CN104601264A - Multi-antenna spectrum sensing method applicable to high-dimension finite sample conditions - Google Patents

Abstract

The invention relates to a multi-antenna spectrum sensing method applicable to high-dimension finite sample conditions. By adopting relevance between multi-antenna receiving signal components to structure sensing decision and sensing decision threshold based on the random stochastic matrix, the multi-antenna spectrum sensing method includes firstly, continuously sampling multi-antenna receiving signals to form a receiving signal data matrix X; then, calculating relevance measurement indicators among the multi-antenna receiving signal components on this basis, and calculating to obtain sensing decision I; secondly, calculating sensing decision threshold t on the basis of the random stochastic matrix; finally, implementing sensing decision, to be specifically, judging that no spectrum hole exists when the sensing decision I is larger than the preset threshold t, or otherwise, judging that a spectrum hole exists. The multi-antenna spectrum sensing method has the advantages that the method is simple and low in calculation complexity in sensing application of the high-dimension finite sample capacity, efficient total blindness detection under the condition of deficiency in statistical information of master user signals, wireless channels and noise can be realized, and sensing results are reliable and the like.

Description

A kind of multiple antennas frequency spectrum sensing method being applicable to higher-dimension finite sample condition

Technical field

The present invention relates to a kind of method being applied to the perception of extensive multiple antennas cognitive radio system intermediate frequency spectrum, belong to the cognitive radio technology field in radio communication.

Background technology

Multiple antennas cognitive radio technology is the study hotspot of wireless communication field, and effective multiple antennas frequency spectrum perception algorithm is one of key factor realizing this technology.The dimension M receiving data vector in multiple antennas frequency spectrum perception scene is numerically equal to the radical of antenna, and sample size N (namely receiving the number of data vector) is then determined by the number of times carrying out to received signal in detecting period sampling.The number of antennas M in traditional multiple antennas cognitive radio system, sensing node configured is often little, and the research of therefore traditional multiple antennas frequency spectrum sensing method focuses on the design problem considering algorithm under M is far smaller than the condition of N.

In concrete algorithm design, the multiple antennas total blindness frequency spectrum sensing method based on Received signal strength correlation properties attracts wide attention because of its good characteristic.Classical multiple antennas frequency spectrum sensing method comprises: based on diagonal and the detection method (CAVD) of the ratio of off diagonal element absolute value, based on eigenvalue of maximum with the detection method (MMED) of the ratio of minimal eigenvalue and based on the detection method (EMED) of Received signal strength energy with the ratio of minimal eigenvalue.Above-mentioned three kinds of methods overcome the fatal defects that sharply worsens of perceptual performance when energy detection method meets with incorrect noise phenomenon, and show excellent detection perform when multiple antennas Received signal strength exists correlation.But, CAVD, MMED and EMED tri-kinds of methods be all conceived to solve receive data vector dimension M be far smaller than sample size N condition under frequency spectrum perception problem.

In order to improve wireless frequency spectrum efficiency further, strengthen the network coverage and power system capacity, extensive MIMO (multiple-input and multiple-output) technology is just becoming the study hotspot of industrial quarters and academia.Under this background, can predict the number of antennas that secondary user's in extensive multiple antennas cognitive radio system in the future or dedicated sense equipment configure certainly will be very big.Now, in the middle of the aware application of detecting period very critical, the new situation of higher-dimension (namely M is very large) finite sample (namely N is limited) will be there is: the dimension M that multiple antennas receives data vector becomes comparable with sample size N, and M is greater than N even.Under such conditions, the multiple antennas frequency spectrum perception algorithm of the classics such as CAVD, MMED and EMED will meet with an actual difficult problem in application process: on the one hand, the precision of the theoretical perception decision threshold of above-mentioned three kinds of classical ways cannot ensure the needs realizing reliable perception in a new condition.Its reason is, is resolved decision threshold accurately owing to solving, and the decision threshold corresponding to above method is all under M is far smaller than the assumed conditions of N, adopt approximate means to obtain.When M is close to N, when M is greater than N even, the application conditions of approximate decision threshold becomes no longer to be set up, and causes cannot obtaining reliable sensing results in perception judging process; On the other hand, be far smaller than the condition of N at the M of routine under, the computation complexity of above-mentioned three kinds of classical multiple antennas frequency spectrum sensing methods depends primarily on the size of N.And at M close to N, under M is greater than the condition of N even, the size of M be can not ignore relative to N, now the computation complexity of algorithm sharply increases along with the increase receiving data vector dimension M.For MMED and EMED, in perception judging process, carry out the operand that Eigenvalues Decomposition brings is M ³rank, the application sharply increased seriously limiting traditional algorithm of the computation complexity obviously brought by the increase of dimension M under the sensed condition of the higher-dimension finite samples such as extensive multiple antennas cognitive radio.

In sum, under the higher-dimension finite sample conditions such as extensive multiple antennas cognitive radio, the design of efficient multiple antennas frequency spectrum perception algorithm will be faced with new challenges.Just under this background, the present invention proposes a kind of multiple antennas total blindness frequency spectrum sensing method being applicable to higher-dimension finite sample application scenarios based on Received signal strength correlated characteristic.

Summary of the invention

Technical problem: the present invention proposes a kind of multiple antennas frequency spectrum sensing method being applicable to higher-dimension finite sample condition.The method have realize simple, computation complexity is low, can realize the advantages such as efficient total blindness detects and sensing results is reliable under the statistical information shortage condition of primary user's signal, wireless channel and noise, can be advantageously applied to the frequency spectrum cavity-pocket test problems under higher-dimension finite sample conditions such as solving extensive multiple antennas cognitive radio system.

Technical scheme: a kind of multiple antennas frequency spectrum sensing method that can be applicable to higher-dimension finite sample condition that the present invention proposes, it utilizes the correlation between multiple antennas Received signal strength component to construct perception judgement amount, and based on Random Matrices Theory design perception decision threshold: first continuous sampling is carried out to multiple antennas Received signal strength, and form Received signal strength data matrix X; Then calculate the relativity measurement index between multiple antennas Received signal strength component on this basis, and calculate perception judgement amount l thus; Secondly the result based on Random Matrices Theory calculates perception decision threshold τ; Finally implement perception judgement: when perception judgement amount l is greater than the threshold value τ of setting, then judge that frequency spectrum cavity-pocket does not exist, otherwise then judge that frequency spectrum cavity-pocket exists.

The described multiple antennas frequency spectrum sensing method be applicable under higher-dimension finite sample condition based on multiple antennas Received signal strength correlative character, its concrete steps are:

1) sensing node is sampled to the signal on M root reception antenna to the 1st that it configures at moment n, obtains Received signal strength data vector x (n)=[x ₁(n), x ₂(n) ..., x _m(n)] ^t, wherein, subscript T representing matrix matrix transpose operation accords with;

2) step 1 is repeated), sensing node implements N continuous sampling altogether to multiline reception signal, obtain N number of Received signal strength data vector x (1), x (2) ..., x (N), and these Received signal strength data vectors are arranged in the form of matrix:

$X=\left[\begin{array}{cccc}{x}_1\left(1\right)& x_1\left(2\right)& ...& x_1\left(N\right)\\ x_2\left(1\right)& x_2\left(2\right)& ...& x_2\left(N\right)\\ .& .& & .\\ .& .& ...& .\\ .& .& & .\\ x_M\left(1\right)& x_M\left(2\right)& ...& x_M\left(N\right)\end{array}\right];$

3) the relativity measurement index between i-th and jth root antenna receiving signal component is calculated respectively to 1≤i≤M and i≤j≤M $R_{i,j}=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_i\left(n\right)x_j\left(n\right);$

4) perception judgement amount is calculated: $l=\frac{{\mathrm{\Σ}}_{i=1}^M{R}_{i,i}^2+2{\mathrm{\Σ}}_{i=1}^M{\mathrm{\Σ}}_{j>i}^M{R}_{i,j}^2}{{\left({\mathrm{\Σ}}_{i=1}^M{R}_{i,i}\right)}^2};$

5) perception decision threshold is calculated: wherein, P _fAfor target false alarm probability, Q ^-1(P _fA) be that inverse horse khoum function is at P _fAthe value at place;

6) perception judgement is implemented:

If perception judgement amount l is greater than decision threshold τ, then judge that frequency spectrum cavity-pocket does not exist; If perception judgement amount l is less than decision threshold τ, then judge that frequency spectrum cavity-pocket exists.

The described method based on Random Matrices Theory calculating perception decision threshold is: under higher-dimension finite sample condition, and proving that perception judgement amount l obeys average based on Random Matrices Theory is 1/N+1/M+1/MN _,variance is 4/ (MN) ²real Gaussian Profile, when target false alarm probability is given as P _fAtime corresponding perception decision threshold τ pass through ) calculate, wherein, Q ^-1(P _fA) be that inverse horse khoum function is at P _fAthe functional value at place.

All explanation of symbols

Beneficial effect: beneficial effect of the present invention is mainly reflected in following three aspects:

1) cognitive method provided by the present invention utilizes large dimension Random Matrices Theory to have derived the analytical expression of perception decision threshold, the simple and reliable results of computing formula; Meanwhile, the calculating of perception judgement amount is without the need to carrying out the decomposition of characteristic value, and realization is simple and complexity is low.New method efficiently solves the frequency spectrum perception problem under the higher-dimension finite sample conditions such as extensive multiple antennas cognitive radio system;

2) cognitive method provided by the present invention can realize efficient detection under the statistical information shortage condition of primary user's signal, wireless channel and noise, simultaneously the calculating of decision threshold is without the need to the knowledge of noise variance, is a kind of total blindness's multiple antennas frequency spectrum sensing method applied widely;

3) cognitive method provided by the present invention be equally applicable to Received signal strength data vector dimension in conventional multiple antennas cognitive radio scene be less than sample size condition under frequency spectrum perception problem.

Accompanying drawing explanation

Fig. 1 is a kind of realization flow figure being applicable to the multiple antennas frequency spectrum sensing method of higher-dimension finite sample condition.

Fig. 2 is target false alarm probability P _fA=0.1, Received signal strength data vector dimension and sample size are close to the performance comparison figure of the present invention under the condition of (M=80, N=100) and the classical multiple antennas frequency spectrum sensing method of existing CAVD, MMED and EMED tri-kinds.

Fig. 3 is target false alarm probability P _fA=0.1, Received signal strength data vector dimension is greater than the performance comparison figure of the present invention and the classical multiple antennas frequency spectrum sensing method of existing CAVD, MMED and EMED tri-kinds under the condition of sample size (M=100, N=80).

Embodiment

A kind of multiple antennas frequency spectrum sensing method that can be applicable to higher-dimension finite sample condition provided by the present invention, it utilizes the correlation between multiple antennas Received signal strength component to construct perception judgement amount, and based on Random Matrices Theory design perception decision threshold: first continuous sampling is carried out to multiple antennas Received signal strength, and form Received signal strength data matrix X; Then calculate the relativity measurement index between multiple antennas Received signal strength component on this basis, and calculate perception judgement amount l thus; Secondly the result based on Random Matrices Theory calculates perception decision threshold τ; Finally implement perception judgement: then judge that frequency spectrum cavity-pocket does not exist when perception judgement amount l is greater than the threshold value τ of setting, otherwise then judge that frequency spectrum cavity-pocket exists.

Below the design of the multiple antennas frequency spectrum sensing method being applicable to higher-dimension finite sample condition is described in detail.

(1) Mathematical Modeling

If sensing node configuration M root reception antenna.Sensing node carries out sampling in the n-th sampling instant to the signal on M root reception antenna and obtains M × 1 dimension Received signal strength data vector x (n)=[x ₁(n) ... x _m(n)] ^t.Because Received signal strength is primary user's signal component and the superposing of noise, therefore Received signal strength data vector can be expressed as x (n)=s (n)+η (n), η (n) and s (n) represents that primary user's Received signal strength component vector of white Gaussian noise vector sum after wireless channel is tieed up in M × 1 respectively here.

In perception, sensing node implements N sampling continuously to multi-antenna signal, and makes based on obtained N number of reception data vector the judgement whether frequency spectrum cavity-pocket exist.Use H ₀represent that primary user's signal does not exist (frequency spectrum cavity-pocket existence), H ₁represent that primary user's signal exists (frequency spectrum cavity-pocket does not exist).Then multiple antennas frequency spectrum perception problem mathematically can be expressed as described binary hypothesis test model:

$\left\{\begin{array}{cc}{H}_0:x\left(n\right)=\mathrm{\η}\left(n\right),& i=1,...,N\\ H_1:x\left(n\right)=s\left(n\right)+\mathrm{\η}\left(n\right),& i=1,...,N\end{array}\right.---\left(1\right)$

(2) implementation method

N number of Received signal strength data vector x (1) that sensing node continuous sampling is obtained, x (2) ..., x (N) is expressed as following matrix form:

$X=\left[\begin{array}{cccc}{x}_1\left(1\right)& x_1\left(2\right)& ...& x_1\left(N\right)\\ x_2\left(1\right)& x_2\left(2\right)& ...& x_2\left(N\right)\\ .& .& & .\\ .& .& ...& .\\ .& .& & .\\ x_M\left(1\right)& x_M\left(2\right)& ...& x_M\left(N\right)\end{array}\right]---\left(2\right)$

The appearance of primary user's signal not only changes the energy size of Received signal strength, changes the dependency structure between multiple antennas Received signal strength component simultaneously.This phenomenon by embody a concentrated reflection of Received signal strength data vector the change of each element size of sampling covariance matrix on, this defined matrix is:

${\hat{R}}_x=\frac{1}{N}{\mathrm{XX}}^T---\left(3\right)$

For convenience of implementing, N being multiplied by formula (3) both sides simultaneously, and matrix of consequence is expressed as:

Here, R _i,j(1≤i≤M, 1≤j≤M) is the i-th row jth column element.Convolution (3) and (4) can calculate:

$R_{i,j}=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_i\left(n\right)x_j\left(n\right)---\left(5\right)$

R _i,jin fact provide the one tolerance of the correlation size between i-th and jth root antenna receiving signal component.Noticing has R as i ≠ j _i,j=R _j,iset up, therefore only need calculate 1≤i≤M when calculating these Measure Indexes and meet the part of i≤j≤M.

When primary user's signal does not occur can be approximated to be a diagonal matrix, and when primary user's signal occurs, because between multiple antennas Received signal strength, the existence of correlation makes it is no longer a diagonal matrix.True based on this, the present invention devises new perception judgement amount:

$l=\frac{{\mathrm{\Σ}}_{i=1}^M{R}_{i,i}^2+2{\mathrm{\Σ}}_{i=1}^M{\mathrm{\Σ}}_{j>i}^M{R}_{i,j}^2}{{\left({\mathrm{\Σ}}_{i=1}^M{R}_{i,i}\right)}^2}---\left(6\right)$

Classical CAVD, MMED and EMEM tri-kinds of multiple antennas frequency spectrum perception algorithms are conceived to when analyzing the statistical nature of judgement amount obtain APPROXIMATE DISTRIBUTION under reception data vector dimension M is far smaller than the condition of sample size N.Its result M arrive greatly its value close to or be greater than N higher-dimension finite sample condition under aware application in the middle of become no longer valid.The large dimension random theory risen recently then solves this class problem and provides powerful, its current research result is utilized to prove: the dimension M receiving data vector becomes very large, its value close to or when being greater than N, under the condition that primary user's signal does not occur, judgement amount l obeys that average is 1/N+1/M+1/MN, variance is 4/ (MN) ²real Gaussian Profile.Therefore, when being P to the false alarm probability that sets the goal _fAtime, can calculate corresponding perception decision threshold is easily:

$\mathrm{\τ}=\frac{1}{N}+\frac{1}{M}+\frac{1}{\mathrm{NM}}(1+{2Q}^{-1}\left(P_{\mathrm{FA}}\right))---\left(7\right)$

Wherein, Q ^-1(P _fA) be that inverse horse khoum function is at P _fAthe value at place.

Comprehensive above analysis is known, frequency spectrum sensing method designed by the present invention does not relate to the complex calculation such as the Eigenvalues Decomposition of sampling covariance matrix when calculating perception judgement amount, realization is simple and computation complexity is low, to derive based on large dimension Random Matrices Theory the analytical expression of perception decision threshold simultaneously, the multiple antennas frequency spectrum perception problem under higher-dimension finite sample condition can have been solved well.Meanwhile, the structure of perception judgement amount corresponding to new method and the calculating of decision threshold do not rely on wireless channel, primary user's signal statistics feature and noise variance information, are a kind of total blindness's multiple antennas frequency spectrum sensing methods.

(3) concrete implementation step

Here combine analytic process above and flow chart 1, the implementation step of a kind of multiple antennas frequency spectrum sensing method that can be applicable under higher-dimension finite sample condition involved in the present invention be further described:

A () sensing node is sampled to the Received signal strength on M root reception antenna at moment n, obtain Received signal strength data vector x (n)=[x ₁(n), x ₂(n) ..., x _m(n)] ^t;

(b) repeat abovementioned steps to multiline reception signal carry out N time sample obtain x (1), x (2) ..., x (N), on this basis the Received signal strength data vector obtained is expressed as such as formula the matrix form represented by (2);

C () utilizes formula (5) to calculate the relativity measurement index R between i-th and jth root antenna receiving signal component respectively to 1≤i≤j≤M _i,j;

D () utilizes formula (6) to calculate perception judgement amount l;

E () utilizes formula (7) to calculate perception decision threshold τ;

F () implements perception judgement: if perception judgement amount l is greater than decision threshold τ, then judge that frequency spectrum cavity-pocket does not exist; If perception judgement amount l is less than decision threshold τ, then judge that frequency spectrum cavity-pocket exists.

Finally, beneficial effect of the present invention is verified by numerical simulation.Offered target false alarm probability P in simulation process _fA=0.1.Fig. 2 gives the performance comparison figure of Received signal strength data vector dimension M and sample size N close to the present invention under the condition of (M=80, N=100) and the classical multiple antennas frequency spectrum sensing method of existing CAVD, MMED and EMED tri-kinds.Simulation result shows, method provided by the present invention is obviously better than the classical multiple antennas total blindness frequency spectrum sensing method of existing MMED, EMED and CAVD tri-kinds in detection perform.Fig. 3 give Received signal strength data vector dimension be greater than sample size (M=100, N=80) condition under performance comparison figure.Simulation result shows, on the one hand, the classical detection probability corresponding to MMED, EMED and CAVD method is all close to 0, and it shows that these three kinds of methods to be greater than under the condition of sample size N all cisco unity malfunctions and losing efficacy at dimension M; On the other hand, method provided by the invention still efficiently can be implemented the detection of frequency spectrum cavity-pocket in this case and provide reliable sensing results, reveals good robust property to the change list receiving data vector dimension.In sum, simulation results show the present invention is that multiple antennas frequency spectrum perception problem under higher-dimension finite sample condition provides a kind of effective solution.

Claims (3)

1. one kind is applicable to the multiple antennas frequency spectrum sensing method of higher-dimension finite sample condition, it is characterized in that the method utilizes the correlation between multiple antennas Received signal strength component to construct perception judgement amount, and calculate perception decision threshold based on Random Matrices Theory: first multiple antennas Received signal strength is sampled, and form reception data matrix; Then the relativity measurement index structure perception judgement amount between multiple antennas Received signal strength component is utilized on this basis; Secondly perception decision threshold is calculated based on Random Matrices Theory; Finally implement perception judgement: then judge that frequency spectrum cavity-pocket does not exist when perception judgement amount is greater than the threshold value of setting, then judge that frequency spectrum cavity-pocket exists when this perception judgement amount is less than this threshold value.

2. the multiple antennas frequency spectrum sensing method being applicable to higher-dimension finite sample condition according to claim 1, is characterized in that the method concrete steps are:

1) sensing node is sampled to the signal on M root reception antenna to the 1st that it configures at moment n, obtains M × 1 and ties up Received signal strength data vector x (n)=[x ₁(n), x ₂(n) ..., x _m(n)] ^t, wherein, subscript T representing matrix matrix transpose operation accords with;

2) step 1 is repeated), sensing node implements N continuous sampling altogether to multiline reception signal, obtain N number of Received signal strength data vector: x (1), x (2) ..., x (N), and these Received signal strength data vectors are arranged in the form of matrix:

$X=\left[\begin{array}{cccc}{x}_1& x_1\left(2\right)& \·\·\·& x_1\left(N\right)\\ x_2\left(1\right)& x_2\left(2\right)& \·\·\·& x_2\left(N\right)\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ x_M\left(1\right)& x_M\left(2\right)& \·\·\·& x_M\left(N\right)\end{array}\right];$

3) the relativity measurement index between i-th and jth root antenna receiving signal component is calculated respectively to 1≤i≤M and i≤j≤M $R_{i,j}=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_i\left(n\right)x_j\left(n\right);$

4) perception judgement amount is calculated: $l=\frac{{\mathrm{\Σ}}_{i=1}^M{R}_{i,i}^2+2{\mathrm{\Σ}}_{i=1}^M{\mathrm{\Σ}}_{j>i}^M{R}_{i,j}^2}{{\left({\mathrm{\Σ}}_{i=1}^M{R}_{i,i}\right)}^2};$

5) perception decision threshold is calculated: wherein, P _fAfor target false alarm probability, Q ^-1(P _fA) be that inverse horse khoum function is at P _fAthe value at place;

6) perception judgement is implemented:

If perception judgement amount l is greater than decision threshold τ, then judge that frequency spectrum cavity-pocket does not exist; If perception judgement amount l is less than decision threshold τ, then judge that frequency spectrum cavity-pocket exists.

3. the multiple antennas frequency spectrum sensing method being applicable to higher-dimension finite sample condition according to claim 1, it is characterized in that the described method based on Random Matrices Theory calculating perception decision threshold is: under higher-dimension finite sample condition, prove that perception judgement amount l obedience average is 1/N+1/M+1/MN, variance is 4/ (MN) based on Random Matrices Theory ²real Gaussian Profile, when target false alarm probability is given as P _fAtime corresponding perception decision threshold τ pass through calculate, wherein, Q ^-1(P _fA) be that inverse horse khoum function is at P _fAthe functional value at place.