CN103490848B - The method and device that a kind of degree of rarefication is estimated - Google Patents

Abstract

The invention provides the method and device that a kind of degree of rarefication is estimated, it is applicable to signal processing technology field, described method includes: according to number and the preset schedule strategy of the sampled point calculated, send sampling instruction to multiple subscriber terminal equipments, obtain the sampled point needed for current iteration step, the sampled point gathered according to all iterative steps before the sampled point needed for current iteration step, current iteration step and multiple observation vectors of sparse signal to be observed, generate the object function of the least mean-square error that adds up based on L1/L2 mixing norm and multistep；According to described object function, sparse signal to be observed is carried out joint sparse degree estimation, and obtains the result that joint sparse degree is estimated；The result estimated according to joint sparse degree, estimates described joint sparse degree to carry out terminating judgement.The invention enables poor in signal to noise ratio condition and without under the actual application scenarios of any prior information, still can effectively perform degree of rarefication accurately and estimate, minimize the total sampling overhead of system simultaneously.

Description

The method and device that a kind of degree of rarefication is estimated

Technical field

The invention belongs to signal processing technology field, particularly relate to the method and device that a kind of degree of rarefication is estimated.

Background technology

In the prior art, the degree of rarefication (number of nonzero element in sparse coefficient vector) of signal is mostly assumed It is previously known, and is chosen for guaranteeing correctly to rebuild original signal institute as prior information according to this degree of rarefication The number of the sampled point that need to collect.But in actual applications, this prior information of the degree of rarefication of signal has can Can be limited.Such as, in the detection application of cognitive radio (Cognitive Radio, CR) broader frequency spectrum In scene, owing at CR user's frequency spectrum detection node, the degree of rarefication of wideband received signal directly depends on mandate The spectrum occupancy of user (PU, Primary User), but due between PU system and CR system Generally without direct interaction information, generally have to use maximum statistic (the i.e. maximum degree of rarefication) generation of degree of rarefication The number being defined as rebuilding the sampled point needed for original signal is gone for actual degree of rarefication.Sparse for solving the above Degree prior information this problem limited, at present, it is proposed that degree of rarefication estimates (SOE, Sparsity Order Estimation) concept and method.

In the technical scheme that existing degree of rarefication is estimated, generally by unique user, according to maximum degree of rarefication, this is limited Prior information S_maxDetermine the number of required sampled point, and to actual degree of rarefication S_nzCarry out the list of unique user Step degree of rarefication is estimated, prior art there is problems in that

1, generally signal to noise ratio is relatively low in actual applications, and under low signal-to-noise ratio scene, unique user degree of rarefication is estimated The degree of rarefication of signal cannot correctly would generally be estimated by meter due to having a strong impact on of noise.

2, existing single step degree of rarefication method of estimation, needs according to maximum degree of rarefication S_maxDetermine the individual of sampled point Number is with to actual degree of rarefication S_nzEstimate, because S_nz<S_maxAnd required sampling overhead is the increasing of degree of rarefication Function, so by S_maxDetermined by the number of sampled point will be greater than the number of actually required sampled point, lead Cause to exist during degree of rarefication is estimated sampling overhead waste.

3 and when in real system without any hypothesis prior information, existing single step degree of rarefication method of estimation is by nothing Method determines that the number of required sampled point, correctly to estimate actual degree of rarefication, makes the existing method cannot be effective Perform.

To sum up, how to realize low signal-to-noise ratio and without any prior information scene under still can be to sparse signal Carrying out correctly degree of rarefication to estimate, in minimizing system simultaneously, total sampling overhead is urgently to be resolved hurrily during degree of rarefication is estimated Problem.

Summary of the invention

A kind of method that the purpose of the embodiment of the present invention is to provide degree of rarefication to estimate, it is intended to solve in low noise Compare and without sparse signal correctly being carried out degree of rarefication estimation and be under any prior information scene The problem that total sampling overhead of uniting is bigger.

To achieve these goals, the following technical scheme of embodiment of the present invention offer:

The embodiment of the present invention is achieved in that a kind of method that degree of rarefication is estimated, described method include:

The number of the sampled point needed for calculating current iteration step；

The number of the sampled point according to described calculating and preset schedule strategy, send sampling instruction to multiple terminals Equipment, so that terminal unit is according to described sampling instruction, performs to sample to obtain sampled data, described sampling Instruction includes: the number of the sampled point that terminal unit needs obtain and terminal unit are for generating sampling matrix Random seed sequence；

Receive the sampled data that each local terminal reports, as adopting needed for the current iteration step obtained Sampling point；

Adopt according to iterative steps all before the sampled point needed for described current iteration step, current iteration step Collection sampled point and multiple observation vectors of sparse signal to be observed, generate based on L1/L2 mixing norm and Multistep adds up the object function of least mean-square error；

According to described object function, sparse signal to be observed is carried out joint sparse degree estimation, and obtains associating The result that degree of rarefication is estimated, described result includes: the degree of rarefication estimated value of sparse signal to be observed and described dilute Dredge the positional information of nonzero element corresponding to degree estimated value；

The result estimated according to described joint sparse degree, estimates described joint sparse degree to carry out terminating judgement.

The embodiment of the present invention additionally provides a kind of method that degree of rarefication is estimated, described method includes:

The sampling instruction that reception system sends, described sampling instruction includes: the sampling that terminal unit needs obtain The number of point and terminal unit are for generating the random seed sequence of sampling matrix；

According to described sampling instruction, terminal unit performs sampling to obtain sampled data；

By described sampled data reporting system.

The embodiment of the present invention additionally provides the device that a kind of degree of rarefication is estimated, described device includes:

Computing unit, for calculating the number of the sampled point needed for current iteration step；

Transmitting element, according to number and the preset schedule strategy of the sampled point of described calculating, sends sampling instruction To multiple terminal units, so that terminal unit is according to described sampling instruction, perform to sample to obtain sampled data, Described sampling instruction includes: number and the terminal unit of the sampled point that terminal unit needs obtain are adopted for generation The random seed sequence of sample matrix；

Receive unit, for receiving the sampled data that each local terminal reports, currently change as obtain Sampled point needed for riding instead of walk suddenly；

Signal generating unit, before according to the sampled point needed for described current iteration step, current iteration step All iterative steps gather sampled point and multiple observation vectors of sparse signal to be observed, generate based on L1/L2 mixing norm and multistep add up the object function of least mean-square error；

Estimation unit, for according to described object function, carries out joint sparse degree to sparse signal to be observed and estimates Meter, and obtain the result that joint sparse degree is estimated, described result includes: the degree of rarefication of sparse signal to be observed The positional information of the nonzero element that estimated value is corresponding with described degree of rarefication estimated value；

Decision unit, for the result estimated according to described joint sparse degree, estimates described joint sparse degree Carry out terminating judgement.

The embodiment of the present invention additionally provides the device that a kind of degree of rarefication is estimated, described device includes:

Receiving unit, for receiving the sampling instruction that system sends, described sampling instruction includes: terminal unit The number of the sampled point that needs obtain and terminal unit are for generating the random seed sequence of sampling matrix；

Sampling unit, for according to described sampling instruction, terminal unit performs sampling to obtain sampled data；

Report unit, for by described sampled data reporting system.

The embodiment of the present invention compared with prior art, has the beneficial effects that: needed for calculating current iteration step The number of sampled point, according to number and the preset schedule strategy of the sampled point of described calculating, sends sampling instruction To multiple subscriber terminal equipments, obtain the sampled point needed for current iteration step, walk according to described current iteration Sampled point that before rapid required sampled point, current iteration step, all iterative steps gather and to be observed dilute Dredge multiple observation vectors of signal, generate and add up least mean-square error based on L1/L2 mixing norm and multistep Object function, according to described object function, carries out joint sparse degree estimation, and obtains sparse signal to be observed Take the result that joint sparse degree is estimated；The result estimated according to described joint sparse degree, to described joint sparse Degree estimation carries out terminating judgement.It is many to same signal to be observed that the embodiment of the present invention realizes utilization system The joint sparse of individual observation vector, and the learning gain during multi-Step Iterations so that at signal to noise ratio bar Part is poor and without under the actual application scenarios of any prior information, still can effectively perform degree of rarefication accurately Estimate, minimize the total sampling overhead of system simultaneously.

Accompanying drawing explanation

In order to be illustrated more clearly that the technical scheme of the embodiment of the present invention, in embodiment being described below The required accompanying drawing used is briefly described, it should be apparent that, the accompanying drawing in describing below is only Some embodiments of the present invention, for those of ordinary skill in the art, are not paying creative labor On the premise of Dong, it is also possible to obtain other accompanying drawing according to these accompanying drawings.

Fig. 1 is the flow chart of the realization of the method that the degree of rarefication that the embodiment of the present invention one provides is estimated；

Fig. 2 is the flow chart of the realization of the method that the degree of rarefication that the embodiment of the present invention two provides is estimated；

Fig. 3 is the flow chart of the realization of the method that the degree of rarefication that the embodiment of the present invention three provides is estimated；

Fig. 4 is the structure chart of the device that the degree of rarefication that the embodiment of the present invention four provides is estimated；

Fig. 5 is the structure chart of the device that the degree of rarefication that the embodiment of the present invention four provides is estimated；

Fig. 6 is the structure chart of the device that the degree of rarefication that the embodiment of the present invention six provides is estimated.

Detailed description of the invention

In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with accompanying drawing and reality Execute example, the present invention is further elaborated.Only should be appreciated that specific embodiment described herein Only in order to explain the present invention, it is not intended to limit the present invention.

Embodiments providing a kind of degree of rarefication method of estimation, described method includes:

The number of the sampled point needed for calculating current iteration step；

The number of the sampled point according to described calculating and preset schedule strategy, send sampling instruction to multiple terminals Equipment, so that terminal unit is according to described sampling instruction, performs to sample to obtain sampled data, described sampling Instruction includes: the number of the sampled point that terminal unit needs obtain and terminal unit are for generating sampling matrix Random seed sequence；

Receive the sampled data that each local terminal reports, as adopting needed for the current iteration step obtained Sampling point；

Adopt according to iterative steps all before the sampled point needed for described current iteration step, current iteration step Collection sampled point and multiple observation vectors of sparse signal to be observed, generate based on L1/L2 mixing norm and Multistep adds up the object function of least mean-square error；

According to described object function, sparse signal to be observed is carried out joint sparse degree estimation, and obtains associating The result that degree of rarefication is estimated, described result includes: the degree of rarefication estimated value of sparse signal to be observed and described dilute Dredge the positional information of nonzero element corresponding to degree estimated value；

The result estimated according to described joint sparse degree, estimates described joint sparse degree to carry out terminating judgement.

The embodiment of the present invention additionally provides a kind of degree of rarefication method of estimation, and described method includes:

The sampling instruction that reception system sends, described sampling instruction includes: the sampling that terminal unit needs obtain The number of point and terminal unit are for generating the random seed sequence of sampling matrix；

According to described sampling instruction, terminal unit performs sampling to obtain sampled data；

By described sampled data reporting system.

The embodiment of the present invention additionally provides a kind of degree of rarefication estimation unit, and described device includes:

Computing unit, for calculating the number of the sampled point needed for current iteration step；

Transmitting element, according to number and the preset schedule strategy of the sampled point of described calculating, sends sampling instruction To multiple terminal units, so that terminal unit is according to described sampling instruction, perform to sample to obtain sampled data, Described sampling instruction includes: number and the terminal unit of the sampled point that terminal unit needs obtain are adopted for generation The random seed sequence of sample matrix；

Receive unit, for receiving the sampled data that each local terminal reports, currently change as obtain Sampled point needed for riding instead of walk suddenly；

Signal generating unit, before according to the sampled point needed for described current iteration step, current iteration step All iterative steps gather sampled point and multiple observation vectors of sparse signal to be observed, generate based on L1/L2 mixing norm and multistep add up the object function of least mean-square error；

Estimation unit, for according to described object function, carries out joint sparse degree to sparse signal to be observed and estimates Meter, and obtain the result that joint sparse degree is estimated, described result includes: the degree of rarefication of sparse signal to be observed The positional information of the nonzero element that estimated value is corresponding with described degree of rarefication estimated value；

Decision unit, for the result estimated according to described joint sparse degree, estimates described joint sparse degree Carry out terminating judgement.

The embodiment of the present invention additionally provides a kind of degree of rarefication estimation unit, and described device includes:

Receiving unit, for receiving the sampling instruction that system sends, described sampling instruction includes: terminal unit The number of the sampled point that needs obtain and terminal unit are for generating the random seed sequence of sampling matrix；

Sampling unit, for according to described sampling instruction, terminal unit performs sampling to obtain sampled data；

Report unit, for by described sampled data reporting system.

Below in conjunction with specific embodiment, the realization of the present invention is described in detail:

Embodiment one

Fig. 1 shows the flowchart of the method that the degree of rarefication that the embodiment of the present invention one provides estimates, describes in detail As follows:

In S101, the number of the sampled point needed for calculating current iteration step；

In S102, according to number and the preset schedule strategy of the sampled point of described calculating, send sampling and refer to Order, to multiple terminal units, so that terminal unit is according to described sampling instruction, performs sampling to obtain hits According to, described sampling instruction includes: number and the terminal unit of the sampled point that terminal unit needs obtain are used for giving birth to Become the random seed sequence of sampling matrix；

In S103, receive the sampled data that each local terminal reports, as the current iteration obtained Sampled point needed for step；

In the present embodiment, send sampling instruction extremely multiple terminal units, receive what each local terminal reported Sampled data, it is achieved combine acquisition sampled point by multiple terminal units.

In S104, according to institute before the sampled point needed for described current iteration step, current iteration step There are sampled point and multiple observation vectors of sparse signal to be observed that iterative step gathers, generate based on L1/L2 Mixing norm and multistep add up the object function of least mean-square error；

In S105, according to described object function, sparse signal to be observed is carried out joint sparse degree estimation, And obtaining the result that joint sparse degree is estimated, described result includes: the degree of rarefication of sparse signal to be observed is estimated It is worth the positional information of the nonzero element corresponding with described degree of rarefication estimated value；

In S106, the result estimated according to described joint sparse degree, estimate to carry out to described joint sparse degree Terminate judgement.

In the present embodiment, S106 particularly as follows: according to described joint sparse degree estimate result, it is judged that described Close whether degree of rarefication estimated result meets termination judgment condition, the most then terminate joint sparse degree and estimate, and Output degree of rarefication estimated result, if it is not, then perform step S101.

In the present embodiment, the number of the sampled point needed for calculating current iteration step, according to adopting of described calculating The number of sampling point and preset schedule strategy, send sampling instruction extremely multiple subscriber terminal equipments, and acquisition is current repeatedly Ride instead of walk rapid required sampled point, according to the sampled point needed for described current iteration step, current iteration step it Before all iterative steps gather sampled point and multiple observation vectors of sparse signal to be observed, generate based on L1/L2 mixing norm and multistep add up the object function of least mean-square error, according to described object function, right Sparse signal to be observed carries out joint sparse degree estimation, and obtains the result that joint sparse degree is estimated；According to institute State the result that joint sparse degree is estimated, estimate described joint sparse degree to carry out terminating judgement.The present invention implements Example realizes the utilization system joint sparse to multiple observation vectors of same signal to be observed, Yi Jiduo Learning gain in step iterative process so that poor in signal to noise ratio condition and without any prior information reality should With under scene, still can effectively perform degree of rarefication accurately and estimate, minimize the total sampling overhead of system simultaneously.

Embodiment two

Embodiment of the present invention application scenarios can be multiple LUTs composition that base station is corresponding with base station System, wherein assumes there be J LUT, below by way of intrasystem base station side and subscriber terminal side The method estimating degree of rarefication illustrates, but the application scenarios of the method for embodiment of the present invention degree of rarefication estimation Being not limited, what Fig. 2 showed the method that the degree of rarefication that the embodiment of the present invention two provides estimates realizes flow process Figure, details are as follows:

In S201, parameter initialization；

In the present embodiment, initialized parameter includes: the degree of rarefication estimated value of a upper iterative step Currently Number M of the sampled point that all iterative steps gather before iterative step_pt, base station initiated after, due to without any Prior information, first base station starts to initialize from less initial value, thus avoids initialization value to be more than reality Degree of rarefication and the unnecessary sampling overhead that causes, such as, can initialize Wherein, wherein S₀For in advance If degree of rarefication initial value, S₀Can be 0, described The position of corresponding nonzero element Initial value For empty set I.e. M_pt=0。

In S202, the number of the sampled point needed for calculating current iteration step；

In the present embodiment, S202 is specifically as follows: according to the degree of rarefication estimated value of a upper iterative step When Number M of the sampled point that all iterative steps gather before front iterative step_pt, preset adjacent twice iteration step The number smallest incremental Δ of rapid sampled point, calculates number M of sampled point needed for current iteration step_r, tool Body is as follows:

$M_r=\max \{M\left({\hat{S}}_p\right),M_{\mathrm{pt}}+\mathrm{\&Delta;}\}-M_{\mathrm{pt}}$

Wherein, max{, for taking maxima operation, M () is that degree of rarefication estimates that the number of required sampled point calculates Function, Wherein N is signal dimension, For the operation that rounds up.

In the present embodiment, for adjacent twice iterative step, identical Δ can be set, i.e. change for whole A Δ is used, it is also possible to different Δs is set, i.e. multiple Δs, example is used for whole iterative process for process As, when iterative process starts, can Δ value be arranged is larger, and when close to estimated value, permissible Δ value arranged is smaller.

In S203, according to number and the preset schedule strategy of the sampled point of described calculating, send sampling and refer to Order is to multiple subscriber terminal equipments, so that terminal unit is according to described sampling instruction, performs sampling and adopts to obtain Sample data, described sampling instruction includes: the number of the sampled point that terminal unit needs obtain and terminal unit are used In the random seed sequence generating sampling matrix；

In the present embodiment, each LUT equipment, according to described sampling instruction, dispatches the sampling money of self Source is to obtain sampled data, and wherein sampling resource can be that the software of LUT equipment, hardware etc. can For carrying out the resource sampled.

In the present embodiment, preset schedule strategy can use: the first scheduling strategy: all intrasystem local use The number of the sampled point needed for the mean allocation system current iteration step of family, such as, jth local user will Allocated acquisition The number of the sampled point needed for this step individual, or, the second scheduling strategy: According to the performance of intrasystem local user, according to the sampled point needed for weight distribution system current iteration step Number, wherein, the performance of local user may include that channel quality corresponding to local user, processes energy Power and battery power consumption etc., for example, it is possible to distribution jth local user Adopting needed for this step individual The number of sampling point, wherein α_jFor jth, local user distributes weight, when jth local user channel quality, When disposal ability and battery power consumption are preferable, then α_jRelatively big, on the contrary the least, and meet ∑_jα_j=1。

In the present embodiment, S203 is specifically as follows: base station is according to the number of the sampled point of described calculating and presets Scheduling strategy sends sampling and instructs to LUT equipment each in system, so that each local user in system Terminal unit starts to perform low speed sampling, according to the number of the sampled point needing to obtain of distribution, it is thus achieved that local Sampled data, and this sampled data is reported base station.Owing in system, each LUT equipment is only It is responsible for local low speed sampling, and the sampled data reporting base station that will obtain through low speed sampling, base station adopts The number of sampling point calculates renewal, scheduling of resource, degree of rarefication estimation, compares the operations such as judgement, therefore, can drop The computation complexity of low local node and energy consumption, and then increase the operation life-span of system.

Wherein, sampling instruction may include that each local user the adopting of required collection when performing low speed sampling The number of sampling point and each LUT equipment are used for generating the random seed sequence of sampling matrix, with As a example by the instruction that jth local user receives, its instruction received is: jth local user is performing low speed During sampling, the number of the sampled point of required collection is M_{R, j}, and LUT equipment is used for generating sampling The random seed sequence of matrix.

For the ease of user understand, LUT equipment is treated observation signal carry out low speed sampling carry out Illustrate, as a example by jth LUT equipment, its treat observation signal carry out low speed sampling obtained Sampled data be:

y_j=Φ_jx+n_j,j=1,...,J,

Wherein, x is sparse signal to be observed, by the vector representation of N × 1, Φ_jFor low speed sampling matrix, By a M_{R, j}The random matrix of × N represents (M_{R, j}< < N), the most alternative low speed is adopted Sample matrix includes: Gaussian matrix, shellfish make great efforts matrix, analog information transition matrix (Analog-to-Information Conversion, AIC) etc., matrix element can be according to the random seed sequence of base station transmission and according to a certain square It (if selecting Gaussian matrix, is then such as, according to Gauss that the battle array concrete distribution rule involved by type generates Distribution generates each element of low speed sampling matrix), n_jFor making an uproar at jth LUT equipment Sound, by a M_{R, j}The vector representation of × 1, y_jThe local sampling obtained by jth LUT equipment Data, by a M_{R, j}This sampled data is also reported by the vector representation of × 1 by each LUT equipment To base station.

For sparse signal x to be observed, its coefficient corresponding on one group of orthogonal basis of a certain transformation space to Amount is sparse, and therefore, sparse signal x to be observed can be expressed as:

x=Ψθ

Wherein, θ is that in the coefficient vector of N × 1, and this coefficient vector, number S of nonzero element is far smaller than letter Number vectorial dimension N(S < < N, the most openness), wherein, the size of S is the degree of rarefication of signal, is The unknown quantity that embodiment of the present invention technical scheme is to be estimated, Ψ is the rarefaction representation matrix (its each of N × N Column vector is i.e. referred to as one group of orthogonal basis of transformation space), the type of rarefaction representation matrix depends on concrete reality Border is applied, such as: openness due to signal in cognitive radio (Cognitive Radio, CR) system It is embodied in frequency domain so rarefaction representation matrix is a discrete Fourier transform (DFT) (Discrete Fourier Transform, DFT) matrix, in image is applied, usually a wavelet transform matrix or discrete cosine become Change (Discrete Cosine Transform, DCT) matrix, i.e. x and θ is one to one, is same signal Two kinds of expressions.

In S204, receive the sampled data that each local terminal reports, as the current iteration obtained Sampled point needed for step；

In the present embodiment, base station collect the sampled data that each LUT reports, for the most effectively Utilize whole sampled datas that system in multi-Step Iterations is obtained, collect the M that kth step obtains_rThis step system individual After sampled point, retain all M gathered in k step all k-1 step in the past the most simultaneously_ptIndividual sampled point.

In S205, according to institute before the sampled point needed for described current iteration step, current iteration step There are sampled point and multiple observation vectors of sparse signal to be observed that iterative step gathers, generate based on L1/L2 Mixing norm and multistep add up the object function of least mean-square error；

In the present embodiment, owing to the observed object of each LUT is same sparse signal, pass through Comprehensively utilize multiple LUT inherent joint sparse to multiple observation vectors of same sparse signal Property, and the sampling during multi-Step Iterations is additive, base station side set up based on L1/L2 mixing norm and The add up object function of least mean-square error of multistep is as follows:

$\widehat{\mathrm{\&Theta;}}=[{\widehat{\mathrm{\&theta;}}}_1,...,{\widehat{\mathrm{\&theta;}}}_J]=\underset{\mathrm{\&Theta;}}{\arg \min }\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\left(\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}{\left|{\mathrm{\&theta;}}_{n,j}\right|}^2\right)}^{1/2}+\mathrm{\&omega;}\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}{\left|\right|\left[\begin{array}{c}{Y}_{k-1,j}\\ y_{k,j}\end{array}\right]-\left[\begin{array}{c}{\mathrm{\&Omega;}}_{k-1,j}\\ {\mathrm{\&Phi;}}_{k,j}\end{array}\right]\mathrm{\&Psi;\&theta;}\left|\right|}_2^2---\left(1\right)$

Wherein, Y_{K-1, j}For jth local user all k-1 iteration steps before current kth time iterative step The cumulative vector of all sampled points gathered in rapid, Wherein T is transposition operation, Ω_k-1,jFor jth local user before current kth time iterative step in all k-1 iterative steps used by The accumulated matrix of sampling matrix, y_{K, j}For jth local user in current kth The sampling point vector gathered in secondary iterative step, Φ_{K, j}For jth local user at current kth time iterative step In used sampling matrix, ω is that compromise is openness and the weighted value of error, θ be N × 1 signal to be observed to Amount, wherein, N is the dimension of signal vector, and in this vector, the number of nonzero element is far smaller than signal vector Dimension (i.e. this signal to be observed has openness), Ψ is the rarefaction representation matrix of N × N, and J is local The number of user's (terminal), θ_jFor by each corresponding for signal θ to be observed observation vector, Θ is by each θ_jDo The matrix of column vector composition, and j ∈ [1 ..., J], θ_{N, j}For θ_jIn nth elements, For to be observed dilute Dredge the joint estimate of multiple observation vectors of signal.

In S206, according to described object function, sparse signal to be observed is carried out joint sparse degree estimation, And obtaining the result that joint sparse degree is estimated, described result includes: the degree of rarefication of sparse signal to be observed is estimated It is worth the positional information of the nonzero element corresponding with described degree of rarefication estimated value；

In the present embodiment, according to object function (1), sparse signal to be observed carried out joint sparse degree estimation, And obtain joint sparse degree estimate result particularly as follows:

${\hat{S}}_c=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}\left(\frac{1}{J}\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}\right|{\widehat{\mathrm{\&theta;}}}_j|\&GreaterEqual;\mathrm{\&lambda;})$

${\hat{S}\mathrm{upp}}_c=\left\{\right\{q_1,...,q_{{\hat{S}}_c}\left\}\right|(1/J){\mathrm{\&Sigma;}}_{j=1}^J{\widehat{\mathrm{\&theta;}}}_{q_i,j}\&GreaterEqual;\mathrm{\&lambda;},q_i\&Element;\{1,...,N\}\}$

Wherein, For the degree of rarefication estimated value of current sparse signal to be observed, non-in sparse signal the most to be observed The number of neutral element, For described degree of rarefication estimated value The positional information of corresponding nonzero element,For The location index that individual nonzero element is corresponding, λ is default degree of rarefication decision threshold, generally may be used Set according to nonzero element amplitude in sparse signal, such as, may be set to the half of nonzero element average.

In S207, it is judged that whether current degree of rarefication estimated result exception occurs, the most then perform S203, To carry out new joint sparse degree estimation, if it is not, then perform S208.

In the present embodiment, by judging whether current degree of rarefication estimated result exception occurs, if occurring abnormal, Then re-execute the sampling of current procedures, to carry out new joint sparse degree estimation once, improve degree of rarefication The robustness estimated.

In the present embodiment, when degree of rarefication estimated value Time, it is believed that current degree of rarefication estimated result occurs abnormal, Wherein η is that default degree of rarefication estimates abnormal thresholding, may be set to β N, and 0 < β < 1, N is vectorial dimension Degree, during because signal has openness, in vector, number S of nonzero element is far smaller than the dimension of signal vector Degree N(S < < N), when degree of rarefication estimated value occurs Time deviated from the openness principle of signal, So it is abnormal to be judged as that estimated result occurs.

In S208, the result estimated according to described joint sparse degree, it is judged that described joint sparse degree estimates knot Whether fruit meets terminates judgment condition, the most then perform S209, if it is not, then perform S202.

In S209, terminate joint sparse degree and estimate, and export degree of rarefication estimated result.

Optionally, S208 can one realize in the following ways:

Judge the positional information of nonzero element corresponding to current degree of rarefication estimated value and last degree of rarefication estimated value The positional information of corresponding nonzero element is the most consistent, i.e. Wherein, For the positional information of nonzero element corresponding to last degree of rarefication estimated value, Estimate for current degree of rarefication The positional information of the nonzero element that evaluation is corresponding.

The most then terminate joint sparse degree to estimate, and export degree of rarefication estimated result；

If it is not, then redirect the number performing to calculate the sampled point needed for current iteration step, to carry out the most once Joint sparse degree is estimated.

Optionally, S208 can also two realize in the following ways:

Judge that current degree of rarefication estimated value is the most equal, i.e. with last degree of rarefication estimated value Wherein,For current degree of rarefication estimated value, Last degree of rarefication estimated value；

If unequal, then redirect the number performing to calculate the sampled point needed for current iteration step, new to carry out One time joint sparse degree is estimated；

If it is equal, it is judged that the positional information of the nonzero element that current degree of rarefication estimated value is corresponding is sparse with the last time The positional information of the nonzero element that degree estimated value is corresponding is the most consistent, i.e.

If consistent, then terminate joint sparse degree and estimate, and export degree of rarefication estimated result；

If inconsistent, then redirect the number performing to calculate the sampled point needed for current iteration step, new to carry out One time joint sparse degree is estimated.

From both above termination judgement mode, judgment condition one " judges that current degree of rarefication estimated value is with upper One time degree of rarefication estimated value is the most equal " it is contained in judgment condition two and " judges that current degree of rarefication estimated value is corresponding The positional information of the nonzero element nonzero element corresponding with last degree of rarefication estimated value positional information whether Unanimously ", when i.e. judgment condition two is set up, judgment condition one is necessarily set up, but adjudicates when judgment condition one is set up Condition two is not necessarily set up, it is notable that two numerical value are only compared by judgment condition one, and sentence Certainly two ordered series of numbers need to be compared by condition two, it is achieved that to the judgement of condition two compared to realizing condition The judgement of one is more complicated, in actual applications, and can be according to the hardware computing resource of real system with to algorithm The implementation of the selecting factors both the above judgements such as execution efficiency requirement.

Embodiment three

Fig. 3 shows the flowchart of the method that the degree of rarefication that the embodiment of the present invention three provides estimates, describes in detail As follows:

In S301, the sampling instruction that the system that receives sends, described sampling instruction includes: terminal unit needs The number of sampled point to be obtained and terminal unit are for generating the random seed sequence of sampling matrix.

In the present embodiment, sampling instruction may include that each LUT equipment is performing low speed sampling The number of sampled point that Shi Suoxu gathers and each LUT equipment for generate sampling matrix with Machine subsequence, as a example by the instruction that jth local user receives, its instruction received is: jth is local User's number of the sampled point of required collection when performing low speed sampling is M_{R, j}, and LUT sets It is ready for use on the random seed sequence generating sampling matrix.

In S302, according to described sampling instruction, terminal unit performs sampling to obtain sampled data；

In the present embodiment, terminal unit receives described instruction of sampling, and according to described sampling instruction, scheduling is certainly The sampling resource of body to obtain sampled data, wherein sampling resource can be LUT equipment software, Hardware etc. may be used for the resource carrying out sampling.

In the present embodiment, the terminal unit of S302 performs sampling to obtain the process of sampled data and above-mentioned enforcement The execution process of step S203 in example two is similar to, and details see the description of above-described embodiment two.

In S303, by described sampled data reporting system.

In the present embodiment, each LUT equipment is merely responsible for local low speed sampling, and will adopt through low speed Sample obtain sampled data reporting system, by system carry out sampled point number calculate renewal, scheduling of resource, Degree of rarefication estimates, compare the operation such as judgement, therefore, it is possible to decrease the computation complexity of local node and energy consumption, And then increase the operation life-span of system.

Embodiment four

Fig. 4 shows the structure chart of the device that the degree of rarefication that the embodiment of the present invention four provides estimates, for the ease of Illustrate, illustrate only the part relevant to the embodiment of the present invention.

Described device includes: computing unit 41, transmitting element 42, receive unit 43, signal generating unit 44, Estimation unit 45 and decision unit 46.

Computing unit 41, for calculating the number of the sampled point needed for current iteration step；

Transmitting element 42, according to number and the preset schedule strategy of the sampled point of described calculating, sends sampling and refers to Order is to multiple subscriber terminal equipments, so that terminal unit is according to described sampling instruction, performs sampling and adopts to obtain Sample data, described sampling instruction includes: the number of the sampled point that terminal unit needs obtain and terminal unit are used In the random seed sequence generating sampling matrix；

Receive unit 43, for receiving the sampled data that each local terminal reports, current as obtain Sampled point needed for iterative step；

Signal generating unit 44, for according to the sampled point needed for described current iteration step, current iteration step it Before all iterative steps gather sampled point and multiple observation vectors of sparse signal to be observed, generate based on L1/L2 mixing norm and multistep add up the object function of least mean-square error；

Estimation unit 45, for according to described object function, carries out joint sparse degree to sparse signal to be observed Estimating, and obtain the result that joint sparse degree is estimated, described result includes: sparse signal to be observed sparse The positional information of the nonzero element that degree estimated value is corresponding with described degree of rarefication estimated value；

Decision unit 46, for the result estimated according to described joint sparse degree, estimates described joint sparse degree Meter carries out terminating judgement.

The degree of rarefication estimation unit that the embodiment of the present invention provides can use the embodiment of the method one in aforementioned correspondence In, details see the description of above-described embodiment one, do not repeat them here.

Embodiment five

Fig. 5 shows the structure chart of the device that the degree of rarefication that the embodiment of the present invention five provides estimates, for the ease of Illustrate, illustrate only the part relevant to the embodiment of the present invention.

Described device includes: computing unit 51, transmitting element 52, receive unit 53, signal generating unit 54, Estimation unit 55, abnormal deciding means 56 and decision unit 57.

Wherein, described preset schedule strategy is:

First scheduling strategy: needed for all intrasystem local user's mean allocation system current iteration step The number of sampled point；Or

Second scheduling strategy: according to the performance of intrasystem local user, currently change according to weight distribution system The number of the sampled point needed for riding instead of walk suddenly.

Optionally, described computing unit 51, specifically for the degree of rarefication estimated value according to a upper iterative step Number M of the sampled point that all iterative steps gather before current iteration step_pt, preset adjacent twice iteration The number smallest incremental Δ of the sampled point of step, calculates number M of sampled point needed for current iteration step_r, Specific as follows:

$M_r=\max \{M\left({\hat{S}}_p\right),M_{\mathrm{pt}}+\mathrm{\&Delta;}\}-M_{\mathrm{pt}}$

Wherein, max{, for taking maxima operation, M () is that degree of rarefication estimates that the number of required sampled point calculates Function, Wherein N is signal dimension, For the operation that rounds up.

Optionally, abnormal deciding means 56, it is used for judging whether current degree of rarefication estimated result exception occurs, When judging that described current degree of rarefication estimated result occurs abnormal, then start transmitting element 52, to carry out new one Secondary joint sparse degree is estimated, when judging that described current degree of rarefication estimated result does not occurs abnormal, starts judgement Unit 57.

Optionally, the nonzero element that described decision unit 57 is corresponding specifically for judging current degree of rarefication estimated value The positional information of the positional information nonzero element corresponding with last degree of rarefication estimated value the most consistent, if so, Then terminate joint sparse degree to estimate, and export degree of rarefication estimated result, if it is not, then start computing unit 51, To carry out new joint sparse degree estimation.

Optionally, described decision unit 57 is specifically for judging current degree of rarefication estimated value and last degree of rarefication Estimated value is the most equal；

If unequal, then start computing unit 51, to carry out new joint sparse degree estimation；

If it is equal, it is judged that the positional information of the nonzero element that current degree of rarefication estimated value is corresponding is sparse with the last time The positional information of the nonzero element that degree estimated value is corresponding is the most consistent；

If consistent, then terminate joint sparse degree and estimate, and export degree of rarefication estimated result；

If inconsistent, then start computing unit 51, to carry out new joint sparse degree estimation.

Optionally, described signal generating unit 54 is according to the sampled point needed for described current iteration step, current iteration Sampled point that before step, all iterative steps gather and multiple observation vectors of sparse signal to be observed, raw Become based on L1/L2 mixing norm and multistep add up least mean-square error object function particularly as follows:

$\widehat{\mathrm{\&Theta;}}=[{\widehat{\mathrm{\&theta;}}}_1,...,{\widehat{\mathrm{\&theta;}}}_J]=\underset{\mathrm{\&Theta;}}{\arg \min }\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\left(\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}{\left|{\mathrm{\&theta;}}_{n,j}\right|}^2\right)}^{1/2}+\mathrm{\&omega;}\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}{\left|\right|\left[\begin{array}{c}{Y}_{k-1,j}\\ y_{k,j}\end{array}\right]-\left[\begin{array}{c}{\mathrm{\&Omega;}}_{k-1,j}\\ {\mathrm{\&Phi;}}_{k,j}\end{array}\right]\mathrm{\&Psi;\&theta;}\left|\right|}_2^2$

Wherein, Y_{K-1, j}For jth local user all k-1 iteration steps before current kth time iterative step The cumulative vector of all sampled points gathered in rapid, Wherein T is transposition operation, Ω_k-1,jFor jth local user before current kth time iterative step in all k-1 iterative steps used by The accumulated matrix of sampling matrix, y_{K, j}For jth local user current The sampling point vector gathered in k iterative step, Φ_{K, j}For jth local user at current kth time iteration step Sampling matrix used in Zhou, ω is the openness weighted value with error of compromise, and θ is the signal to be observed of N × 1 In vector and this vector, the number of nonzero element is far smaller than dimension N of signal vector, and Ψ is the sparse of N × N Representing matrix, J is the number of local user, θ_jFor by each corresponding for signal θ to be observed observation vector, Θ For by each θ_jDo column vector composition matrix, j ∈ [1 ..., J], θ_n,jFor θ_jIn nth elements, Joint estimate for multiple observation vectors of sparse signal to be observed；

Described estimation unit 55, according to described object function, carries out joint sparse degree to sparse signal to be observed Estimate, and obtain joint sparse degree estimate result particularly as follows:

${\hat{S}}_c=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}\left(\frac{1}{J}\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}\right|{\widehat{\mathrm{\&theta;}}}_j|\&GreaterEqual;\mathrm{\&lambda;})$

${\hat{S}\mathrm{upp}}_c=\left\{\right\{q_1,...,q_{{\hat{S}}_c}\left\}\right|(1/J){\mathrm{\&Sigma;}}_{j=1}^J{\widehat{\mathrm{\&theta;}}}_{q_i,j}\&GreaterEqual;\mathrm{\&lambda;},q_i\&Element;\{1,...,N\}\}$

Wherein, For the degree of rarefication estimated value of current sparse signal to be observed, Estimate for described degree of rarefication Value The positional information of corresponding nonzero element, For The location index that individual nonzero element is corresponding, λ is Degree of rarefication decision threshold.

The degree of rarefication estimation unit that the embodiment of the present invention provides can use the embodiment of the method two in aforementioned correspondence In, details see the description of above-described embodiment two, do not repeat them here.

Embodiment six

Fig. 6 shows the structure chart of the device that the degree of rarefication that the embodiment of the present invention six provides estimates, for the ease of Illustrate, illustrate only the part relevant to the embodiment of the present invention.

Described device includes: receives unit 61, sampling unit 62 and reports unit 63.

Receiving unit 61, for receiving the sampling instruction that system sends, described sampling instruction includes: terminal sets The number of the standby sampled point needing to obtain and terminal unit are for generating the random seed sequence of sampling matrix；

Sampling unit 62, for according to described sampling instruction, terminal unit performs sampling to obtain sampled data；

Report unit 63, for by described sampled data reporting system.

The degree of rarefication estimation unit that the embodiment of the present invention provides can use the embodiment of the method three in aforementioned correspondence In, details see the description of above-described embodiment three, do not repeat them here.

It should be noted that in said apparatus embodiment, included unit is according to function logic Carry out dividing, but be not limited to above-mentioned division, as long as being capable of corresponding function；It addition, The specific name of each functional unit, also only to facilitate mutually distinguish, is not limited to the protection of the present invention Scope.

It addition, one of ordinary skill in the art will appreciate that the whole or portion realizing in the various embodiments described above method The program that can be by step by step completes to instruct relevant hardware, and corresponding program can be stored in a meter In calculation machine read/write memory medium, described storage medium, such as ROM/RAM, disk or CD etc..

The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all at this Any amendment, equivalent and the improvement etc. made within bright spirit and principle, should be included in the present invention Protection domain within.

Claims (14)

1. the method that a degree of rarefication is estimated, it is characterised in that described method includes:

The number of the sampled point needed for calculating current iteration step；

The number of the sampled point according to described calculating and preset schedule strategy, send sampling instruction to multiple terminals Equipment, so that terminal unit is according to described sampling instruction, performs to sample to obtain sampled data, described sampling Instruction includes: the number of the sampled point that terminal unit needs obtain and terminal unit are for generating sampling matrix Random seed sequence；

Receive the sampled data that each local terminal reports, as adopting needed for the current iteration step obtained Sampling point；

Adopt according to iterative steps all before the sampled point needed for described current iteration step, current iteration step Collection sampled point and multiple observation vectors of sparse signal to be observed, generate based on L1/L2 mixing norm and Multistep adds up the object function of least mean-square error；

According to described object function, sparse signal to be observed is carried out joint sparse degree estimation, and obtains associating The result that degree of rarefication is estimated, described result includes: the degree of rarefication estimated value of sparse signal to be observed and described dilute Dredge the positional information of nonzero element corresponding to degree estimated value；

The result estimated according to described joint sparse degree, estimates described joint sparse degree to carry out terminating judgement.

2. the method for claim 1, it is characterised in that described preset schedule strategy is:

First scheduling strategy: needed for all intrasystem local user's mean allocation system current iteration step The number of sampled point；Or

Second scheduling strategy: according to the performance of intrasystem local user, currently change according to weight distribution system The number of the sampled point needed for riding instead of walk suddenly.

3. the method for claim 1, it is characterised in that needed for described calculating current iteration step The number of sampled point particularly as follows:

Degree of rarefication estimated value according to a upper iterative stepBefore current iteration step, all iterative steps are adopted Number M of the sampled point of collection_pt, the number smallest incremental Δ of sampled point of adjacent twice iterative step preset, Number M of the sampled point needed for calculating current iteration step_r, specific as follows:

$M_r=max\{M\left({\hat{S}}_p\right),M_{pt}+\mathrm{\&Delta;}\}-M_{pt}$

Wherein, max{, for taking maxima operation, M () is that degree of rarefication estimates that the number of required sampled point calculates Function,Wherein N is signal dimension,For the operation that rounds up.

4. the method for claim 1, it is characterised in that described according to the estimation of described joint sparse degree Result, to described joint sparse degree estimate carry out terminate judgement before, described method also includes:

Judge whether current degree of rarefication estimated result exception occurs；

When judging that described current degree of rarefication estimated result occurs abnormal, redirect execution adopting according to described calculating The number of sampling point and preset schedule strategy, send sampling instruction extremely multiple subscriber terminal equipments, so that terminal sets Standby according to described sampling instruction, perform sampling to obtain sampled data, estimate carrying out a new joint sparse degree Meter；

When judging that described current degree of rarefication estimated result does not occurs abnormal, then perform according to described joint sparse The result that degree is estimated, estimates described joint sparse degree to carry out terminating judgement.

5. the method for claim 1, it is characterised in that described according to the estimation of described joint sparse degree Result, to described joint sparse degree estimate carry out terminate judgement particularly as follows:

Judge the positional information of nonzero element corresponding to current degree of rarefication estimated value and last degree of rarefication estimated value The positional information of corresponding nonzero element is the most consistent；

The most then terminate joint sparse degree to estimate, and export degree of rarefication estimated result；

If it is not, then redirect the number performing to calculate the sampled point needed for current iteration step, to carry out the most once Joint sparse degree is estimated.

6. the method for claim 1, it is characterised in that described according to the estimation of described joint sparse degree Result, to described joint sparse degree estimate carry out terminate judgement particularly as follows:

Judge that current degree of rarefication estimated value is the most equal with last degree of rarefication estimated value；

If unequal, then redirect the number performing to calculate the sampled point needed for current iteration step, new to carry out One time joint sparse degree is estimated；

If it is equal, it is judged that the positional information of the nonzero element that current degree of rarefication estimated value is corresponding is sparse with the last time The positional information of the nonzero element that degree estimated value is corresponding is the most consistent；

If consistent, then terminate joint sparse degree and estimate, and export degree of rarefication estimated result；

If inconsistent, then redirect the number calculating the sampled point needed for current iteration step, to carry out the most once Joint sparse degree is estimated.

7. the method for claim 1, it is characterised in that described according to described current iteration step institute The sampled point that needs, the sampled point that all iterative steps gather before current iteration step and sparse letter to be observed Number multiple observation vectors, generate and add up the target of least mean-square error based on L1/L2 mixing norm and multistep Function particularly as follows:

$\widehat{\mathrm{\&Theta;}}=\&lsqb;{\widehat{\mathrm{\&theta;}}}_1,...,{\widehat{\mathrm{\&theta;}}}_J\&rsqb;=\underset{\mathrm{\&Theta;}}{\arg \min }\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\left(\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}|{\mathrm{\&theta;}}_{n,j}{|}^2\right)}^{1/2}+\mathrm{\&omega;}\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}\left|\right|\left[\begin{array}{c}{Y}_{k-1,j}\\ y_{k,j}\end{array}\right]-\left[\begin{array}{c}{\mathrm{\&Omega;}}_{k-1,j}\\ {\mathrm{\&Phi;}}_{k,j}\end{array}\right]\mathrm{\&Psi;}\mathrm{\&theta;}|{|}_2^2$

Wherein, Y_k-1,jFor jth local user all k-1 iteration steps before current kth time iterative step The cumulative vector of all sampled points gathered in rapid,Wherein T is transposition operation, Ω_k-1,jFor jth local user before current kth time iterative step in all k-1 iterative steps used by The accumulated matrix of sampling matrix,y_k,jFor jth local user current The sampling point vector gathered in k iterative step, Φ_k,jFor jth local user at current kth time iteration step Sampling matrix used in Zhou, ω is the openness weighted value with error of compromise, and θ is the signal to be observed of N × 1 In vector and this vector, the number of nonzero element is far smaller than dimension N of signal vector, and Ψ is the sparse of N × N Representing matrix, J is the number of local user, θ_jFor by each corresponding for signal θ to be observed observation vector, Θ For by each θ_jDo column vector composition matrix, j ∈ [1 ..., J], θ_n,jFor θ_jIn nth elements, Joint estimate for multiple observation vectors of sparse signal to be observed；

Described according to described object function, sparse signal to be observed is carried out joint sparse degree estimation, and obtains Joint sparse degree estimate result particularly as follows:

${\hat{S}}_c=\underset{n=1}{\overset{N}{\&Sigma;}}\left(\frac{1}{J}\underset{j=1}{\overset{J}{\&Sigma;}}\right|{\widehat{\mathrm{\&theta;}}}_j|\&GreaterEqual;\mathrm{\&lambda;})$

$\hat{S}{\mathrm{upp}}_c=\left\{\right\{q_1,...,q_{{\hat{S}}_c}\left\}\right|(1/J){\mathrm{\&Sigma;}}_{j=1}^J{\widehat{\mathrm{\&theta;}}}_{q_i,j}\&GreaterEqual;\mathrm{\&lambda;},q_i\&Element;\{1,...,N\}\}$

Wherein,For the degree of rarefication estimated value of current sparse signal to be observed,Estimate for described degree of rarefication ValueThe positional information of corresponding nonzero element, q₁...,ForThe location index that individual nonzero element is corresponding, λ is Degree of rarefication decision threshold, describedRepresent the i-th unit of the jth correspondence observation vector treating observation signal The estimated value of element.

8. the device that a degree of rarefication is estimated, it is characterised in that described device includes:

Computing unit, for calculating the number of the sampled point needed for current iteration step；

Transmitting element, according to number and the preset schedule strategy of the sampled point of described calculating, sends sampling instruction To multiple terminal units, so that terminal unit is according to described sampling instruction, perform to sample to obtain sampled data, Described sampling instruction includes: number and the terminal unit of the sampled point that terminal unit needs obtain are adopted for generation The random seed sequence of sample matrix；

Receive unit, for receiving the sampled data that each local terminal reports, currently change as obtain Sampled point needed for riding instead of walk suddenly；

Signal generating unit, before according to the sampled point needed for described current iteration step, current iteration step All iterative steps gather sampled point and multiple observation vectors of sparse signal to be observed, generate based on L1/L2 mixing norm and multistep add up the object function of least mean-square error；

Estimation unit, for according to described object function, carries out joint sparse degree to sparse signal to be observed and estimates Meter, and obtain the result that joint sparse degree is estimated, described result includes: the degree of rarefication of sparse signal to be observed The positional information of the nonzero element that estimated value is corresponding with described degree of rarefication estimated value；

Decision unit, for the result estimated according to described joint sparse degree, estimates described joint sparse degree Carry out terminating judgement.

9. device as claimed in claim 8, it is characterised in that described preset schedule strategy is:

First scheduling strategy: needed for all intrasystem local user's mean allocation system current iteration step The number of sampled point；Or

Second scheduling strategy: according to the performance of intrasystem local user, currently change according to weight distribution system The number of the sampled point needed for riding instead of walk suddenly.

10. device as claimed in claim 8, it is characterised in that described computing unit, specifically for root Degree of rarefication estimated value according to a upper iterative stepBefore current iteration step, all iterative step collections adopts Number M of sampling point_pt, the number smallest incremental Δ of sampled point of adjacent twice iterative step preset, calculate and work as Number M of the sampled point needed for front iterative step_r, specific as follows:

$M_r=max\{M\left({\hat{S}}_p\right),M_{pt}+\mathrm{\&Delta;}\}-M_{pt}$

Wherein, max{, for taking maxima operation, M () is that degree of rarefication estimates that the number of required sampled point calculates Function,Wherein N is signal dimension,For the operation that rounds up.

11. devices as claimed in claim 8, it is characterised in that device also includes:

Abnormal deciding means, is used for judging whether current degree of rarefication estimated result exception occurs, described when judging When current degree of rarefication estimated result occurs abnormal, then start transmitting element, to carry out a new joint sparse degree Estimate, when judging that described current degree of rarefication estimated result does not occurs abnormal, start decision unit.

12. devices as claimed in claim 8, it is characterised in that described decision unit, specifically for sentencing The positional information of the nonzero element that disconnected current degree of rarefication estimated value is corresponding is corresponding with last degree of rarefication estimated value The positional information of nonzero element is the most consistent, the most then terminate joint sparse degree and estimate, and export degree of rarefication Estimated result, if it is not, then start computing unit, to carry out new joint sparse degree estimation.

13. devices as claimed in claim 8, it is characterised in that described decision unit, specifically for sentencing Disconnected current degree of rarefication estimated value is the most equal with last degree of rarefication estimated value；

If unequal, then start computing unit, to carry out new joint sparse degree estimation；

If it is equal, it is judged that the positional information of the nonzero element that current degree of rarefication estimated value is corresponding is sparse with the last time The positional information of the nonzero element that degree estimated value is corresponding is the most consistent；

If consistent, then terminate joint sparse degree and estimate, and export degree of rarefication estimated result；

If inconsistent, then start computing unit, to carry out new joint sparse degree estimation.

14. devices as claimed in claim 8, it is characterised in that described signal generating unit according to described currently The sampled point of all iterative step collections and treating before sampled point needed for iterative step, current iteration step Multiple observation vectors of observation sparse signal, generate and add up lowest mean square based on L1/L2 mixing norm and multistep The object function of error particularly as follows:

$\widehat{\mathrm{\&Theta;}}=\&lsqb;{\widehat{\mathrm{\&theta;}}}_1,...,{\widehat{\mathrm{\&theta;}}}_J\&rsqb;=\underset{\mathrm{\&Theta;}}{\arg \min }\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\left(\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}|{\mathrm{\&theta;}}_{n,j}{|}^2\right)}^{1/2}+\mathrm{\&omega;}\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}\left|\right|\left[\begin{array}{c}{Y}_{k-1,j}\\ y_{k,j}\end{array}\right]-\left[\begin{array}{c}{\mathrm{\&Omega;}}_{k-1,j}\\ {\mathrm{\&Phi;}}_{k,j}\end{array}\right]\mathrm{\&Psi;}\mathrm{\&theta;}|{|}_2^2$

Wherein, Y_k-1,jFor jth local user all k-1 iteration steps before current kth time iterative step The cumulative vector of all sampled points gathered in rapid,Wherein T is transposition operation, Ω_k-1,jFor jth local user before current kth time iterative step in all k-1 iterative steps used by The accumulated matrix of sampling matrix,y_k,jFor jth local user current The sampling point vector gathered in k iterative step, Φ_k,jFor jth local user at current kth time iteration step Sampling matrix used in Zhou, ω is the openness weighted value with error of compromise, and θ is the signal to be observed of N × 1 In vector and this vector, the number of nonzero element is far smaller than dimension N of signal vector, and Ψ is the sparse of N × N Representing matrix, J is the number of local user, θ_jFor by each corresponding for signal θ to be observed observation vector, Θ For by each θ_jDo column vector composition matrix, j ∈ [1 ..., J], θ_n,jFor θ_jIn nth elements, Joint estimate for multiple observation vectors of sparse signal to be observed；

Described estimation unit, according to described object function, carries out joint sparse degree to sparse signal to be observed and estimates Meter, and obtain joint sparse degree estimate result particularly as follows:

${\hat{S}}_c=\underset{n=1}{\overset{N}{\&Sigma;}}\left(\frac{1}{J}\underset{j=1}{\overset{J}{\&Sigma;}}\right|{\widehat{\mathrm{\&theta;}}}_j|\&GreaterEqual;\mathrm{\&lambda;})$

$\hat{S}{\mathrm{upp}}_c=\left\{\right\{q_1,...,q_{{\hat{S}}_c}\left\}\right|(1/J){\mathrm{\&Sigma;}}_{j=1}^J{\widehat{\mathrm{\&theta;}}}_{q_i,j}\&GreaterEqual;\mathrm{\&lambda;},q_i\&Element;\{1,...,N\}\}$

Wherein,For the degree of rarefication estimated value of current sparse signal to be observed,Estimate for described degree of rarefication ValueThe positional information of corresponding nonzero element, q₁...,ForThe location index that individual nonzero element is corresponding, λ is Degree of rarefication decision threshold, describedRepresent the i-th unit of the jth correspondence observation vector treating observation signal The estimated value of element.