CN104218920A - Partitioning concurrence based adaptive digital beamforming method and implementing device thereof - Google Patents

Abstract

The invention discloses a partitioning concurrence based adaptive digital beamforming method and an implementing device thereof. The partitioning concurrence based adaptive digital beamforming method includes (1) transforming two-dimensional baseband data received by a digital array into one-dimensional data vector; (2) partitioning the one-dimensional data vector and corresponding weight vector thereof and a constraint matrix into same M blocks, subjecting the M blocks to parallel processing by M processing modules to obtain M block processing results and directly obtain baseband data output by a comprehensive processing module after adaptive digital wave beams are formed and relevant parameters required for next calculation; (3) repeating according to the step (2) to obtain an adaptive iteration updating structure. The implementing device comprises N unit antennas and corresponding channel receivers and analog/digital converters and the M parallel processing modules and the comprehensive processing module. The partitioning concurrence based adaptive digital beamforming method and the implementing device thereof have the advantages that partitioning is flexible, little mutual information among the blocks is generated, operand of adaptive digital beamforming is effectively reduced and transmission and interface pressure of data convergence is alleviated.

Description

A kind of self-adaptive numerical integration algorithm method of block parallel and implement device thereof

Technical field

The invention belongs to adaptive array signal process field, particularly a kind of self-adaptive numerical integration algorithm method of block parallel and implement device thereof.

Background technology

Self-adaptive numerical integration algorithm is the new technology grown up on array antenna and signal transacting basis, its basic thought is weighted process with the directivity function of control antenna battle array by receiving data to bay, Antenna Array Pattern is made to form high-gain on desired signal direction, produce darker zero to fall in interference signal direction, reach the object of airspace filter.Self-adaptive numerical integration algorithm algorithm can according to the weighted factor of each array element of change self-adaptative adjustment of signal, interference environment, reaches and ensures to expect that sense gain suppresses the object of interference while constant.

But in the application of the large array system of reality, there are the large and data of operand and come together the problems such as difficult.The method being usually used in reducing large array adaptive array computational complexity comprises dimensionality reduction and contraction, and these two kinds of methods have limitation; One of subject matter of dimensionality reduction to produce graing lobe, and partial adaptivity process then can reduce the degree of freedom of array, thus reduce the performance of interference and clutter recognition.The processing method of block parallel has the advantages that to reduce process operand and reduce the base band data amount of coming together, document 1 (Adaptive array beamforming based on an efficient technique.IEEE Trans.AP-44,1996 (8): 1094-1101) a kind of beamforming algorithm (ERLS) of block parallel is disclosed, its Wave beam forming performance is by the impact of block form, and when the dimension of maximum sub-weight vector is larger, the operand of ERLS is very large.Document 2 (Sub-array RLS adaptive algorithm.IEEE Electronics letters, 1999,35 (13): 1061-1062) a kind of asynchronous block parallel algorithm based on RLS is disclosed, each iteration upgrades a weight coefficient needs M-2 beat more than ERLS (M is the block number of piecemeal), and iterative processing efficiency is low.

Summary of the invention

The object of the invention is the self-adaptive numerical integration algorithm method and the implement device thereof that provide a kind of block parallel, effectively reduces operand and the processing time of self-adaptive numerical integration algorithm.

The technical solution realizing the object of the invention is: a kind of self-adaptive numerical integration algorithm method of block parallel, comprises the following steps:

Step 1, be one dimension base band data vector x (n) by the two-dimensional base band data transformation received;

Step 2, weight vectors w (n) determining one dimension base band data vector x (n) correspondence and constraint matrix C;

If the beam position expected is (u ₁, v ₁), the steering vector of this beam position as constraint matrix, i.e. C=a (u ₁, v ₁);

Wherein, u ₁=sin θ ₁, v ₁=cos θ ₁sin φ ₁, θ ₁and φ ₁be respectively the angle of pitch and the azimuth of the arrival bearing of desired signal;

Step 3, respectively piecemeal process is carried out to weight vectors w (n) of one dimension base band data vector x (n) and correspondence and constraint matrix C;

Step 4, make n=0; Initializes weights vector w (0)=a (u ₁, v ₁), integrated treatment module return parameters S _dQ(0)=0 and μ _y(0)=0; M parallel processing module carries out the process of block level respectively to the data block obtained after constraint matrix C piecemeal process in step 3, and M is positive integer;

The block level of i-th parallel processing module is treated to:

$S_i=C_i^H{C}_i$

Wherein S _ifor the mould of piecemeal constraint; C _ifor i-th data block of constraint matrix C; I=1,2...M;

Modules is by the result S after process _ibe sent to integrated treatment module, integrated treatment module is handled as follows again

$S={\left(\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{S}_i\right)}^{-1}$

The S that step 5, a M parallel processing module returns the data block obtained after weight vectors w (n) of one dimension base band data vector x (n) in step 3 and correspondence and the process of constraint matrix C piecemeal and integrated treatment module _dQ(n) and μ _yn () carries out the process of block level;

The block level of i-th parallel processing module is treated to:

$Q_i\left(n\right)=C_i^H{x}_i\left(n\right)$

$y_i\left(n\right)=w_i^H\left(n\right)x_i\left(n\right)$

$D_i\left(n\right)=C_i^H{w}_i\left(n\right)$

w _i(n+1)＝w _i(n)+S _DQ(n)C _i-μ _y(n)x _i(n)

X _i(n) and w _in () is respectively i-th data block of one dimension base band data vector x (n) and weight vectors w (n), Q _i(n) and D _in () is block level process intermediate object program, implication is the result of calculation on the right of equation, i=1,2...M; y _in () is block stages of digital Wave beam forming result;

By Q _i(n), y _i(n), D _i(n) and w _i(n+1) integrated treatment module is sent to;

The block level result of M parallel processing module is handled as follows by step 6, integrated treatment module:

$Q\left(n\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{Q}_i\left(n\right)$

$y\left(n\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{y}_i\left(n\right)$

$D\left(n\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{D}_i\left(n\right)$

μ _y(n+1)＝μy ^H(n)

S _DQ(n+1)＝S[1-D(n)+μ _y(n)Q(n)]

Wherein, Q (n) and D (n) is intermediate object program, and implication is the result of calculation on the right of equation; μ is step factor;

By S _dQand μ (n+1) _y(n+1) M parallel processing module is sent to respectively; The output that y (n) is self-adaptive numerical integration algorithm;

Step 7, when expect beam position constant time, make n=n+1, jump into step 5; When expecting that beam position changes, upgrade (u ₁, v ₁), jump into step 4;

By above-mentioned steps, namely obtain continuous print wave beam y (n) and export.

A self-adaptive numerical integration algorithm device for block parallel, comprises channel receiver and analog to digital converter, a M parallel processing module and 1 integrated treatment module of N number of element antenna and correspondence; Wherein, N, M are positive integer, and N > M;

The signal that N number of element antenna receives is respectively by forming N number of base band data passage after channel receiver and analog to digital conversion, N number of base band data passage is divided into M group and is sent to M parallel processing module respectively, M parallel processing module respectively with integrated treatment module generation closed loop, integrated treatment module utilizes the result of M parallel modules to form digital beam and exports, and calculates corresponding parameter feedback simultaneously and is used as to calculate next time to parallel modules.

The present invention compared with prior art, there is following advantage: (1) block parallel of the present invention and the classification processing and utilizing design feature of antenna array and the array distributed layout of channel receiver, avoids the data transmission bottle neck problem that mass data converges; (2) the present invention adopts parallel processing structure, greatly reduces the operand of self-adaptive numerical integration algorithm, shortens the processing time; (3) two-dimensional base band data acquisition piecemeal of the present invention comes together process, reduces transmission and interface pressure that data centralization comes together.

Accompanying drawing explanation

Fig. 1 is the self-adaptive numerical integration algorithm method flow diagram of block parallel of the present invention.

Fig. 2 is the array element distribution schematic diagram of the two-dimensional array antenna of the embodiment of the present invention.

Fig. 3 (a) for the embodiment of the present invention is at (-10 °, 10 °), the face battle array radiate pattern in (10 °, 25 °) and (30 ° ,-15 °) direction;

Fig. 3 (b) for the embodiment of the present invention is at (0 °, 20 °), the face battle array radiate pattern in (0 °, 35 °) and (0 ° ,-30 °) direction.

Embodiment

Below in conjunction with accompanying drawing, the present invention is further illustrated.

Composition graphs 1, a kind of self-adaptive numerical integration algorithm method of block parallel, comprises the following steps:

Step 1, by N _r× N _cthe two-dimensional base band data transformation that the two-dimensional rectangle digital array antenna that individual element antenna is formed receives is one dimension base band data vector

$x\left(n\right)={\left[\begin{array}{ccc}{x}_{11}\left(n\right)& x_{21}\left(n\right)...x_{N_{r}1}\left(n\right)& x_{12}\left(n\right)...x_{N_{r}2}\left(n\right)...x_{\mathrm{kl}}\left(n\right)...x_{N_r{N}_c}\left(n\right)\end{array}\right]}^T$

Wherein, x _kln () is the base band data that on the n-th sampling instant two-dimensional rectangle front, row k l row place element antenna receives, 1≤k≤N _r, 1≤l≤N _c, N _rand N _cbe positive integer; N=1,2...;

Step 2, weight vectors w (n) determining one dimension base band data vector x (n) correspondence and constraint matrix C;

Weight vectors w (n) is expressed as:

$w\left(n\right)={\left[\begin{array}{ccc}{w}_{11}\left(n\right)& w_{21}\left(n\right)...w_{N_{r}1}\left(n\right)& w_{12}\left(n\right)...w_{\mathrm{kl}}\left(n\right)...w_{N_r{N}_c}\left(n\right)\end{array}\right]}^T$

Wherein, w _kln () is x _kln weight coefficient that () is corresponding, 1≤k≤N _r, 1≤l≤N _c;

The steering vector of desired signal is defined as:

$a_d(u_1,v_1)=a_v\left(v_1\right)\⊗a_u\left(u_1\right)$

$a_u\left(u_1\right)={[1,\exp \{j\frac{{2\mathrm{\πd}}_1}{\mathrm{\λ}}(2-1)u_1\},...\exp \{j\frac{{2\mathrm{\πd}}_1}{\mathrm{\λ}}(N_r-1)u_1\}]}^T$

$a_v\left(v_1\right)={[1,\exp \{j\frac{{2\mathrm{\πd}}_2}{\mathrm{\λ}}(2-1)v_1\},...\exp \{j\frac{{2\mathrm{\πd}}_2}{\mathrm{\λ}}(N_c-1)v_1\}]}^T$

Wherein a _u(u ₁) be N _r× 1 dimensional vector, a _v(v ₁) be N _c× 1 dimensional vector; u ₁=sin θ ₁, v ₁=cos θ ₁sin φ ₁, θ ₁and φ ₁for the angle of pitch and the azimuth of the arrival bearing of desired signal, d ₁and d ₂be respectively the ranks spacing between element antenna, λ is wavelength; a _d(u ₁, v ₁) be (N _r× N _c) × 1 dimensional vector; for kronecker inner product, be defined as

$P\⊗Q=\left\{\begin{array}{cccc}{P}_{11}Q& P_{12}Q& ...& P_{1q}Q\\ P_{21}Q& P_{22}Q& ...& P_{2q}Q\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ P_{p1}Q& P_{p2}Q& ...& P_{\mathrm{pq}}Q\end{array}\right\}$

Wherein, $P=\left\{\begin{array}{cccc}{P}_{11}& P_{12}& ...& P_{1q}\\ P_{21}& P_{22}& ...& P_{2q}\\ .& & .& .\\ .& & .& .\\ .& & .& .\\ P_{p1}& P_{p2}& ...& P_{\mathrm{pq}}\end{array}\right\}$

The computing formula of the steering vector after normalization is:

$(u_1,v_1)=\frac{a_d(u_1,v_1)}{\sqrt{a_d^H}(u_1,v_1)a_d(u_1,v_1)}$

If the beam position expected is (u ₁, v ₁), the steering vector of this beam position as constraint matrix, namely

$C={\left[\begin{array}{ccc}{c}_{11}& c_{21}...c_{N_{r}1}& c_{12}...c_{\mathrm{kl}}...c_{N_r{N}_c}\end{array}\right]}^T=a(u_1,v_1);$

Step 3, respectively piecemeal process is carried out to weight vectors w (n) of one dimension base band data vector x (n) and correspondence and constraint matrix C; Be specially:

By one dimension base band data vector x (n) piecemeal be wherein x _in () is N _ithe vector of × 1,1≤i≤M, and n=N _r× N _c, N _ifor the dimension of each sub-block, M is total sub-block number; Weight vectors w (n) and constraint matrix C resolve into following form:

$\left(n\right)={{[w}_1^T\left(n\right),w_2^T\left(n\right),...,w_M^T\left(n\right)]}^T$

$C={[C_1^T,C_2^T,...,C_M^T]}^T;$

Step 4, make n=0; Initializes weights vector w (0)=a (u ₁, v ₁), integrated treatment module return parameters S _dQ(0)=0 and μ _y(0)=0; M parallel processing module carries out the process of block level respectively to the data block obtained after constraint matrix C piecemeal process in step 3;

The block level of i-th parallel processing module is treated to:

$S_i=C_i^H{C}_i$

Wherein S _ifor the mould of piecemeal constraint; C _ifor i-th data block of constraint matrix C; I=1,2...M;

Modules is by the result S after process _ibe sent to integrated treatment module, integrated treatment module is handled as follows again

$S={\left(\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{S}_i\right)}^{-1}$

The S that step 5, a M parallel processing module returns the data block obtained after weight vectors w (n) of one dimension base band data vector x (n) in step 3 and correspondence and the process of constraint matrix C piecemeal and integrated treatment module _dQ(n) and μ _yn () carries out the process of block level;

The block level of i-th parallel processing module is treated to:

$Q_i\left(n\right)=C_i^H{x}_i\left(n\right)$

$y_i\left(n\right)=w_i^H\left(n\right)x_i\left(n\right)$

$D_i\left(n\right)=C_i^H{w}_i\left(n\right)$

w _i(n+1)＝w _i(n)+S _DQ(n)C _i-μ _y(n)x _i(n)

X _i(n) and w _in () is respectively i-th data block of one dimension base band data vector x (n) and weight vectors w (n), Q _i(n) and D _in () is block level process intermediate object program, implication is the result of calculation on the right of equation, i=1,2...M; y _in () is block stages of digital Wave beam forming result;

By Q _i(n), y _i(n), D _i(n) and w _i(n+1) integrated treatment module is sent to;

The block level result of M parallel processing module is handled as follows by step 6, integrated treatment module:

$Q\left(n\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{Q}_i\left(n\right)$

$y\left(n\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{y}_i\left(n\right)$

$D\left(n\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{D}_i\left(n\right)$

μ _y(n+1)＝μy ^H(n)

S _DQ(n+1)＝S[1-D(n)+μ _y(n)Q(n)]

Wherein, Q (n) and D (n) is intermediate object program, and implication is the result of calculation on the right of equation; μ is step factor; The span of μ is r _xX=E (x (n) x (n) ^h), i.e. x (n) x (n) ^hmathematic expectaion, tr (R _xX) be R _xXtrace of a matrix;

By S _dQand μ (n+1) _y(n+1) M parallel processing module is sent to respectively; The output that y (n) is self-adaptive numerical integration algorithm;

Step 7, when expect beam position constant time, make n=n+1, jump into step 5; When expecting that beam position changes, upgrade (u ₁, v ₁), jump into step 4;

By above-mentioned steps, namely obtain continuous print wave beam y (n) and export.

A self-adaptive numerical integration algorithm device for block parallel, comprises channel receiver and analog to digital converter, a M parallel processing module and 1 integrated treatment module of N number of element antenna and correspondence; Wherein, N, M are positive integer, and N > M;

The signal that N number of element antenna receives is respectively by forming N number of base band data passage after channel receiver and analog to digital conversion, N number of base band data passage is divided into M group and is sent to M parallel processing module respectively, M parallel processing module respectively with integrated treatment module generation closed loop, integrated treatment module utilizes the result of M parallel modules to form digital beam and exports, and calculates corresponding parameter feedback simultaneously and is used as to calculate next time to parallel modules.

Below in conjunction with specific embodiment, the present invention will be further described.

Embodiment 1

Under Matlab environment, the face battle array algorithm of block parallel is emulated.Suppose the uniform surface battle array of 16 × 16, array-element antenna directional diagram isotropism, array element distance is half-wavelength.Suppose: receive in data and comprise additive white Gaussian noise; Signal is from (0 °, 0 °) direction, and signal noise ratio (SNR) is 0dB; Interference signal and signal irrelevant, and interference-to-noise ratio (INR) is 30dB; Equal and the N of the dimension of every sub-piecemeal _i=32, total block count M=8.Get consider randomness, all simulation results are all that the result after 100 independent operatings is averaged and obtains.In order to the side lobe performance reached, add 30dB Dolph-Chebyshev in the algorithm and retrain.

Fig. 3 (a) be three interference signals respectively from (-10 °, 10 °), the face battle array radiate pattern in (10 °, 25 °) and (30 ° ,-15 °) direction; At three interference radiating way, the degree of depth that directional diagram zero falls into is respectively :-58dB ,-51dB and-60dB.

Fig. 3 (b) be three interference signals respectively from (0 °, 20 °), the face battle array radiate pattern in (0 °, 35 °) and (0 ° ,-30 °) direction is in horizontal angle the cross section, angle of pitch direction at place; As seen from the figure, at three interference radiating way, the degree of depth that directional diagram zero falls into is respectively :-55dB ,-54dB and-56dB.

Above-mentioned simulation result shows that the face battle array LCMV adaptive beam-forming algorithm of block parallel is identical with not having block parallel performance before treatment, can adaptive suppression interference signal while sense gain is expected in guarantee, and interference rejection capability is strong.

Piecemeal of the present invention is flexible, and point interblock interactive information is few, effectively can be reduced the operand of self-adaptive numerical integration algorithm by piecemeal process, reduces transmission and interface pressure that data come together.

Claims (6)

1. a self-adaptive numerical integration algorithm method for block parallel, is characterized in that, comprise the following steps:

Step 1, be one dimension base band data vector x (n) by the two-dimensional base band data transformation received;

Step 2, weight vectors w (n) determining one dimension base band data vector x (n) correspondence and constraint matrix C;

If the beam position expected is (u ₁, v ₁), the steering vector of this beam position as constraint matrix, i.e. C=a (u ₁, v ₁);

Wherein, u ₁=sin θ ₁, v ₁=cos θ ₁sin φ ₁, θ ₁and φ ₁be respectively the angle of pitch and the azimuth of the arrival bearing of desired signal;

Step 3, respectively piecemeal process is carried out to weight vectors w (n) of one dimension base band data vector x (n) and correspondence and constraint matrix C;

Step 4, make n=0; Initializes weights vector w (0)=a (u ₁, v ₁), integrated treatment module return parameters S _dQ(0)=0 and μ _y(0)=0; M parallel processing module carries out the process of block level respectively to the data block obtained after constraint matrix C piecemeal process in step 3, and M is positive integer;

The block level of i-th parallel processing module is treated to:

$S_i=C_i^H{C}_i$

Wherein S _ifor the mould of piecemeal constraint; C _ifor i-th data block of constraint matrix C; I=1,2...M;

Modules is by the result S after process _ibe sent to integrated treatment module, integrated treatment module is handled as follows

$S={\left(\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{S}_i\right)}^{-1}$

The S that step 5, a M parallel processing module returns the data block obtained after weight vectors w (n) of one dimension base band data vector x (n) in step 3 and correspondence and the process of constraint matrix C piecemeal and integrated treatment module _dQ(n) and μ _yn () carries out the process of block level;

The block level of i-th parallel processing module is treated to:

$Q_i\left(n\right)=C_i^H{x}_i\left(n\right)$

$y_i\left(n\right)=w_i^H\left(n\right)x_i\left(n\right)$

$D_i\left(n\right)=C_i^H{w}_i\left(n\right)$

w _i(n+1)＝w _i(n)+S _DQ(n)C _i-μ _y(n)x _i(n)

X _i(n) and w _in () is respectively i-th data block of one dimension base band data vector x (n) and weight vectors w (n), Q _i(n) and D _in () is block level process intermediate object program, implication is the result of calculation on the right of equation, i=1,2...M; y _in () is block stages of digital Wave beam forming result;

By Q _i(n), y _i(n), D _i(n) and w _i(n+1) integrated treatment module is sent to;

The block level result of M parallel processing module is handled as follows by step 6, integrated treatment module:

$Q\left(n\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{Q}_i\left(n\right)$

$y\left(n\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{y}_i\left(n\right)$

$D\left(n\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{D}_i\left(n\right)$

μ _y(n+1)＝μy ^H(n)

S _DQ(n+1)＝S[1-D(n)+μ _y(n)Q(n)]

Wherein, Q (n) and D (n) is intermediate object program, and implication is the result of calculation on the right of equation; μ is step factor;

By S _dQand μ (n+1) _y(n+1) M parallel processing module is sent to respectively; The output that y (n) is self-adaptive numerical integration algorithm;

Step 7, when expect beam position constant time, make n=n+1, jump into step 5; When expecting that beam position changes, upgrade (u ₁, v ₁), jump into step 4;

By above-mentioned steps, obtain continuous print wave beam y (n) and export.

2. the self-adaptive numerical integration algorithm method of block parallel according to claim 1, is characterized in that, is that one dimension base band data vector x (n) is specially by the two-dimensional base band data transformation received described in step 1:

By N _r× N _cthe two-dimensional base band data transformation that the two-dimensional rectangle digital array antenna that individual element antenna is formed receives is one dimension base band data vector

$x\left(n\right)={\left[\begin{array}{ccc}{x}_{11}\left(n\right)& x_{21}\left(n\right)...x_{N_{r}1}\left(n\right)& x_{12}\left(n\right)...x_{N_{r}2}\left(n\right)...x_{\mathrm{kl}}\left(n\right)...x_{N_r{N}_c}\left(n\right)\end{array}\right]}^T$

Wherein, x _kln () is the base band data that on the n-th sampling instant two-dimensional rectangle front, row k l row place element antenna receives, 1≤k≤N _r, 1≤l≤N _c, N _rand N _cbe positive integer, n=1,2....

3. the self-adaptive numerical integration algorithm method of block parallel according to claim 2, is characterized in that, determines weight vectors w (n) that one dimension base band data vector x (n) is corresponding and constraint matrix C described in step 2

Be specially:

Weight vectors w (n) is expressed as:

$w\left(n\right)={\left[\begin{array}{ccc}{w}_{11}\left(n\right)& w_{21}\left(n\right)...w_{N_{r}1}\left(n\right)& w_{12}\left(n\right)...w_{\mathrm{kl}}\left(n\right)...w_{N_r{N}_c}\left(n\right)\end{array}\right]}^T$

Wherein, w _kln () is x _kln weight coefficient that () is corresponding;

The steering vector of desired signal is defined as:

$a_d(u_1,v_1)=a_v\left(v_1\right)\⊗a_u\left(u_1\right)$

$a_u\left(u_1\right)={[1,\exp \{j\frac{{2\mathrm{\πd}}_1}{\mathrm{\λ}}(2-1)u_1\},...\exp \{j\frac{{2\mathrm{\πd}}_1}{\mathrm{\λ}}(N_r-1)u_1\}]}^T$

$a_v\left(v_1\right)={[1,\exp \{j\frac{{2\mathrm{\πd}}_2}{\mathrm{\λ}}(2-1)v_1\},...\exp \{j\frac{{2\mathrm{\πd}}_2}{\mathrm{\λ}}(N_c-1)v_1\}]}^T$

Wherein a _u(u ₁) be N _r× 1 dimensional vector, a _v(v ₁) be N _c× 1 dimensional vector; u ₁=sin θ ₁, v ₁=cos θ ₁sin φ ₁, θ ₁and φ ₁for the angle of pitch and the azimuth of the arrival bearing of desired signal, d ₁and d ₂be respectively the ranks spacing between element antenna, λ is wavelength; a _d(u ₁, v ₁) be (N _r× N _c) × 1 dimensional vector; for kronecker inner product, be defined as:

$P\⊗Q=\left\{\begin{array}{cccc}{P}_{11}Q& P_{12}Q& ...& P_{1q}Q\\ P_{21}Q& P_{22}Q& ...& P_{2q}Q\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ P_{p1}Q& P_{p2}Q& ...& P_{\mathrm{pq}}Q\end{array}\right\}$

Wherein, $P=\left\{\begin{array}{cccc}{P}_{11}& P_{12}& ...& P_{1q}\\ P_{21}& P_{22}& ...& P_{2q}\\ .& & .& .\\ .& & .& .\\ .& & .& .\\ P_{p1}& P_{p2}& ...& P_{\mathrm{pq}}\end{array}\right\}$

The computing formula of the steering vector after normalization is:

$(u_1,v_1)=\frac{a_d(u_1,v_1)}{\sqrt{a_d^H}(u_1,v_1)a_d(u_1,v_1)}$

If the beam position expected is (u ₁, v ₁), the steering vector of this beam position as constraint matrix, namely

$C={\left[\begin{array}{ccc}{c}_{11}& c_{21}...c_{N_{r}1}& c_{12}...c_{\mathrm{kl}}...c_{N_r{N}_c}\end{array}\right]}^T=a(u_1,v_1)$

Wherein, 1≤k≤N _r, 1≤l≤N _c; u ₁=sin θ ₁, v ₁=cos θ ₁sin φ ₁, θ ₁and φ ₁for the angle of pitch and the azimuth of the arrival bearing of desired signal.

4. the self-adaptive numerical integration algorithm method of block parallel according to claim 1, it is characterized in that, described in step 3, piecemeal process is carried out respectively to weight vectors w (n) of one dimension base band data vector x (n) and correspondence and constraint matrix C and be specially:

By one dimension base band data vector x (n) piecemeal be , wherein x _in () is N _ithe vector of × 1,1≤i≤M, and n=N _r× N _c, N _ifor the dimension of each sub-block, M is total sub-block number; Weight vectors w (n) and constraint matrix C resolve into following form:

$\left(n\right)={{[w}_1^T\left(n\right),w_2^T\left(n\right),...,w_M^T\left(n\right)]}^T$

$C={[C_1^T,C_2^T,...,C_M^T]}^T.$

5. the self-adaptive numerical integration algorithm method of block parallel according to claim 1, is characterized in that, in step 6, the span of step factor μ is

Wherein, R _xX=E (x (n) x (n) ^h), i.e. x (n) x (n) ^hmathematic expectaion, tr (R _xX) be R _xXtrace of a matrix.

6. based on an implement device for the self-adaptive numerical integration algorithm method of block parallel described in claim 1, it is characterized in that, comprise channel receiver and analog to digital converter, a M parallel processing module and 1 integrated treatment module of N number of element antenna and correspondence; Wherein, N, M are positive integer, and N > M;

The signal that N number of element antenna receives is respectively by forming N number of base band data passage after channel receiver and analog to digital conversion, N number of base band data passage is divided into M group and is sent to M parallel processing module respectively, M parallel processing module respectively with integrated treatment module generation closed loop, integrated treatment module utilizes the result of M parallel modules to form digital beam and exports, and calculates corresponding parameter feedback simultaneously and is used as to calculate next time to parallel modules.