sampling period Δ t to the Received signal strength y on a jth hydrophone _jt () is sampled, obtain receiving sparse signal vector y _j.

4-3, by dictionary C, receives vectorial y _j, utilize SBL algorithm, reconstruct x.

4-4, finds the respective column in dictionary C corresponding to nonzero element in x, then finds the Doppler factor of equal value corresponding to each footpath.

The Doppler factor method of estimation of a kind of MIMO-OFDM underwater sound communication system based on management loading theory of the present invention.For the feature of designed synchronizing pilot, a kind of new sparse signal treatment technology is introduced MIMO-OFDM underwater sound communication system and carries out Doppler's estimation, the method of the accurate estimating Doppler factor of a kind of energy is proposed, solve traditional correlation technique estimated result to be forbidden, do not consider the shortcoming of the sparse characteristic of MIMO-OFDM communication system transmission.The method is made up of three parts, the frame structure design of underwater sound MIMO-OFDM signal, the rarefaction representation of signal and sparse Bayesian algorithm (SBL), utilize this feature openness of underwater acoustic channel to complete the estimation to MIMO-OFDM underwater sound Doppler in conjunction with management loading theory.The precision that mimo channel Doppler estimates is substantially increased while simplifying Doppler's estimating step.

Claims (7)

1. a Doppler estimation for MIMO-OFDM underwater sound communication system, is characterized in that, comprises the steps:

S1: the sparse signal that receiving terminal receives was sampled with the sampling interval, obtains sparse vector y;

S2: by receiving all known synchronizing pilot signal s (t) of both sides in conjunction with channel characteristic, choosing suitable accuracy, constructing super complete dictionary;

S3: accumulate all sample values in one-period: y=Cx+n, receives vector and regards the linear combination respectively arranged with the x element C that is coefficient as;

S4: by super complete dictionary C, receive sparse vector y, utilize SBL algorithm, reconstruction signal x;

S5: find the respective column in super complete dictionary C corresponding to nonzero element in x, then find the Doppler factor of equal value corresponding to each footpath.

2. the Doppler estimation of MIMO-OFDM underwater sound communication system according to claim 1, is characterized in that, described step S1 is specially: considering one has N _tindividual transducer, N _rthe underwater sound MIMO-OFDM communication system of individual hydrophone, suppose in the MIMO underwater sound condition of sparse channel between i-th transducer and a jth hydrophone, in a symbol period, multipath fading is approximate constant, and each transducer is different to the equivalent Doppler factor of hydrophone; Definition Doppler factor ν _p, time delay factor τ _pparameter sets:

ν _p∈{ν ₁,ν ₂,...,ν _l,...,ν _L}τ _p∈{τ ₁,τ ₂,...,τ _k,...,τ _K}

Wherein ν _l=ν ₁+ (l-1) Δ ν, τ _k=τ ₁+ (k-1) Δ τ, Δ ν, Δ τ represent the discrete precision that Doppler and time delay are estimated, can adjust flexibly as required, L, K represent Doppler factor, time delay value number that mimo channel is possible respectively; Assuming that transmitting terminal synchronizing pilot signal is s (t), after the sparse time varying channel of MIMO, a jth hydrophone Received signal strength is expressed as y _j(t).

3. the Doppler estimation of MIMO-OFDM underwater sound communication system according to claim 1, it is characterized in that, described step S2 is specially: set the duration of s (t) as [0, T], at receiving terminal, y (t) is sampled with interval of delta t, make τ in described S1 step ₁=0, sampling interval Δ t=Δ τ, and omit the Δ t in above formula, abbreviation is also write as vector form and can be expressed as y [i]=c [i] ^tx+n [i].

4. the Doppler estimation of MIMO-OFDM underwater sound communication system according to claim 1, it is characterized in that, described step S3 is specially: accumulate all sample values in one-period: y=Cx+n, the vector form of Received signal strength shown in considering, receives vector and regards the linear combination respectively arranged with the x element C that is coefficient as.

5. the Doppler estimation of MIMO-OFDM underwater sound communication system according to claim 1, it is characterized in that, described step S4 is specially: adopt the MIMO-OFDM underwater sound Doppler estimation based on management loading, from the sparse characteristic of underwater acoustic channel, based on super complete dictionary, MIMO Doppler estimation problem is modeled as rarefaction representation problem, management loading algorithm is utilized to solve, under the prerequisite of known y and C, SBL algorithm reconstructed amplitude fading matrix x at that time with self-adaptative adjustment.

6. the Doppler estimation of MIMO-OFDM underwater sound communication system according to claim 1, it is characterized in that, described step S5 is specially: the sparse matrix x reconstructed according to described step S4, find corresponding element in the corresponding C matrix of its non-vanishing vector, then find the Doppler factor of each corresponding mimo channel.

7. the Doppler estimation of MIMO-OFDM underwater sound communication system according to claim 1, is characterized in that, in described step S4, the specific implementation step of SBL algorithm is:

The first step, input: reception vector y, I × (L × K) of I dimension ties up super complete dictionary Matrix C, I=T/ Δ t

Second step, initialization: hyper parameter γ ⁽⁰⁾>=0, β ⁽⁰⁾> 0, number of iterations k=1, ε > 0;

3rd step, algorithm iteration:

A. remember Λ=diag (1/ γ), utilize

$p\left(x\right|y,\mathrm{\γ},\mathrm{\β})={\left(2\mathrm{\π}\right)}^{-\frac{N}{2}}|\mathrm{\Σ}{|}^{-\frac{1}{2}}\exp (-\frac{1}{2}{\left(x-\mathrm{\μ}\right)}^T{\mathrm{\Σ}}^{-1}(x-\mathrm{\μ}\left)\right)$

Σ＝(βC ^TC+Λ) ^-1

Estimate μ respectively ^(k-1), Σ ^(k-1);

B. A=β is remembered ^-1i+C Λ ^-1c ^t, utilize

L(γ|β)＝log|A|+y ^TA ^-1y

${\mathrm{\γ}}_i^{(k+1)}={\mathrm{\μ}}_i^2+{\mathrm{\Σ}}_{ii}$

Undated parameter γ, β, obtain γ respectively ^(k), β ^(k);

C. judge whether to meet || γ ^(k)-γ ^(k+1)||≤ε, if meet, goes to the 4th step; Discontented then k=k+1, repeats the 3rd step 1;

4th step, exports the estimated value of sparse vector: