CN102594515A - Node data transmitting method and device of sensor network and node data reconfiguring method and device of sensor network - Google Patents

Abstract

The invention discloses a node data transmitting method of a sensor network and relates to the technical field of a wireless sensor network, wherein the sensor network totally comprises nw network nodes. The method comprises the following steps of: S1: acquiring the node data xj of the nw network nodes in the sensor network; S2: compressing the node data xj of each network node by using a preset random compression matrix to obtain the node compression data tj transmitted by each network node every time, and carrying out transmission frequency compression on the network data formed by tj to obtain mw data, wherein i is equal to 1, 2,..., mw and represents i times of transmission, and wm is less than nw; and S3: transmitting the compression data of the nw network nodes to a fusion center through mw times of transmission. In the fusion center, the data xj originally acquired by each sensor is recovered by a node data reconfiguring method of the sensor network. According to the invention, the quantity of data transmitted and processed in the sensor network can be reduced.

Description

Sensor network nodes data transmission method for uplink and device, reconstructing method and device

Technical field

The present invention relates to the wireless sensor network technology field, particularly a kind of sensor network nodes data transmission method for uplink and device, reconstructing method and device.

Background technology

Wireless sensor network is widely used in highdensity radar array; Field such as fence, intelligent parking lot; In the face of complicated perception environment; Desirable wireless sensor network needs that sensor node has that stronger signal obtains, the wireless transmission capability of computing ability and high-energy storage area and efficient stable, satisfies these condition needs and pays huge software and hardware cost but be generally, and but may not obtain desirable effect.

Usually adopt the compression sensing technology to carry out the signal processing of sensor network in the prior art, Fig. 1 is the structural framing sketch map of sensor network system in the prior art, and is as shown in Figure 1; There is a fusion center (Fusion Center in the wireless sensor network system; Be called for short FC), comprise a large amount of wireless sensor nodes in the sensor network, be distributed in the physical environment that will observe; The final purpose node of wireless sensor node all is FC, and FC carries out the fusion and the reconstruct of data.

Supposing has n in the sensor network _wIndividual sensor node, the data of each node are x _j, j represents j network node, and the span of j does, j=1,2 ..., n _wWith the convergence of all nodes do together

Be called network data (network data), if interstitial content is bigger, network data X _wWill be very big, each node processing X _wAt least need n _wInferior communication, mobile and processing in network brings very big difficulty for data for this, but also concerns the safety problem of sensing data.

How not only quick but also realize that effectively the transmission of data and processing are the problems that the existing sensors network presses for solution.

Summary of the invention

The technical problem that (one) will solve

The technical problem that the present invention will solve is: how to reduce the transmission quantity of data in sensor network.

(2) technical scheme

For solving the problems of the technologies described above a kind of sensor network nodes data transmission method for uplink, may further comprise the steps:

S1: be total to n in the pick-up transducers network _wThe node data x of individual network node _j, wherein,

Wherein R representes real number, and the data that collect belong to real number field, n _pExpression node data x _jDimension, n _wNode number in the expression sensor network, j representes common n _wJ network node of individual network node;

S2: use the node data x of preset condensation matrix at random to each network node _jCarry out processed compressed, obtain the node packed data t of transmission each time of each network node _j, and to said t _jThe network data of forming

Carry out the compression of the number of transmissions, obtain m _wIndividual data

Wherein the i value is i=1,2 ..., m _w, represent i transmission, and m _w＜n _w

S3: simultaneously with n _wThe packed data of individual network node

Through m _wInferior transmission is sent to fusion center, and the data of fusion center reception each time are n _wThe m that individual node sends _wThe mixing of individual signal, each signal all are j=1,2 ..., n _wIndividual network node sends packed data

Blended data, be expressed as

Wherein subscript i is the i time transmission, and the span of i is i=1,2 ..., m _w, experience m altogether _wWhole network data t is accomplished in inferior transmission _wTransmission.

Wherein, said condensation matrix at random is the m that is generated at random as the seed of randomizer by the physical address of each network node _wIndividual m _p* n _pThe gaussian random matrix on rank:

Comprise two stochastic variables:

First stochastic variable is condensation matrix A at random _j, j=1,2 ..., n _w, A _jBe by the n that seed produced of each network node physical address separately as pseudorandom number generator _wIndividual m _p* n _pThe condensation matrix at random of dimension;

Second stochastic variable does

Be by the m of each network node physical address separately as the seed generation of pseudorandom number generator _wIndividual random number.

Wherein, step S2 specifically comprises:

Utilize condensation matrix A at random _jNode data x to each network node _jCarry out processed compressed, obtain the node packed data t of each network node _j=A _jX _j, wherein subscript j represents j network node, and the span of j is j=1,2 ..., n _w；

Utilize said second stochastic variable

Compress whole network data t _wThe number of transmissions, obtain the packed data of the i time of each network node transmission

Wherein, x _jThe node data of representing j network node, said node data x _jDimension be n _p

The present invention also provides a kind of sensor network nodes data reconstruction method, is used for the data that the above-mentioned sensor network nodes data transmission method for uplink of reconstruct sends, and may further comprise the steps:

A1: according to m _wThe data y that inferior fusion center receives ⁱSecond stochastic variable with said condensation matrix at random

The packed data t of the said sensor network of reconstruct _w

A2: according to the packed data t of said sensor network _wThe first stochastic variable A with said condensation matrix at random _j, use alternately restructing algorithm, the node data x of each network node in the said sensor network of reconstruct _j

Wherein, Fusion center is according to the prior physical address of each sensor network nodes of prevision, recovers said condensation matrix at random and generates said

by the physical address of each network node at random as the seed of randomizer

Wherein, said steps A 1 specifically comprises:

Fusion center receives m _wThe data y of inferior transmission ⁱ, the span of i is i=1,2 ..., m _wUse said gaussian random matrix j=1,2 ..., n _wApplication second stochastic variable

Recover network packed data t _w, the applied compression sensing algorithm minimizes mathematical operation l ₁Norm, application of formula t _w=argmin|t _w| ₁, constraints does

Recover network data $t_w={[t_1,t_2,...,t_{n_w}]}^T.$

Wherein, said steps A 2 specifically comprises:

A2.1: iteration for the first time, iterations iter=1, the remaining matrix R of initialization ^IterBe said network data t _w, like R ^Iter=t _w, the intermediate variable of preset computing approaches matrix

Be m _p* n _wFull null matrix;

A2.2: the iter time iteration, remaining matrix R is calculated in iterations iter＞1 ^IterWith the said first stochastic variable A of condensation matrix at random _jIn maximum row of the degree of correlation (notion of the degree of correlation is the coefficient sum that two vectors are done inner product) number: ind ^Iter

${\mathrm{ind}}^{\mathrm{iter}}=\arg \underset{\mathrm{ind}=\mathrm{1,2},...n_p}{\max }\underset{k=1}{\overset{n_p}{\mathrm{\&Sigma;}}}\frac{<r_k^{\mathrm{iter}-1}\&CenterDot;a_j^{\mathrm{ind}}>}{{\left|\right|a_j^{\mathrm{ind}}\left|\right|}_2^2}$

Wherein,

Represent remaining matrix R in the last iteration ^IterIn k row,

Represent the said first stochastic variable A of condensation matrix at random _jInd row; L in the expression mathematics ₂Norm; K representes the intermediate variable of computing; Symbol

Be illustrated in all variable i nd=1,2..., n _pValue in, can obtain maximum

Value, symbol

Expression

With

Inner product;

A2.3: make set omega ^Iter=Ω ^Iter∪ ind ^Iter, symbol ∪ representes to get two union of sets collection; Form m _pThe matrix on * iter rank

Promptly from said matrix A _jIn choose Ω ^IterThe submatrix that the row of set representative are formed, the initial data of each node of estimating with the method that minimizes mean square deviation: Iteration obtains to comprise the vector of iter * 1 element each time

Order

Obtain the matrix that approaches of iteration each time

Remaining matrix

A2.4: if iter＜100, then iter=iter+1 promptly continues next iteration, repeating step A2.2; If iter=100 then stops iteration;

A2.5: data output node estimated value

and then network data

application of formula

that obtains estimating are calculated mean square deviation; If mean square deviation is less than said threshold value; Execution in step A2.6; If mean square deviation is not less than said threshold value, execution in step A2.7;

A2.6: utilize said condensation matrix at random

to multiply each other, carry out all restructing algorithms more again with corresponding estimated data

;

A2.7: data output node

Wherein, said threshold value is 10 ^-4

The present invention also provides a kind of sensor network nodes data sending device, comprising:

Data acquisition module is used for pick-up transducers network n altogether _wThe node data x of individual network node _j, wherein,

Wherein R representes real number, and the data that collect belong to real number field, n _pExpression node data x _jDimension, n _wNode number in the expression sensor network, j representes common n _wJ network node of individual network node;

Data compressing module is used to use the node data x of preset condensation matrix at random to each network node _jCarry out processed compressed, obtain the node packed data t of transmission each time of each network node _j, and to said t _jThe network data of forming

Carry out the compression of the number of transmissions, obtain m _wIndividual data

Wherein the i value is i=1,2 ..., m _w, represent i transmission, and m _w＜n _w

Data transmission blocks is used for simultaneously with n _wThe packed data of individual network node

Through m _wInferior transmission is sent to fusion center, and the data of fusion center reception each time are n _wThe m that individual node sends _wThe mixing of individual signal, each signal all are j=1,2 ..., n _wIndividual network node sends packed data

Blended data, be expressed as

Wherein subscript i is the i time transmission, and the span of i is i=1,2 ..., m _w, experience m altogether _wWhole network data t is accomplished in inferior transmission _wTransmission.

The present invention also provides a kind of sensor network nodes data reconstruction device, is used for the data that the above-mentioned sensor network nodes data sending device of reconstruct sends, and comprising:

Send the number of times reconstructed module, be used for according to m _wThe data y that inferior fusion center receives ⁱSecond stochastic variable with said condensation matrix at random

The packed data t of the said sensor network of reconstruct _w

Send the data reconstruction module, be used for packed data t according to said sensor network _wThe first stochastic variable A with said condensation matrix at random _j, use alternately restructing algorithm, the node data x of each network node in the said sensor network of reconstruct _j

(3) beneficial effect

The present invention is through carrying out the compression of data volume and data transmission times to the node data of sensor network, thereby reduces the transmission quantity of data in sensor network.

Description of drawings

Fig. 1 is the structural representation of sensor network;

Fig. 2 is a kind of sensor network nodes data transmission method for uplink flow chart of the embodiment of the invention;

Fig. 3 is the rarefaction representation curve chart of network data in Fig. 2 method;

Fig. 4 is the sketch map of the data compression of step S2 among Fig. 2;

Fig. 5 is a kind of sensor network nodes data reconstruction method flow diagram of the embodiment of the invention.

Embodiment

Compression sensing (Compressed Sensing is called for short CS) is a kind of new sensing or Sampling techniques, has very big difference, CS theory to show that the collection quantity of signal can be far smaller than traditional sampled data with the traditional data sampling.The realization of CS is based on two criterions, sparse property and incoherence.Sparse property is meant that " information rate " of continuous time signal possibly be far smaller than its shared bandwidth, and perhaps the quantity of the discrete-time signal degree of freedom is far smaller than the length of signal.More precisely, CS is based on a fact, and natural signal representes to be sparse or compressible down in suitable substrate, for example discrete cosine transform, Fourier transform etc. through after some conversion.Incoherence has been represented a kind of antithesis characteristic, is similar to frequency and time relation.That is to say that sampled signal ground waveform under particular substrate is very intensive.Generally, most signals can find at the bottom of the suitable transform-based and be expressed as sparse form, like voice signal, picture signal and some measurement data etc.; Under these substrates; The factor quantity of expression signal seldom, therefore, CS can well realize compressed encoding.To degree of rarefication is the signal of K, need not sample N time (supposing that N is a signal length), and only need c * K uncorrelated measuring-signal can realize compression; Wherein c is a constant, and value is 2～4, has replaced original method that K sparse signal is adopted N time; K＜＜N; Therefore, transducer can only send low volume data to merging node, merges node and still can also in any form these signals be handled by these signals of reconstruct.

The node data sending method of the sensor network that Fig. 2 provides for the embodiment of the invention one, as shown in Figure 2, comprising:

Step 201, in sensor network, the node data of each network node is obtained in sampling, each node data is sparse under same substrate.

N is arranged in the sensor network _wIndividual node, each network node passes through n _wInferior sampling obtains the node data of each network node

Wherein, R representes real number, n _pExpression node data x _jDimension, then the node data of each network node is n _p* 1 vector, owing to have correlation between the node data of each network node, the node data of supposing each network node is at same substrate ψ _mSparse down.After the all-network node is all finished sparse processing in the sensor network, can obtain network data

Step 202, use condensation matrix at random the node data of each network node and the number of times of each network node transmission data are carried out processed compressed; Obtain the node packed data of transmission each time of each network node, the physical address that said condensation matrix at random is each network node generates.

In the present embodiment, each sensor node j utilizes local mac address to produce random matrix A as the seed of pseudorandom number generator _jAs long as know the MAC Address of each node, FC also can obtain all n easily _wThe A of condensation matrix at random of individual node _jCompression process is to use random matrix A _jData x with this node collection _jMultiply each other and obtain t _j, t _j=A _jX _j, t _jBe that dimension is m _pVector, comprise m _pIndividual component, wherein m _p＜n _pThereby, realized the compression of each network node data, promptly from n _pThe data x of dimension _jBoil down to m _pThe data t of dimension _jWith n _wThe packed data t of individual node _jBe combined as the form of matrix, just constitute the network packed data, for

T wherein _j=A _jX _j, A _jBe that dimension is m _p* n _p, x _jDimension be n _p, back t multiplies each other _jDimension be m _p

Whole network data spatially is correlated with; If find and the corresponding description node of network topology between the substrate of data dependence, can rarefaction representation network data

so network data also can utilize the method compression of compression sensing.Network data t _wUnder the substrate of figure wavelet field, can obtain rarefaction representation.

For example, the figure wavelet coefficient is ψ _{I, j}, be expressed as

Wherein

Be normalization factor, N ' _h(v _j) expression node v _jIn the network topology, ring domain node data.The figure wavelet transformation representes in the network topology, the difference between the mean value of ring domain node data, reflection strictly according to the facts the correlation in node data space, satisfy network data t _wSparse property on the space of figure small echo.Fig. 3 has shown network data t _wSparse property in the figure wavelet field.Wherein dotted line is an initial data, and solid line is figure small echo and the data after the conversion, and present embodiment selects Mexican Hat as wavelet coefficient.

Further, for each network node, be seed with the physical address of this network node, adopt a pseudo-random generator to generate one group of pseudo random number and form,

Number of times to each network node transmission data carries out processed compressed, that is to say, uses

Can be with network data t _wThe number of times of transmission is by n _wInferior boil down to m _wInferior; In the present embodiment, will at each node

Node packed data t with each network node _jMultiply each other, obtain the data of the i time transmission $t_j^i={\mathrm{\&alpha;}}_j^i{t}_j={\mathrm{\&alpha;}}_j^i{A}_j\&CenterDot;x_j.$

Step 203, with each network node each time the transmission each network node physical address corresponding of node packed data

send to fusion center.

Each the network node physical address corresponding of node packed data

of transmission each time that is specially each network node sends to fusion center.Said process as shown in Figure 4.

Correspondingly, this network node corresponding nodes packed data of each network node in the sensor network passes through m _wInferior transmission is sent to fusion center, the initial data that fusion center is gathered according to the data reconstruction that receives.

Each network node and fusion center all are furnished with an omnidirectional antenna; And the transmission channel between each network node and the fusion center is an arrowband additive white Gaussian noise access channel; That is to say; The node packed data of transmission each time of each network node all can receive the pollution of additive white Gaussian noise, is expressed as Z wherein _jFor zero-mean, the variance of each network node does

Additive white Gaussian noise.Sensor network all is a wireless environment simultaneously; Also need consider the influence of channel factors to data reconstruct; Because sensor network all is in observing environment, to lay in advance to accomplish, so each sensor node knows the distance of self and fusion center, thereby can estimate channel factors h _j, then when each transmission, the influence that can remove channel factors in advance, promptly

The embodiment of the invention adopts condensation matrix at random that the data and the number of transmissions of network node are compressed; Can solve in the existing sensors network can't be not only fast but also realize the problem of the transmission and the processing of whole network image data effectively; Simultaneously; Reduce the data volume of the transmission and the processing of network node, prolonged the life-span of sensor network nodes, weakened the effect in " energy cavity ".

Fig. 5 is the schematic flow sheet of the data reconstruction method of fusion center provided by the invention; In data transmission procedure, node packed data and physical address corresponding that each network node in the said sensor network is transmitted each time send to fusion center, correspondingly; Fusion center can adopt method shown in Figure 5 to carry out reconstruct to the data that receive; Thereby obtain the node data of each network node in the said sensor network, as shown in Figure 5, said method specifically comprises:

Step 501, fusion center obtain the condensation matrix at random of said sensor network according to each the network node corresponding physical layer address that receives.

For instance, fusion center obtains the condensation matrix at random of said sensor network according to each the network node corresponding physical layer address that receives

Wherein subscript i is the i time transmission, and the span of i is i=1,2 ..., m _wSubscript j represents j network node, and the span of j is j=1, and 2 ..., n _w

Fusion center can be according to each network node physical address corresponding in the communication protocol receiving sensor network that is provided with in advance.Wherein, Fusion center can be told each network node physical address corresponding automatically according to communication protocol; Each network node physical address corresponding to receive is a seed; Be used to generate the identical pseudo-random generator of pseudo-random generator that condensation matrix adopted at random when adopting with node data that compresses each network node and node data the number of transmissions; The identical condensation matrix at random of condensation matrix at random that adopts when generating with node data that compresses each network node and node data the number of transmissions specifically comprises:

With each network node corresponding physical layer address is seed, adopts to generate at random that the used pseudo-random generator of condensation matrix generates the condensation matrix at random

that each network node transmits application each time

The condensation matrix at random that each network node is transmitted each time application combines, and forms the condensation matrix at random of said sensor network:

${\left\{{\mathrm{\&alpha;}}_j^i{A}_j\right\}}_{i=1}^{m_w}=\left[\begin{array}{cccc}{A}_1^1& A_2^1& ...& A_{n_w}^1\\ A_1^2& A_2^2& ...& A_{n_w}^2\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ A_1^{m_w}& A_2^{m_w}& ...& A_{n_w}^{m_w}\end{array}\right],j=\mathrm{1,2},...,n_w,A_j\&Element;R^{m_p\&times;n_p}.$

Step 502, according to the node packed data that each network node of receiving transmits each time, obtain the packed data of said sensor network.

For instance, the node packed data that transmits each time according to each network node that receives

obtains the packed data of said sensor network.

For instance, said sensor network adopts analogue transmission mechanism, at m _wIn the inferior transmission, the node amount of compressed data that successively each network node is transmitted each time sends to fusion center, in transmission each time, and n _wIndividual node sends the corresponding nodes packed data simultaneously, and mixes at fusion center.

Be that the data that fusion center receives each time are:

Fusion center m _wThe inferior data that receive are:

$y_w={[y^1,y^2,...,y^{m_w}]}^T=\left[\begin{array}{cccc}{a}_1^1& a_2^1& ...& a_{n_w}^1\\ a_1^2& a_2^2& ...& a_{n_w}^2\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ a_1^{m_w}& a_2^{m_w}& ...& a_{n_w}^{m_w}\end{array}\right]\left[\begin{array}{c}{t}_1\\ t_2\\ .\\ .\\ .\\ t_{n_w}\end{array}\right].$

Further for instance, each network node all is furnished with an omnidirectional antenna with merging, and the channel between each network node and the fusion center is arrowband additive white Gaussian noise access channel, and the Gaussian noise variance of fusion center does And fusion center known with each network node apart from d _j, according to h _j=1/d _j ²Can calculate the large scale fading factor h of each network node corresponding transmission channel _j

Suppose that said sensor network and fusion center are synchronous, specifically comprise: 1, carrier synchronization, promptly each network node all is equipped with local synclator reception carrier frequency; 2, time synchronized is promptly used for each time transmission channel, and relevant timing error all is far smaller than signal transmission time Tc; 3, Phase synchronization, promptly transmission channel reaches fusion center based on the relevant mode of phase place.

Under the synchronous hypothesis of said sensor network and fusion center, use arrowband additive white Gaussian noise access channel, accordingly, the data that fusion center receives each time are:

$y^i=\underset{j=1}{\overset{n_w}{\mathrm{\&Sigma;}}}\sqrt{h_j}{t}_j^i+z_0=\underset{j=1}{\overset{n_w}{\mathrm{\&Sigma;}}}\frac{1}{h_j}{A}_j^i(x_j+z_j)+z_0$

z ₀Zero-mean, the variance of expression fusion center do

Additive white Gaussian noise.Accordingly, fusion center is used the decoding function:

Accordingly, use second condensation matrix at random

Recover the node packed data of each network node, obtain the packed data of said sensor network, the applied compression sensing algorithm minimizes mathematical operation l ₁Norm, application of formula t _w=argmin|t _w| ₁, constraints does

Recover network data

Step 503, according to the packed data of said sensor network and the condensation matrix at random of said sensor network, use multinode and replace restructing algorithm, the node data of each network node in the said sensor network of reconstruct.Be specially:

Utilization obtains the packed data t of said sensor network _wThe node data x of each network node of reconstruct _jThat is input m, _p* n _wThe observing matrix on rank

M wherein _pBe observation signal length, n _wFor multinode replaces the node number in the restructing algorithm, the node number n in the promptly said sensor network _wInput m _p* n _wThe first condensation matrix A at random on rank _j, j=1,2 ..., n _w, n wherein _pBe signal x _jDimension; Output n _p* n _wThe matrix x on rank _j, j=1,2 ..., n _wMiddle output variable is: m _p* n _wRank approach matrix

Output m _p* n _wThe remaining matrix R on rank; Output index set omega _Ind, and point out that the row number of non-0 element distribute.

Multinode replaces restructing algorithm and specifically describes as follows:

Step 1, iteration: Iter=1 for the first time, initialization R ₀=t _w, and

Full null matrix;

Step 2, the iter time iteration: calculate the maximum row of the degree of correlation:

${\mathrm{ind}}^{\mathrm{iter}}=\arg \underset{\mathrm{ind}=\mathrm{1,2},...n_p}{\max }\underset{k=1}{\overset{n_p}{\mathrm{\&Sigma;}}}\frac{<r_k^{\mathrm{iter}-1}\&CenterDot;a_j^{\mathrm{ind}}>}{{\left|\right|a_j^{\mathrm{ind}}\left|\right|}_2^2}$

Wherein,

Represent remaining matrix R in the last iteration ^IterIn k row,

Represent the said first stochastic variable A of condensation matrix at random _jInd row;

L in the expression mathematics ₂Norm; K representes the intermediate variable of computing; Symbol

Be illustrated in all variable i nd=1,2..., n _pValue in, can obtain maximum

Value, symbol

Expression

With

Inner product;

Step 3, make set omega ^Iter=Ω ^Iter∪ ind ^Iter, form m _pThe matrix of * iter

Promptly from A _jIn choose Ω ^IterThe submatrix that the row of set representative are formed;

Use the initial data of each node of the method estimation that minimizes mean square deviation:

Iteration obtains to comprise the vector of iter * 1 element each time

Order

Obtain the matrix that approaches of iteration each time

Remaining matrix $R^{\mathrm{Iter}}=t_w-{\hat{t}}^{\mathrm{Iter}};$

Step 4, if iter＜100, then iter=iter+1 promptly continues next iteration, repeating step 2; If iter=100 then stops iteration;

Step 5, data output node estimated value

The data combination of each node is become whole network data:

Application of formula

Calculate mean square deviation, if mean square deviation is less than threshold value 10 ^-4, execution in step 6 is if mean square deviation is not less than this threshold value, execution in step 7;

Step 6, general be condensation matrix at random

Act on the node data estimated value

Obtain Again recovery nodes packed data t again _w, turn back to step 1;

Step 7: data output node then

Be the data x of node in the corresponding sensor network _jEstimated value.

Need explanation to be, above-mentioned threshold value can specifically define according to concrete application, and the threshold value of present embodiment is preferably 10 ^-4

The fusion center of present embodiment is deciphered processing through the decoding function to the data that receive; Physical address according to each network node that receives recovers condensation matrix at random; The node packed data that transmits each time according to each network node that receives recovers the packed data of sensor network; Further use multinode combined reconstruction algorithm progressively the iterative approach True Data can realize that to obtain the node data of sensor network fusion center not only fast but also carry out the high accuracy reconstruct of data effectively.

The present invention also provides a kind of sensor network nodes data sending device, comprising:

Data acquisition module is used for pick-up transducers network n altogether _wThe node data x of individual network node _j, wherein,

Wherein R representes real number, and the data that collect belong to real number field, n _pExpression node data x _jDimension, n _wNode number in the expression sensor network, j representes common n _wJ network node of individual network node;

Data compressing module is used to use the node data x of preset condensation matrix at random to each network node _jCarry out processed compressed, obtain the node packed data t of transmission each time of each network node _j, and to said t _jThe network data of forming

Carry out the compression of the number of transmissions, obtain m _wIndividual data

Wherein the i value is i=1,2 ..., m _w, represent i transmission, and m _w＜n _w

Data transmission blocks is used for simultaneously with n _wThe packed data of individual network node

Through m _wInferior transmission is sent to fusion center, and the data of fusion center reception each time are n _wThe m that individual node sends _wThe mixing of individual signal, each signal all are j=1,2 ..., n _wIndividual network node sends packed data

Blended data, be expressed as

Wherein subscript i is the i time transmission, and the span of i is i=1,2 ..., m _w, experience m altogether _wWhole network data t is accomplished in inferior transmission _wTransmission.

Mode such as above-mentioned data transmission method for uplink that this device sends the sensor network nodes data are said, repeat no more here.

The present invention also provides a kind of sensor network nodes data reconstruction device, is used for the data that the above-mentioned sensor network nodes data sending device of reconstruct sends, and this reconfiguration device is integrated on the fusion center, comprising:

Send the number of times reconstructed module, be used for according to m _wThe data y that inferior fusion center receives ⁱSecond stochastic variable with said condensation matrix at random

The packed data t of the said sensor network of reconstruct _w

Send the data reconstruction module, be used for packed data t according to said sensor network _wThe first stochastic variable A with said condensation matrix at random _j, use alternately restructing algorithm, the node data x of each network node in the said sensor network of reconstruct _j

The mode such as the above-mentioned data reconstruction method of these device reconstruct sensor network nodes data are said, repeat no more here.

Above execution mode only is used to explain the present invention; And be not limitation of the present invention; The those of ordinary skill in relevant technologies field under the situation that does not break away from the spirit and scope of the present invention, can also be made various variations and modification; Therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.

Claims (10)

1. a sensor network nodes data transmission method for uplink is characterized in that, may further comprise the steps:

S1: be total to n in the pick-up transducers network _wThe node data x of individual network node _j, wherein,

Wherein R representes real number, and the data that collect belong to real number field, n _pExpression node data x _jDimension, n _wNode number in the expression sensor network, j representes common n _wJ network node of individual network node;

S2: use the node data x of preset condensation matrix at random to each network node _jCarry out processed compressed, obtain the node packed data t of transmission each time of each network node _j, and to said t _jThe network data of forming Carry out the compression of the number of transmissions, obtain m _wIndividual data

Wherein the i value is i=1,2 ..., m _w, represent i transmission, and m _w＜n _w

S3: simultaneously with n _wThe packed data of individual network node

Through m _wInferior transmission is sent to fusion center, and the data of fusion center reception each time are n _wThe m that individual node sends _wThe mixing of individual signal, each signal all are j=1,2 ..., n _wIndividual network node sends packed data Blended data, be expressed as

Wherein subscript i is the i time transmission, and the span of i is i=1,2 ..., m _w, experience m altogether _wWhole network data t is accomplished in inferior transmission _wTransmission.

2. sensor network nodes data transmission method for uplink as claimed in claim 1 is characterized in that, said condensation matrix at random is the m that is generated at random as the seed of randomizer by the physical address of each network node _wIndividual m _p* n _pThe gaussian random matrix on rank:

Comprise two stochastic variables:

First stochastic variable is condensation matrix A at random _j, j=1,2 ..., n _w, A _jBe by the n that seed produced of each network node physical address separately as pseudorandom number generator _wIndividual m _p* n _pThe condensation matrix at random of dimension;

Second stochastic variable does

Be by the m of each network node physical address separately as the seed generation of pseudorandom number generator _wIndividual random number.

3. sensor network nodes data transmission method for uplink as claimed in claim 2 is characterized in that step S2 specifically comprises:

Utilize condensation matrix A at random _jNode data x to each network node _jCarry out processed compressed, obtain the node packed data t of each network node _j=A _jX _j, wherein subscript j represents j network node, and the span of j is j=1,2 ..., n _w

Utilize said second stochastic variable

Compress whole network data t _wThe number of transmissions, obtain the packed data of the i time of each network node transmission Wherein, x _jThe node data of representing j network node, said node data x _jDimension be n _p

4. a sensor network nodes data reconstruction method is characterized in that, is used for the data that each described sensor network nodes data transmission method for uplink of reconstruct such as claim 1～3 sends, and may further comprise the steps:

A1: according to m _wThe data y that inferior fusion center receives ⁱSecond stochastic variable with said condensation matrix at random

The packed data t of the said sensor network of reconstruct _w

A2: according to the packed data t of said sensor network _wThe first stochastic variable A with said condensation matrix at random _j, use alternately restructing algorithm, the node data x of each network node in the said sensor network of reconstruct _j

5. sensor network nodes data reconstruction method as claimed in claim 4; It is characterized in that; Fusion center is according to the prior physical address of each sensor network nodes of prevision, recovers said condensation matrix at random and generates said by the physical address of each network node at random as the seed of randomizer

6. sensor network nodes data reconstruction method as claimed in claim 5 is characterized in that, said steps A 1 specifically comprises:

Fusion center receives m _wThe data y of inferior transmission ⁱ, the span of i is i=1,2 ..., m _w

Use said gaussian random matrix j=1,2 ..., n _wApplication second stochastic variable

Recover network packed data t _w, the applied compression sensing algorithm minimizes mathematical operation l ₁Norm, application of formula t _w=argmin|t _w| ₁, constraints does

Recover network data $t_w={[t_1,t_2,...,t_{n_w}]}^T.$

7. sensor network nodes data reconstruction method as claimed in claim 5 is characterized in that, said steps A 2 specifically comprises:

A2.1: iteration for the first time, iterations iter=1, the remaining matrix R of initialization ^IterBe said network data t _w, like R ^Iter=t _w, the intermediate variable of preset computing approaches matrix

Full null matrix;

A2.2: the iter time iteration, remaining matrix R is calculated in iterations iter＞1 ^IterWith the said first stochastic variable A of condensation matrix at random _jIn maximum row of the degree of correlation number: ind ^Iter:

${\mathrm{ind}}^{\mathrm{iter}}=\arg \underset{\mathrm{ind}=\mathrm{1,2},...n_p}{\max }\underset{k=1}{\overset{n_p}{\mathrm{\&Sigma;}}}\frac{<r_k^{\mathrm{iter}-1}\&CenterDot;a_j^{\mathrm{ind}}>}{{\left|\right|a_j^{\mathrm{ind}}\left|\right|}_2^2},$

Wherein,

Represent remaining matrix R in the last iteration ^IterIn k row,

Represent the said first stochastic variable A of condensation matrix at random _jInd row;

L in the expression mathematics ₂Norm; K representes the intermediate variable of computing; Symbol

Be illustrated in all variable i nd=1,2..., n _pValue in, can obtain maximum

Value, symbol

Expression

With

Inner product;

A2.3: make set omega ^Iter=Ω ^Iter∪ ind ^Iter, symbol ∪ representes to get two union of sets collection; Form m _pThe matrix on * iter rank

Promptly from said matrix A _jIn choose Ω ^IterThe submatrix that the row of set representative are formed, the initial data of each node of estimating with the method that minimizes mean square deviation: Iteration obtains to comprise the vector of iter * 1 element each time

Order

Obtain the matrix that approaches of iteration each time

Remaining matrix

A2.4: if iter＜100, then iter=iter+1 promptly continues next iteration, repeating step A2.2; If iter=100 then stops iteration;

A2.5: data output node estimated value

and then network data

application of formula

that obtains estimating are calculated mean square deviation; If mean square deviation is less than said threshold value; Execution in step A2.6; If mean square deviation is not less than said threshold value, execution in step A2.7;

A2.6: utilize said condensation matrix at random

to multiply each other, carry out all restructing algorithms more again with corresponding estimated data

;

A2.7: data output node

8. sensor network nodes data transmission method for uplink as claimed in claim 7 is characterized in that, said threshold value is 10 ^-4

9. a sensor network nodes data sending device is characterized in that, comprising:

Data acquisition module is used for pick-up transducers network n altogether _wThe node data x of individual network node _j, wherein,

Wherein R representes real number, and the data that collect belong to real number field, n _pExpression node data x _jDimension, n _wNode number in the expression sensor network, j representes common n _wJ network node of individual network node;

Data compressing module is used to use the node data x of preset condensation matrix at random to each network node _jCarry out processed compressed, obtain the node packed data t of transmission each time of each network node _j, and to said t _jThe network data of forming

Carry out the compression of the number of transmissions, obtain m _wIndividual data

Wherein the i value is i=1,2 ..., m _w, represent i transmission, and m _w＜n _w

Data transmission blocks is used for simultaneously with n _wThe packed data of individual network node

Through m _wInferior transmission is sent to fusion center, and the data of fusion center reception each time are n _wThe m that individual node sends _wThe mixing of individual signal, each signal all are j=1,2 ..., n _wIndividual network node sends packed data

Blended data, be expressed as

Wherein subscript i is the i time transmission, and the span of i is i=1,2 ..., m _w, experience m altogether _wWhole network data t is accomplished in inferior transmission _wTransmission.

10. a sensor network nodes data reconstruction device is characterized in that, is used for the data that reconstruct sensor network nodes data sending device as claimed in claim 9 sends, and comprising:

Send the number of times reconstructed module, be used for according to m _wThe data y that inferior fusion center receives ⁱSecond stochastic variable with said condensation matrix at random

The packed data t of the said sensor network of reconstruct _w

Send the data reconstruction module, be used for packed data t according to said sensor network _wThe first stochastic variable A with said condensation matrix at random _j, use alternately restructing algorithm, the node data x of each network node in the said sensor network of reconstruct _j

Images