CN104614703A - Fast super-resolution direction-finding device for two-dimensional broadband signal realized by multiband combined sparse reconstruction method - Google Patents

Abstract

The invention discloses a fast super-resolution direction-finding device for a two-dimensional broadband signal realized by a multiband combined sparse reconstruction method, belongs to the fields of signal estimating and signal processing, and solve the problems that the slow realization speed can be caused by larger calculated amount of a super-resolution direction-finding method for a broadband signal realized by an existing sparse reconstruction technology. The device comprises N DSPs (Digital Signal Processor) and SRAMs (Static Random Access Memory), N DSPs comprise a main DSP and N-1 DSPs; the SRAM is used for storing a spare matrix A(f<k>) of the bandwidth signal, N DSPs are used for uniformly distributing and processing the sampling data of J frequency points in parallel, and adopting the multiband combined sparse reconstruction method to process the sampling data of all frequency points, so as to obtain an angle estimation value of the bandwidth signal. The main DSP processor is used for summarizing and taking the average of angle estimated values obtained by all DSP, and the obtained mean value is a final direction-finding angle value. The fast super-resolution direction-finding device is used for a radar, a sonar, a missile-borne system and a radio detection device.

Description

The two-dimentional broadband signal rapid super-resolution direction-finding device that multiband joint sparse reconstructing method realizes

Technical field

The invention belongs to Signal estimation and signal transacting field.

Background technology

Broadband signal super-resolution direction-finding method can be divided into three major types: the first kind is the method based on maximum likelihood, and under Gaussian noise supposition, the precision based on the algorithm for estimating of maximum likelihood is the highest.Therefore and be of little use although it can reach optimum estimated performance, need the prior imformation of signal source joint Power spectral density, cost function complex structure, operand is large, and easily converges to local extremum; Equations of The Second Kind is the method based on subspace, it comprises again incoherent signal subspace method and coherent signal-subspace method, because these class methods need to process the data of each frequency, and often to carry out pre-estimation to arrival direction, also need to carry out multi-dimensional search simultaneously, particularly very huge for operand 2D signal.And easily cause overall estimated performance lower when low signal-to-noise ratio and fewer snapshots; 3rd class is sparse reconstruct class methods, and this is a kind of method estimating broadband signal DOA of rising in recent ten years.Sparse reconstruct belongs to the category of compressed sensing, by the joint sparse of signal between the sparse expansion structure of broadband array signal model after frequency domain decomposition and frequency band, utilize the methods such as base tracking or orthogonal matching pursuit to estimate to the DOA realizing signal simultaneously, can the special circumstances such as directly processing signals angle intervals is less, relevant, but the calculated amount of tracing process is larger.

The maximum problem of maximum likelihood class methods easily converges to local extremum, and its can cause direct impact to last estimated value; Subspace class methods often need to carry out angle and estimate and take into account focal position reason, also need to carry out multi-dimensional search simultaneously, calculate comparatively complicated; And although sparse reconstruct class methods resolving power is higher, because follow-up tracing process is comparatively loaded down with trivial details, so calculated amount is still larger.

Summary of the invention

The broadband signal super-resolution direction-finding method calculated amount that the object of the invention is to realize to solve existing sparse reconfiguration technique causes more greatly realizing slow problem, the invention provides the two-dimentional broadband signal rapid super-resolution direction-finding device that a kind of multiband joint sparse reconstructing method realizes.

The two-dimentional broadband signal rapid super-resolution direction-finding device that multiband joint sparse reconstructing method of the present invention realizes,

Described direction-finding device comprises 1 SRAM processor and Z dsp processor, and a described Z dsp processor comprises 1 main dsp processor and Z-1 is individual from dsp processor;

SRAM memory, for the sparse matrix of stored wide band signal

Main dsp processor, for receiving the sampled data of J the frequency that hyperchannel wideband digital receiver 2 sends, and will wherein the sampled data uniform distribution of (Z-1) W frequency to Z-1 from dsp processor, wherein J/Z=W, W, Z and J are positive integer; Sparse matrix also for storing according to SRAM memory multiband joint sparse reconstructing method is adopted to obtain the angle estimation value of broadband signal to the sampled data of a remaining W frequency; Also for receiving described multiple broadband signal angle estimation value sent from dsp processor, be averaged the angle estimation value of the broadband signal obtained and the broadband signal angle estimation value summation of reception, the mean value of acquisition is final direction finding angle value;

From dsp processor, for receiving the sampled data of W frequency of distribution; Sparse matrix also for storing according to SRAM memory multiband joint sparse reconstructing method is adopted to obtain the angle estimation value of broadband signal to the sampled data of W frequency; Also for the angle estimation value of acquisition is sent to main dsp processor.The sparse matrix of described sense $\tilde{A}\left(f_k\right)=[a(f_k,{\widetilde{\mathrm{\&phi;}}}_1,{\widetilde{\mathrm{\&theta;}}}_1),a(f_k,{\widetilde{\mathrm{\&phi;}}}_2,{\widetilde{\mathrm{\&theta;}}}_2),...,a(f_k,{\widetilde{\mathrm{\&phi;}}}_p,{\widetilde{\mathrm{\&theta;}}}_p),...,a(f_k,{\widetilde{\mathrm{\&phi;}}}_P,{\widetilde{\mathrm{\&theta;}}}_P)];$

Wherein, vector for array is to direction incident broadband signal is at Frequency point f _kdirection vector, the angle grid set of broadband signal arrival direction $\mathrm{\&Omega;}=(({\widetilde{\mathrm{\&phi;}}}_1,{\widetilde{\mathrm{\&theta;}}}_1),({\widetilde{\mathrm{\&phi;}}}_2,{\widetilde{\mathrm{\&theta;}}}_2),...,({\widetilde{\mathrm{\&phi;}}}_p,{\widetilde{\mathrm{\&theta;}}}_p),...,({\widetilde{\mathrm{\&phi;}}}_P,{\widetilde{\mathrm{\&theta;}}}_P));$

for the position angle of p signal in grid set omega, for the angle of pitch of p signal in grid set omega, p=1,2 ..., P.

Described from dsp processor, according to the sparse matrix that SRAM memory stores the process that the sampled data of W frequency adopts multiband joint sparse reconstructing method to obtain the angle estimation value of broadband signal is comprised:

Step one: according to the sampled data of W sampling frequency, obtain E _s(f _k) and weight matrix W (f _k), 1≤k≤W; Proceed to step 2;

Wherein, weight matrix $W\left(f_k\right)={({\mathrm{\&Sigma;}}_S\left(f_k\right)-{\mathrm{\&sigma;}}^{2}I)}^2{\mathrm{\&Sigma;}}_S^{-1}\left(f_k\right),$ Broadband signal is in frequency f _kdiagonal matrix Σ (the f of the covariance matrix at place _k) by eigenvalue λ ₁, λ ₂λ _mcomposition:

Described eigenwert meets following relation:

λ ₁(f _k)≥λ ₂(f _k)≥…≥λ _N(f _k)>λ _N+1(f _k)＝…＝λ _M(f _k)＝σ ²；

N<M, M are the number of array element in aerial array 1, λ _m(f _k) be frequency f _klocate M eigenwert; σ ²for noise power, I is unit battle array; Σ _s(f _k) be the diagonal matrix of the larger eigenwert of top n in diagonal matrix composition, E _s(f _k) be the larger eigenwert characteristic of correspondence vector of top n in diagonal matrix;

Step 2: establish support set according to the E obtained _s(f _k) and W (f _k), in conjunction with the sparse matrix stored obtain D _k,n, proceed to step 3;

Wherein $D_{k,n}={\tilde{A}}^H\left(f_k\right)R_{k,n},R_{k,n}=Y_{k,n}-\tilde{A}\left(f_k\right){\tilde{T}}_k[:,n],$ Y _k,n=W ^1/2(f _k) [n, n] E _s(f _k) [:, n] and, 1≤n≤N, N is the number of broadband signal;

Step 3: utilize the D obtained _k,n, ask for

Step 4: make λ=0.95, τ _min=0.1, q=1, set Q={t ₀, and τ=D [t ₀], wherein proceed to step 5, t is the index of D, is independent variable;

Step 5: for t ∈ Q, 1≤n≤N, 1≤k≤W, adopts down and out options method to upgrade support set obtain the support set after upgrading proceed to step 6;

Step 6: the support set after utilizing step 5 to upgrade upgrade D, proceed to step 7;

Step 7: the D obtained according to step 6, asks for proceed to step 8;

Step 8: judge whether D [t] is greater than τ, if so, then will gather content update in set Q, simultaneously q=q+1, proceeds to step 9; If not, then τ=λ τ, proceeds to step 9;

Step 9: as q≤N or τ>=τ _mintime, set Q is the set of direction finding angle estimation value, otherwise, proceed to step 5.

Described main dsp processor and adopt shared bus close coupled system or link port cascade loose coupling mode or time division multiplex serial port mode to connect from the connection between dsp processor.

Described $({\widetilde{\mathrm{\&phi;}}}_p,{\widetilde{\mathrm{\&theta;}}}_p)=\arg \underset{{\tilde{T}}_k[:,n],1\&le;n\&le;N,1\&le;k\&le;W}{\min }\frac{1}{2}{\left|\right|Y_{k,n}-\tilde{A}\left(f_k\right){\tilde{T}}_k[:,n]\left|\right|}_2^2+\mathrm{\&tau;}{\left|\right|{\tilde{T}}_k[:,n]\left|\right|}_1$

$s.t.\mathrm{supp}{\tilde{T}}_1[:,1]=...=\mathrm{supp}{\tilde{T}}_1[:,n]=...=\mathrm{supp}{\tilde{T}}_W[:,N];$

Supp{} represents support, Y _k,nfor E _s(f _k) W ^1/2n-th row.

Beneficial effect of the present invention is, direction-finding device of the present invention, according to the joint sparse of signal between frequency band, utilize parallel processing means, the sample point of bay collection is divided into multi-group data, carrying out Frequency Fitting process respectively by often organizing data, afterwards they comprehensively being estimated to the arrival direction (Direction of arrival, DOA) of final signal.The present invention, according to the needs of engineering field Quick Measurement angle, uses multi-disc digital signal processor in actual device simultaneously, adopts parallel processing to realize above method.Substantially increase processing speed, the present invention can be widely used in radar, sonar, missile borne system and other radio detection device.The present invention is when signal relative bandwidth 30%, and signal to noise ratio (S/N ratio) is 0dB, 30 frequencies, under each frequency during fast umber of beats 20, and (1) direction finding precision: 1 °/σ; (2) hardware system operation time :≤50ms.

Accompanying drawing explanation

Fig. 1 is the array signal model described in embodiment one.

Fig. 2 is the principle schematic of broadband signal detection system.

Fig. 3 is the two-dimentional broadband signal rapid super-resolution direction-finding device adopting the multiband joint sparse reconstructing method of shared bus close coupled system to realize in embodiment two.

Fig. 4 is the two-dimentional broadband signal rapid super-resolution direction-finding device adopting the multiband joint sparse reconstructing method of link port cascade loose coupling mode to realize in embodiment three.

Fig. 5 is the two-dimentional broadband signal rapid super-resolution direction-finding device adopting the multiband joint sparse reconstructing method of time division multiplex serial port mode to realize in embodiment four.

Embodiment

Embodiment one: composition graphs 1 and Fig. 2 illustrate present embodiment, broadband signal detection system described in Fig. 1, described detection system comprises broad-band antenna array 1, the two-dimentional broadband signal rapid super-resolution direction-finding device 3 of the multiband joint sparse reconstructing method realization of hyperchannel wideband digital receiver 2 and present embodiment, the two-dimentional broadband signal rapid super-resolution direction-finding device 3 that multiband joint sparse reconstructing method described in present embodiment realizes, described direction-finding device comprises 1 SRAM processor and Z dsp processor, a described Z dsp processor comprises 1 main dsp processor and Z-1 is individual from dsp processor,

SRAM memory, for the sparse matrix of stored wide band signal

Main dsp processor, for receiving the sampled data of J the frequency that hyperchannel wideband digital receiver 2 sends, and will wherein the sampled data uniform distribution of (Z-1) W frequency to Z-1 from dsp processor, wherein J/Z=W, W, Z and J are positive integer; Sparse matrix also for storing according to SRAM memory multiband joint sparse reconstructing method is adopted to obtain the angle estimation value of broadband signal to the sampled data of a remaining W frequency; Also for receiving described multiple broadband signal angle estimation value sent from dsp processor, be averaged the angle estimation value of the broadband signal obtained and the broadband signal angle estimation value summation of reception, the mean value of acquisition is final direction finding angle value;

From dsp processor, for receiving the sampled data of W frequency of distribution; Sparse matrix also for storing according to SRAM memory multiband joint sparse reconstructing method is adopted to obtain the angle estimation value of broadband signal to the sampled data of W frequency; Also for the angle estimation value of acquisition is sent to main dsp processor.

Array signal model as shown in Figure 1, considers have M omnidirectional's array element to constitute a planar antenna array in XY plane, has the irrelevant broadband signal in N number of far field from (φ _n, θ _n) (n=1,2 ..., N) incide on array, wherein φ _nbe the position angle of the n-th signal, θ _nbe the angle of pitch of the n-th signal, ground unrest is zero-mean, variance is σ ²white Gaussian noise, then the signal received by m array element can be expressed as:

$x_m\left(t\right)=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{s}_n(t-{\mathrm{\&tau;}}_{\mathrm{mn}})+n_m\left(t\right),m=\mathrm{1,2},...,M---\left(1\right)$

Wherein ${\mathrm{\&tau;}}_{\mathrm{mn}}=\frac{x_m\cos {\mathrm{\&phi;}}_n\cos {\mathrm{\&theta;}}_n+y_m\sin {\mathrm{\&phi;}}_n\cos {\mathrm{\&theta;}}_n}{c},(m=\mathrm{1,2},...,M;n=\mathrm{1,2},...,N),$ If in observation time △ T, signal sampling number of times is L, then at t _l(l=1,2 ... L) moment, the matrix form that array exports is:

$x\left(t_l\right)=\left[\begin{array}{c}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{s}_n(t_l-{\mathrm{\&tau;}}_{1n})\\ \underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{s}_n(t_l-{\mathrm{\&tau;}}_{2n})\\ .\\ .\\ .\\ \underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{s}_n(t_l-{\mathrm{\&tau;}}_{\mathrm{Mn}})\end{array}\right]+\left[\begin{array}{c}{n}_1\left(t_l\right)\\ n_2\left(t_l\right)\\ .\\ .\\ .\\ n_M\left(t_l\right)\end{array}\right],(l=\mathrm{1,2},...L)---\left(2\right)$

If array signal sub band number is J, then after pair array exports and carries out Fourier transform, the sampling output of array can represent to be become

X(f _k)＝A(f _k)S(f _k)+N(f _k)k＝1,2,…,J (3)

Direction matrix A (f in formula _k) determined by following formula

Vector a (f _k, φ _n, θ _n) for array is to direction (φ _n, θ _n) (n=1,2 ..., N) and incident broadband signal is at Frequency point f _kdirection vector, X (f _k) covariance matrix be:

R(f _k)＝E[X(f _k)X ^H(f _k)]

＝A(f _k)E[S(f _k)S ^H(f _k)]A ^H(f _k)+σ ²I (k＝1,2,…,J) (5)

＝A(f _k)R _S(f _k)A ^H(f _k)+σ ²I

Wherein R _s(f _k)=E [S (f _k) S ^h(f _k)] for signal source be f in frequency _kthe covariance matrix at place.Formula (5) can be write as following form further

$\begin{array}{c}R\left(f_k\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_i\left(f_k\right)e_i\left(f_k\right)e_i^H\left(f_k\right)+\underset{j=N+1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_j\left(f_k\right)e_j\left(f_k\right)e_j^H\left(f_k\right)\\ =\left[\begin{array}{cc}{E}_S\left(f_k\right)& E_N\left(f_k\right)\end{array}\right]\mathrm{\&Sigma;}\left(f_k\right){\left[\begin{array}{cc}{E}_S\left(f_k\right)& E_N\left(f_k\right)\end{array}\right]}^H\\ =E_S\left(f_k\right){\mathrm{\&Sigma;}}_S\left(f_k\right)E_S^H\left(f_k\right)+E_N\left(f_k\right){\mathrm{\&Sigma;}}_N\left(f_k\right)E_N^H\left(f_k\right)\end{array},(k=\mathrm{1,2},...,J)---\left(6\right)$

Diagonal matrix Σ (f _k) be made up of eigenwert

In above formula, eigenwert meets following relation

λ ₁(f _k)≥λ ₂(f _k)≥…≥λ _N(f _k)>λ _N+1(f _k)＝…＝λ _M(f _k)＝σ ²(8)

The direction vector E of signal _s(f _k) be the larger eigenwert characteristic of correspondence vector of top n, E _n(f _k) be all the other M-N less eigenwert characteristic of correspondence vectors, the signal comprehensively on each frequency can obtain Wide-Band Weighted Sub-Space Fitting Direction expression formula

$(\widehat{\mathrm{\&phi;}},\widehat{\mathrm{\&theta;}})=\arg \underset{\mathrm{\&phi;},\mathrm{\&theta;}}{\min }\underset{k=1}{\overset{J}{\mathrm{\&Sigma;}}}{\left|\right|E_S\left(f_k\right)W^{1/2}\left(f_k\right)-A\left(f_k\right)T_k\left|\right|}_F^2---\left(9\right)$

Wherein weight matrix W meets following formula

$W\left(f_k\right)={({\mathrm{\&Sigma;}}_S\left(f_k\right)-{\mathrm{\&sigma;}}^{2}I)}^2{\mathrm{\&Sigma;}}_S^{-1}\left(f_k\right)---\left(10\right)$

If adopt the method for subspace fitting to solve above formula, then need Wideband Focusing and multi-dimensional search, so multifrequency point combined reconstruction can be adopted to process above formula.

Set up transform angle set $\mathrm{\&Omega;}=(({\widetilde{\mathrm{\&phi;}}}_1,{\widetilde{\mathrm{\&theta;}}}_1),({\widetilde{\mathrm{\&phi;}}}_2,{\widetilde{\mathrm{\&theta;}}}_2),...,({\widetilde{\mathrm{\&phi;}}}_p,{\widetilde{\mathrm{\&theta;}}}_p),...,({\widetilde{\mathrm{\&phi;}}}_P,{\widetilde{\mathrm{\&theta;}}}_P)),$ It contains the direction that all angles may arrive, and can obtain following sparse matrix in conjunction with array manifold:

$\tilde{A}\left(f_k\right)=[a(f_k,{\widetilde{\mathrm{\&phi;}}}_1,{\widetilde{\mathrm{\&theta;}}}_1),a(f_k,{\widetilde{\mathrm{\&phi;}}}_2,{\widetilde{\mathrm{\&theta;}}}_2),...,a(f_k,{\widetilde{\mathrm{\&phi;}}}_p,{\widetilde{\mathrm{\&theta;}}}_p),...,a(f_k,{\widetilde{\mathrm{\&phi;}}}_P,{\widetilde{\mathrm{\&theta;}}}_P)]---\left(11\right)$

By the A (f in formula (9) _k) T _kexpand

$A\left(f_k\right)T_k=\tilde{A}\left(f_k\right){\tilde{T}}_k---\left(12\right)$

for row sparse matrix, angle of arrival actual in the corresponding Ω in position of its non-zero row.Each frequency band is corresponding there is identical sparsity structure, formula (9) can be solved by the following method according to this principle

$\underset{{\tilde{P}}_k,1\&le;k\&le;J}{\min }\underset{k=1}{\overset{J}{\mathrm{\&Sigma;}}}{\left|\right|E_S\left(f_k\right)W_k^{1/2}-\tilde{A}\left(f_k\right){\tilde{T}}_k\left|\right|}_F^2$

$\begin{array}{c}s.t.\mathrm{supp}{\tilde{T}}_1[:,1]=...=\mathrm{supp}{\tilde{T}}_1[:,n]=...=\mathrm{supp}{\tilde{T}}_J[:,N],\left|\mathrm{supp}{\tilde{T}}_k\right[:,n\left]\right|=N,\\ 1\&le;n\&le;N,1\&le;k\&le;J\end{array}---\left(13\right)$

Above formula is carried out arranging

$({\widetilde{\mathrm{\&phi;}}}_p,{\widetilde{\mathrm{\&theta;}}}_p)=\arg \underset{{\tilde{T}}_k[:,n],1\&le;n\&le;N,1\&le;k\&le;W}{\min }\frac{1}{2}{\left|\right|Y_{k,n}-\tilde{A}\left(f_k\right){\tilde{T}}_k[:,n]\left|\right|}_2^2+\mathrm{\&tau;}{\left|\right|{\tilde{T}}_k[:,n]\left|\right|}_1$

$s.t.\mathrm{supp}{\tilde{T}}_1[:,1]=...=\mathrm{supp}{\tilde{T}}_1[:,n]=...=\mathrm{supp}{\tilde{T}}_W[:,N];---\left(14\right)$

Supp{} represents support, Y _k,nfor E _s(f _k) W ^1/2n-th row, τ is Dynamic gene.In order to make to be guaranteed with the joint sparse in frequency band between signal band, homotopy Method can be adopted to solve formula (14):

Described from dsp processor, according to the sparse matrix that SRAM memory stores the process that the sampled data of W frequency adopts multiband joint sparse reconstructing method to obtain the angle estimation value of broadband signal is comprised:

Step one: according to the sampled data of W sampling frequency, obtain E _s(f _k) and weight matrix W (f _k), 1≤k≤W; Proceed to step 2;

Wherein, weight matrix $W\left(f_k\right)={({\mathrm{\&Sigma;}}_S\left(f_k\right)-{\mathrm{\&sigma;}}^{2}I)}^2{\mathrm{\&Sigma;}}_S^{-1}\left(f_k\right),$ Broadband signal is in frequency f _kdiagonal matrix Σ (the f of the covariance matrix at place _k) by eigenvalue λ ₁, λ ₂λ _mcomposition:

Described eigenwert meets following relation:

λ ₁(f _k)≥λ ₂(f _k)≥…≥λ _N(f _k)>λ _N+1(f _k)＝…＝λ _M(f _k)＝σ ²；

N<M, M are the number of array element in aerial array 1, λ _m(f _k) be frequency f _klocate M eigenwert; σ ²for noise power, I is unit battle array; Σ _s(f _k) be the diagonal matrix of the larger eigenwert of top n in diagonal matrix composition, E _s(f _k) be the larger eigenwert characteristic of correspondence vector of top n in diagonal matrix;

Step 2: establish support set according to the E obtained _s(f _k) and W (f _k), in conjunction with the sparse matrix stored obtain D _k,n, proceed to step 3;

Wherein $D_{k,n}={\tilde{A}}^H\left(f_k\right)R_{k,n},R_{k,n}=Y_{k,n}-\tilde{A}\left(f_k\right){\tilde{T}}_k[:,n],$ Y _k,n=W ^1/2(f _k) [n, n] E _s(f _k) [:, n] and, 1≤n≤N, N is the number of broadband signal;

Step 3: utilize the D obtained _k,n, ask for

Step 4: make λ=0.95, τ _min=0.1, q=1, set Q={t ₀, and τ=D [t ₀], wherein proceed to step 5, t is the index of D, is independent variable;

Step 5: for t ∈ Q, 1≤n≤N, 1≤k≤W, adopts down and out options method to upgrade support set obtain the support set after upgrading proceed to step 6;

Step 6: the support set after utilizing step 5 to upgrade upgrade D, proceed to step 7;

Step 7: the D obtained according to step 6, asks for proceed to step 8;

Step 8: judge whether D [t] is greater than τ, if so, then will gather content update in set Q, simultaneously q=q+1, proceeds to step 9; If not, then τ=λ τ, proceeds to step 9;

Step 9: as q≤N or τ>=τ _mintime, set Q is the set of direction finding angle estimation value, otherwise, proceed to step 5.

Embodiment two: composition graphs 2 and Fig. 3 illustrate present embodiment, present embodiment is the further restriction to the two-dimentional broadband signal rapid super-resolution direction-finding device that the multiband joint sparse reconstructing method described in embodiment one realizes, broadband signal detection system described in Fig. 2, described detection system comprises broad-band antenna array 1, the broadband signal super-resolution direction-finding device 3 of hyperchannel wideband digital receiver 2 and present embodiment, described direction-finding device 3 adopts 6 digital signal processors, shared bus close coupled system composition multicomputer system is adopted to realize parallel processing.

As shown in Figure 3, digital signal processor adopts the ADSP-TS201S of Analog Device Instruments (ADI) company, adopt 6 dsp processors to above method parallel processing, 6 dsp processors are connected by shared bus close coupled system, after powering on, first program loads to CPLD module 3-7 by PROM module 3-8, to be configured dsp processor 3-1 ~ dsp processor 3-6, program loads to these 6 pieces of DSP processing DSP processor 3-1 ~ 3-6, under first main dsp processor 3-1 calculates each frequency afterwards by bus by FLASH module 3-9 afterwards value, be stored in SRAM module 3-10 afterwards, to call when performing subsequent algorithm, main dsp processor 3-1 starts the observation data receiving J the frequency that hyperchannel wideband digital receiver 2 transmits afterwards, they are divided into Z group, now Z=6, suppose J=30 again, then every sheet DSP can process the observation data of W=30/6=5 frequency, main dsp processor 3-1 by bus by other from dsp processor 3-2 ~ from dsp processor 3-6 be responsible for process observation data pass to them, each dsp processor 3-1 ~ dsp processor 3-6 solves according to the step of above theory deduction afterwards, 5 from dsp processor 3-2 ~ respective estimated value is passed to main dsp processor 3-1 by bus from dsp processor 3-6 afterwards, these results are averaged and draw net result by main dsp processor 3-1 again.Wherein JTAG module 3-11 is responsible for debugging dsp processor 3-1 ~ dsp processor 3-6, and power module 3-12 is responsible for bulk supply, and crystal oscillator module 3-13 is responsible for providing clock, and reseting module 3-14 is responsible for providing reset signal.

Embodiment three: composition graphs 2 and Fig. 4 illustrate present embodiment, present embodiment is the further restriction to the two-dimentional broadband signal rapid super-resolution direction-finding device that the multiband joint sparse reconstructing method described in embodiment one realizes,

Broadband signal detection system described in Fig. 1, described detection system comprises the broadband signal super-resolution direction-finding device 3 of broad-band antenna array 1, hyperchannel wideband digital receiver 2 and present embodiment, described direction-finding device 3 adopts 6 digital signal processors, adopts link port cascade loose coupling mode to form multicomputer system and realizes parallel processing.

As shown in Figure 4, digital signal processor adopts the ADSP-TS201S of Analog Device Instruments (ADI) company, adopt 6 processors to above method parallel processing, 6 dsp processors are connected by link port cascade loose coupling mode, after powering on, first program loads to CPLD module 3-7 by PROM module 3-8, to be configured dsp processor 3-1 ~ dsp processor 3-6, the program of these 6 dsp processors loads to main dsp processor 3-1 by FLASH module 3-9 afterwards, afterwards main dsp processor 3-1 more successively by other from dsp processor 3-2 ~ pass to them by link port one-level one-level from the program of dsp processor 3-6, under main dsp processor 3-1 starts to calculate each frequency afterwards value, be stored in SRAM module 3-10 afterwards, to call when performing subsequent algorithm, main dsp processor 3-1 starts the observation data receiving J the frequency that hyperchannel wideband digital receiver 2 transmits afterwards, they are divided into Z group, now Z=6, suppose J=30 again, then every sheet DSP can process the observation data of W=30/6=5 frequency, main dsp processor 3-1 again by link port by other from dsp processor 3-2 ~ from dsp processor 3-6 be responsible for process observation data one-level one-level successively pass to them, each dsp processor 3-1 ~ dsp processor 3-6 solves according to the step of above theory deduction afterwards, 5 from dsp processor 3-2 ~ respective estimated value is uploaded to main dsp processor 3-1 by link port from dsp processor 3-6 afterwards, these results are averaged and draw net result by main dsp processor 3-1 again.Wherein JTAG module 3-11 is responsible for debugging dsp processor 3-1 ~ dsp processor 3-6, and power module 3-12 is responsible for bulk supply, and crystal oscillator module 3-13 is responsible for providing clock, and reseting module 3-14 is responsible for providing reset signal.

Embodiment four: composition graphs 2 and Fig. 5 illustrate present embodiment, present embodiment is the further restriction to the two-dimentional broadband signal rapid super-resolution direction-finding device that the multiband joint sparse reconstructing method described in embodiment one realizes, broadband signal detection system described in Fig. 1, described detection system comprises broad-band antenna array 1, the broadband signal super-resolution direction-finding device 3 of hyperchannel wideband digital receiver 2 and present embodiment, described direction-finding device 3 adopts 6 digital signal processors, adopt time division multiplex serial port mode to form multicomputer system and realize parallel processing.

As shown in Figure 5, digital signal processor adopts the TMS320C5X of Texas Instruments (TI) company, adopt 6 processors to above method parallel processing, 6 dsp processors are connected by time division multiplex serial port, after powering on, first program loads to CPLD module 3-7 by PROM3-8, to be configured dsp processor 3-1 ~ dsp processor 3-6, program also loads to these 6 pieces of dsp processor 3-1 ~ dsp processor 3-6 by FLASH module 3-9 afterwards, under first main dsp processor 3-1 calculates each frequency afterwards value, be stored in SRAM module 3-10 afterwards, to call when performing subsequent algorithm, main dsp processor 3-1 starts the observation data receiving J the frequency that hyperchannel wideband digital receiver 2 transmits afterwards, they are divided into Z group, now Z=6, suppose J=30 again, then every sheet DSP can process the observation data of W=30/6=5 frequency, main dsp processor 3-1 by time division multiplex serial port by other from dsp processor 3-2 ~ from dsp processor 3-6 be responsible for process observation data pass to them, each dsp processor 3-1 ~ dsp processor 3-6 solves according to the step of above theory deduction afterwards, 5 from dsp processor 3-2 ~ respective estimated value is passed to main dsp processor 3-1 by time division multiplex serial port from dsp processor 3-6 afterwards, these results are averaged and draw net result by main dsp processor 3-1 again.Wherein JTAG module 3-11 is responsible for debugging dsp processor, and power module 3-12 is responsible for bulk supply, and crystal oscillator module 3-13 is responsible for providing clock, and reseting module 3-14 is responsible for providing reset signal.

Claims (4)

1. the two-dimentional broadband signal rapid super-resolution direction-finding device of multiband joint sparse reconstructing method realization, is characterized in that,

Described direction-finding device comprises 1 SRAM processor and Z dsp processor, and a described Z dsp processor comprises 1 main dsp processor and Z-1 is individual from dsp processor;

SRAM memory, for the sparse matrix of stored wide band signal

Main dsp processor, for receiving the sampled data of J the frequency that hyperchannel wideband digital receiver 2 sends, and will wherein the sampled data uniform distribution of (Z-1) W frequency to Z-1 from dsp processor, wherein J/Z=W, W, Z and J are positive integer; Sparse matrix also for storing according to SRAM memory multiband joint sparse reconstructing method is adopted to obtain the angle estimation value of broadband signal to the sampled data of a remaining W frequency; Also for receiving described multiple broadband signal angle estimation value sent from dsp processor, be averaged the angle estimation value of the broadband signal obtained and the broadband signal angle estimation value summation of reception, the mean value of acquisition is final direction finding angle value;

From dsp processor, for receiving the sampled data of W frequency of distribution; Sparse matrix also for storing according to SRAM memory multiband joint sparse reconstructing method is adopted to obtain the angle estimation value of broadband signal to the sampled data of W frequency; Also for the angle estimation value of acquisition is sent to main dsp processor.

2. the two-dimentional broadband signal rapid super-resolution direction-finding device of multiband joint sparse reconstructing method realization according to claim 1, is characterized in that,

The sparse matrix of described sense:

$\tilde{A}\left(f_k\right)=[a(f_k,{\widetilde{\mathrm{\&phi;}}}_1,{\widetilde{\mathrm{\&theta;}}}_1),a(f_k,{\widetilde{\mathrm{\&phi;}}}_2,{\widetilde{\mathrm{\&theta;}}}_2),...,a(f_k,{\widetilde{\mathrm{\&phi;}}}_p,{\widetilde{\mathrm{\&theta;}}}_p),...,a(f_k,{\widetilde{\mathrm{\&phi;}}}_P,{\widetilde{\mathrm{\&theta;}}}_P)];$

Wherein, vector for array is to direction incident broadband signal is at Frequency point f _kdirection vector, the angle grid set of broadband signal arrival direction $\mathrm{\&Omega;}=(({\widetilde{\mathrm{\&phi;}}}_1,{\widetilde{\mathrm{\&theta;}}}_1),({\widetilde{\mathrm{\&phi;}}}_2,{\widetilde{\mathrm{\&theta;}}}_2),...,({\widetilde{\mathrm{\&phi;}}}_p,{\widetilde{\mathrm{\&theta;}}}_p),...,({\widetilde{\mathrm{\&phi;}}}_P,{\widetilde{\mathrm{\&theta;}}}_P));$

for the position angle of p signal in grid set omega, for the angle of pitch of p signal in grid set omega, p=1,2 ..., P.

3. the two-dimentional broadband signal rapid super-resolution direction-finding device of multiband joint sparse reconstructing method realization according to claim 2, is characterized in that,

Described from dsp processor, according to the sparse matrix that SRAM memory stores the process that the sampled data of W frequency adopts multiband joint sparse reconstructing method to obtain the angle estimation value of broadband signal is comprised:

Step one: according to the sampled data of W frequency, obtains E _s(f _k) and weight matrix W (f _k), 1≤k≤W; Proceed to step 2;

Wherein, weight matrix broadband signal is in frequency f _kdiagonal matrix Σ (the f of the covariance matrix at place _k) by eigenvalue λ ₁, λ ₂" λ _mcomposition:

Described eigenwert meets following relation:

λ ₁(f _k)≥λ ₂(f _k)≥…≥λ _N(f _k)>λ _N+1(f _k)＝…＝λ _M(f _k)＝σ ²；

N<M, M are the number of array element in aerial array 1, λ _m(f _k) be frequency f _klocate M eigenwert; σ ²for noise power, I is unit battle array; Σ _s(f _k) be the diagonal matrix of the larger eigenwert of top n in diagonal matrix composition, E _s(f _k) be the larger eigenwert characteristic of correspondence vector of top n in diagonal matrix;

Step 2: establish support set according to the E obtained _s(f _k) and W (f _k), in conjunction with the sparse matrix stored obtain D _k,n, proceed to step 3;

Wherein $D_{k,n}={\tilde{A}}^H\left(f_k\right)R_{k,n},R_{k,n}=Y_{k,n}-\tilde{A}\left(f_k\right){\tilde{T}}_k[:,n],$ Y _k,n=W ^1/2(f _k) [n, n] E _s(f _k) [:, n] and, 1≤n≤N, N is the number of broadband signal;

Step 3: utilize the D obtained _k,n, ask for

Step 4: make λ=0.95, τ _min=0.1, q=1, set Q={t ₀, and τ=D [t ₀], wherein proceed to step 5, t is the index of D, is independent variable;

Step 5: for t ∈ Q, 1≤n≤N, 1≤k≤W, adopts down and out options method to upgrade support set obtain the support set after upgrading proceed to step 6;

Step 6: the support set after utilizing step 5 to upgrade upgrade D, proceed to step 7;

Step 7: the D obtained according to step 6, asks for proceed to step 8;

Step 8: judge whether D [t] is greater than τ, if so, then will gather content update in set Q, simultaneously q=q+1, proceeds to step 9; If not, then τ=λ τ, proceeds to step 9;

Step 9: as q≤N or τ>=τ _mintime, set Q is the set of direction finding angle estimation value, otherwise, proceed to step 5.

4. the two-dimentional broadband signal rapid super-resolution direction-finding device of multiband joint sparse reconstructing method realization according to claim 3, it is characterized in that, described main dsp processor and adopt shared bus close coupled system or link port cascade loose coupling mode or time division multiplex serial port mode to connect from the connection between dsp processor.