CN103297160A - Spectrum sensing method and spectrum sensing device for goodness-of-fit test based on normalized eigenvalues - Google Patents

Abstract

The invention relates to the technical field of wireless communication and discloses a spectrum sensing method and a spectrum sensing device for goodness-of-fit test based on normalized eigenvalues. The method includes receiving signals on authorized spectrum bands by a spectrum sensing device, performing sampling filter on the received signals, then calculating a covariance matrix, performing eigenvalue decomposition on the covariance matrix, sorting eigenvalues from small to large, dividing the eigenvalues by sum of all the eigenvalues to obtain normalized eigenvalues, performing goodness-of-fit test on the normalized eigenvalues, and judging whether signals exist or not according to test results. The method and the device have the advantages of no need of authorized signal features, noise uncertainty and insensitivity and the like, and the device is excellent in performance.

Description

Frequency spectrum sensing method and device based on the test of fitness of fot of normalization characteristic value

Technical field

The present invention relates to wireless communication technology field, particularly relate to a kind of frequency spectrum sensing method and device that need not to send any characteristic information of signal.

Background technology

The problem of radio spectrum resources scarcity becomes more and more outstanding in the growth at full speed along with wireless data service.The proposition of cognitive radio, make flexible Application occupied frequency spectrum resource become possibility.The development of cognitive radio technology has greatly improved the availability of frequency spectrum, has alleviated growing wireless traffic demand and the contradiction between the deficient frequency spectrum resource day by day, is generally believed it is the preferred plan that solves present wireless frequency spectrum utilance problem.Existing studies show that, cognitive radio can be improved capacity of communication system and be improved spectrum management efficient.

In cognitive radio system, how differentiating the signal that whether has authorized user on the frequency spectrum is the problem that at first will solve.This problem is called as frequency spectrum detection or frequency spectrum perception.Frequency spectrum detecting method commonly used comprises: energy measuring, matched filtering detection and cyclo-stationary detect.The energy measuring complexity is low, but is subjected to probabilistic influence of noise, and its mis-behave is serious.Matched filtering excellent performance, but the feature of the known transmission signal of needs.The performance that cyclo-stationary detects is also very excellent, but complexity is higher, when practical application, is subjected to certain limitation.

Can judge on this frequency range it is signal or noise according to the covariance matrix that receives signal, the theoretical foundation of its work is as follows.We know, usually, have under the situation that signal exists, and the covariance matrix of signal is not a diagonal matrix, and when only noise existed, the covariance matrix that receives signal was the matrix that diagonal entry equates.Based on this theoretical foundation, traditional technical scheme has proposed a kind of characteristic value of the covariance matrix that receives signal of utilizing and has carried out frequency spectrum perception, and judgment variables can be made of the characteristic value of covariance matrix.And a kind of judgment variables building method, the i.e. ratio of minimax characteristic value have been provided.In addition, conventional solution also proposes the ratio of a kind of geometric average of utilizing characteristic value and arithmetic average as judgment variables.The advantage of these methods is any prior informations that need not to send signal, also need not any statistical property of noise.But these methods are not utilized the distribution character of characteristic value.

Summary of the invention

Technical problem: in order further to promote the performance of characteristic value detection method, and avoid the probabilistic influence of noise, the present invention proposes a kind of frequency spectrum detecting method and device of the test of fitness of fot based on the normalization characteristic value.

Technical scheme: the frequency spectrum sensing method based on the test of fitness of fot of normalization characteristic value comprises the steps:

(1) receives the wireless signal for the treatment of on the perception frequency range;

(2) carry out sampling filter to received signal, calculate the covariance matrix of signal, be designated as R, its dimension is L * L;

(3) characteristic value of calculating covariance matrix is decomposed, the characteristic value that obtains sorting, and described characteristic value is expressed as σ from small to large ₁, σ ₂..., σ _L

(4) calculate the normalization characteristic value, namely characteristic value divided by all characteristic values and, the result is expressed as

(5) according to the cumulative distribution function F (x) of the normalization characteristic value of noise, calculate the cumulative probability of normalization characteristic value:

The cumulative distribution function F (x) of the normalization characteristic value of described noise calculates by theory, perhaps by emulation F (x) is made into the form of form, and tabling look-up obtains the cumulative probability of normalization characteristic value.

(6) calculate judgment variables T according to the test of fitness of fot, when T greater than predefined thresholding, then judging has authorization signal to exist on this frequency spectrum, when T less than predefined thresholding, then judging does not have authorization signal, i.e. this frequency spectrum free time.

Described judgment variables T adopts the Anderson-Darling check,

$T=-\frac{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}(2l-1)\left\{1n\right[F\left({\widetilde{\mathrm{\σ}}}_l\right)+1n[1-F\left({\widetilde{\mathrm{\σ}}}_{L+1-l}\right)]\}}{L}-L$

Wherein, ln () expression natural logrithm function.

Preferably, described method is applicable to the frequency spectrum perception of single antenna and multiaerial system.

Preferably, described method is applicable to the cooperation perception of multinode.

Frequency spectrum sensing device based on the test of fitness of fot of normalization characteristic value comprises: the normalization computing module of wireless signal samples and filtration module, covariance matrix computing module, characteristic value decomposing module, characteristic value, the cumulative probability computing module of normalization characteristic value, judgment variables computing module and judging module; Wherein,

Described wireless signal samples and filtration module are for the wireless signal that obtains institute's perception frequency range;

Described covariance matrix computing module is used for calculating the covariance matrix for the treatment of perceptual signal;

Described characteristic value decomposing module, the characteristic value decomposition that is used for calculating the covariance matrix for the treatment of perceptual signal;

The normalization computing module of described characteristic value, with characteristic value divided by all characteristic values and, obtain the normalization characteristic value;

The cumulative probability computing module of described normalization characteristic value calculates the cumulative probability of each normalization characteristic value correspondence;

Described judgment variables computing module calculates inspected number according to the cumulative probability of normalization characteristic value; Adopt Anderson-Darling test of fitness of fot method;

Described judging module comprises comparator, is used for relatively judgment variables and thresholding.

The present invention adopts technique scheme, have following beneficial effect: in judgment variables T calculation stages, this method only needs simple addition and comparison operation, compares in the background technology frequency spectrum perception based on characteristic value, complexity is very low, and uncertain insensitive to noise.In addition, the present invention also is applicable to the frequency spectrum perception of multiaerial system.

Description of drawings

Fig. 1 is the frequency spectrum sensing method flow chart of the embodiment of the invention;

Fig. 2 is the frequency spectrum sensing device block diagram of the embodiment of the invention;

Judgment variables computing module block diagram when Fig. 3 is the employing Anderson-Darling test of fitness of fot method of the embodiment of the invention;

Fig. 4 is at the wireless microphone signal, and sampled point is 64 o'clock, and the performance of the embodiment of the invention and background technology is schematic diagram relatively;

Fig. 5 is at the wireless microphone signal, and sampled point is 128 o'clock, and the performance of the embodiment of the invention and background technology is schematic diagram relatively;

Fig. 6 is at multipoint cooperative frequency spectrum perception model, and sampled point is 32 o'clock, and the performance of the embodiment of the invention and background technology is schematic diagram relatively;

Fig. 7 is at multipoint cooperative frequency spectrum perception model, and sampled point is 64 o'clock, and the performance of the embodiment of the invention and background technology is schematic diagram relatively.

Embodiment

Below in conjunction with specific embodiment, further illustrate the present invention, should understand these embodiment only is used for explanation the present invention and is not used in and limits the scope of the invention, after having read the present invention, those skilled in the art all fall within the application's claims institute restricted portion to the modification of the various equivalent form of values of the present invention.

Frequency spectrum detecting method based on the test of fitness of fot of normalization characteristic value as described in Figure 1, may further comprise the steps:

1) frequency spectrum sensing device receives the wireless signal for the treatment of on the perception frequency range;

2) after frequency spectrum sensing device carries out sampling filter to received signal, calculate the covariance matrix of signal, be designated as R, its dimension is L * L;

Below we discuss the calculating of signal covariance matrix under two kinds of scenes respectively.

Scene 1: the covariance matrix of signal can estimate by the method for moving average.Suppose to receive burst and be expressed as,

x(1),x(2),x(N),x(N+1),x(N+2),...,x(nN),x(nN+1),....

Wherein, N represents the over-sampling of signal count (or for many antennas receiving system, N can be modeled as the number of reception antenna).We adopt the method for sliding window to calculate the covariance matrix that receives signal.Suppose smoothing factor M, so,

[formula 1]

Wherein, K represents to calculate the vector number of summation,

$x_n={\left[\begin{array}{cccc}x\left(\mathrm{nN}\right)& x(\mathrm{nN}+1)& ...& x(\mathrm{nN}+N-1)\end{array}\right]}^T$ [formula 2]

${\hat{x}}_n={\left[\begin{array}{cccc}{x}_n^T& x_{n-1}^T& ...& x_{n-M+1}^T\end{array}\right]}^T$ [formula 3]

R is the matrix of a NM * NM, and makes L=NM, and then R is the matrix of a L * L.

So, when only having noise,

[formula 4]

Wherein, w _nDefinition and x _nSimilar.We suppose that noise is white Gaussian noise, and its variance is γ ²From the mathematics strictly, R _wBe not the Wishart matrix, still, R _wCan be approximated to be the Wishart matrix.

Scene 2: the collaborative spectrum sensing model suppose to adopt flat fading channel, and the cooperative node number is L.So, n reception signal constantly can be expressed as,

$x_n=P^{\frac{1}{2}}{H}_n{s}_n+w_n$ [formula 5]

Wherein, x _nBe the reception signal phasor of L * 1, s _nBe the transmission signal phasor of K * 1, H _nBe the Rayleigh fading channel matrix of standard, P is that L * L receives correlation matrix, w _nBe independent identically distributed white Gaussian noise, its variance is γ ²So, the covariance matrix that receives signal can calculate by following formula,

R is the matrix of a L * L.When only having noise,

As can be seen, from mathematics, R _wBe the Wishart matrix.

In order to improve the performance of traditional detection technology, we propose to utilize the distribution character of characteristic value to detect.Its basic principle is that when only having noise, we can obtain the distribution character of characteristic value.Therefore, whether we can check the characteristic value distribution that receives signal to distribute similar to the characteristics of noise value by the test of fitness of fot (Goodness of Fit Test).Be subjected to probabilistic influence of noise for fear of detector, we further carry out the test of fitness of fot according to the normalization characteristic value.For this reason, we carry out step 3) and 4), distribute and calculate characteristic value and the normalization characteristic value of covariance matrix.

3) the frequency spectrum sensing device covariance matrix carries out the characteristic value decomposition, and characteristic value is expressed as σ from small to large ₁, σ ₂..., σ _L

4) frequency spectrum sensing device calculates the normalization characteristic value, namely characteristic value divided by all characteristic values and, the result is expressed as

Below, we will carry out the test of fitness of fot of normalization characteristic value.

5) frequency spectrum sensing device calculates the cumulative probability of normalization characteristic value according to the cumulative distribution function F (x) of the normalization characteristic value of noise:

When carrying out the test of fitness of fot, a committed step is to calculate the cumulative probability of data to be tested.And the cumulative probability of normalization characteristic value is calculated and can be calculated by theoretical formula, also can obtain the cumulative probability of normalization characteristic value by emulation, is stored as table for inquiry.Below, we are according to R under foregoing two kinds of signal models _wComputational methods are derived the approximate expression of the cumulative distribution function of normalization characteristic value when only having noise respectively theoretically.

Here, we suppose L 〉=K, and this is rational in actual conditions.According to R _wThe characteristic value sum equal R _wMark, R so _wThe average of characteristic value sum is L γ ²Therefore, only exist under the situation of noise, when asking the cumulative distribution function of normalization characteristic value, we are with L γ ²As approximate characteristic value and.The cumulative distribution function of the normalization characteristic value that we obtain so, is approximate expression.

According to the cumulative distribution function of the asymptotic characteristic root of Wishart matrix on the mathematics, we can calculate with L γ by following formula ²The cumulative distribution function of characteristic value during for normalization factor,

$F\left(x\right)=\{\begin{array}{cc}0& 0<x\&le;a/\mathrm{<figure-callout\; id="1"\; label="L"\; filenames="HDA00003252520100021.png,HDA00003252520100031.png"\; state="\{\{state\}\}">L</figure-callout>}\\ \frac{1}{2\mathrm{\&pi;}}G\left(\mathrm{Lx}\right)+\frac{1}{2}& a/L<x\&le;b/L\\ 1& x>b/L\end{array}$

Wherein, $a={(1-\sqrt{K/L})}^2,b={(1+\sqrt{K/L})}^2,$

$G\left(x\right)\stackrel{\mathrm{\&Delta;}}{=}\&Integral;\frac{1}{x}\sqrt{(x-a)(b-x)}\mathrm{dx}$

$=\sqrt{(x-a)(b-x)}-\sqrt{\mathrm{ab}}\arcsin \frac{(a+b)x-2\mathrm{ab}}{(b-a)x}-\frac{a<figure-callout\; id="2"\; label="+\; b"\; filenames="HDA00003252520100031.png,HDA00003252520100032.png"\; state="\{\{state\}\}">+</figure-callout>b}{2}\arcsin \left(\frac{a+b-2x}{b-a}\right)$

During actual the realization, we can be made into table with cumulative distribution function, are stored in the memory, in order to reduce computation complexity.Preferably, we can also obtain the cumulative distribution of more accurate normalization characteristic value by emulation, and are made into table.Obtain the cumulative probability of normalization characteristic value by tabling look-up.

6) frequency spectrum sensing device calculates judgment variables T according to the test of fitness of fot.When T greater than predefined thresholding, then described frequency spectrum sensing device is judged has authorization signal to exist on this frequency spectrum, when T less than predefined thresholding, then described frequency spectrum sensing device is judged does not have authorization signal, i.e. this frequency spectrum free time.

The test of fitness of fot has multiple implementation method.One of them implements special case is to adopt the Anderson-Darling check, that is:

$T=-\frac{\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}(2l-1)\left\{1n\right[F\left({\widetilde{\mathrm{\&sigma;}}}_l\right)+1n[1-F\left({\widetilde{\mathrm{\&sigma;}}}_{L+1-l}\right)]\}}{L}-L$ [formula 6]

In sum, by step 1)～6), can obtain the result of frequency spectrum perception.

Realize the cumulative distribution function of normalization characteristic value by tabling look-up, the present invention compares traditional characteristic value frequency spectrum detecting method, and complexity has only increased the calculating of [formula 6], thereby complexity is not high.

Below in conjunction with block diagram, the job step of frequency spectrum sensing method of the present invention is further described.

As shown in Figure 1, at first, frequency spectrum perception equipment receives the wireless signal for the treatment of on the perception frequency range, after carrying out sampling filter to received signal, calculate the covariance matrix R of signal, calculate the characteristic value of covariance matrix and decompose, the characteristic value that obtains sorting is calculated the normalization characteristic value then, be characteristic value divided by all characteristic values and, at last, calculate judgment variables T according to the test of fitness of fot, when T greater than predefined thresholding, then described frequency spectrum sensing device is judged has authorization signal to exist on this frequency spectrum, when T less than predefined thresholding, then described frequency spectrum sensing device is judged does not have authorization signal, i.e. this frequency spectrum free time.

In conjunction with Fig. 2 and Fig. 3, below frequency spectrum sensing device of the present invention is further described.As shown in Figure 2, frequency spectrum sensing device of the present invention comprises: the characteristic value decomposing module of covariance matrix computing module, ordering, normalization characteristic value calculating module, test of fitness of fot judgment variables computing module, judging module.Wherein, wireless signal samples and filtration module are used for obtaining the signal of institute's perception frequency range, the covariance matrix computing module is used for calculating the covariance matrix for the treatment of perceptual signal, the characteristic value decomposing module of ordering is used for asking characteristic value to decompose to covariance matrix, and characteristic value sorted from small to large, the normalization characteristic value calculating module with characteristic value divided by all characteristic values and.Fig. 3 has provided the judgment variables computing module when adopting Anderson-Darling test of fitness of fot method, it comprises the cumulative probability computing module of normalization characteristic value and the computing module of Anderson-Darling check judgment variables T, and wherein the calculating of T is shown in [formula 6].

Below by actual emulation checking, scene 1 and scene have been provided 2 times by Fig. 4～Fig. 7, the performance comparison of the inventive method and background technology.Energy measuring, background technology that we have contrasted under the noiseless uncertainty comprise that minimax characteristic value ratio detects and the geometric average of characteristic value and the detection of arithmetic average ratio, the energy measuring under the noise uncertainty and the art of this patent.The art of this patent and minimax characteristic value ratio detect all insensitive to the noise uncertainty, just, exist the noise uncertainty whether not influence their performance.What provide among the figure is to have under the noise uncertainty performance of the art of this patent and background technology.The noise coefficient of uncertainty is 0.5dB.

Fig. 4 and Fig. 5 are that the signal model with scene 1 is example, and sending signal is the wireless microphone signal, and sampling number is 64 and 128 points, and slippage factor is 16, and oversample factor is 1.As can be seen from Figure 4, this patent method is poorer slightly than the energy measuring under the noiseless uncertainty, and the energy measuring when still obviously being better than existing noise uncertain, minimax characteristic value ratio detect and geometric average and the arithmetic average ratio of characteristic value detect.Along with number of samples is increased to 128 points, as can be seen from Figure 5, this patent method greater than 0.9 o'clock, is better than all other technology of considering in detection probability, comprises accurately known energy measuring of noise variance.

Fig. 6 and Fig. 7 are that the signal model with scene 2 is example, and signal produces and provided by [formula 5].The element that receives correlation matrix P is produced by following method: the i of P matrix is capable, and the j column element equals α ^{| i-j|}, wherein α equals 0.9.Suppose that cooperation antenna number L is 16.Detection performance when Fig. 6 and Fig. 7 have provided K=32 and K=64 respectively.As can be seen from Figures 6 and 7, the embodiment of the invention is poorer slightly than the energy measuring under the noiseless uncertainty, and the energy measuring when still obviously being better than existing noise uncertain, minimax characteristic value ratio detect and geometric average and the arithmetic average ratio of characteristic value detect.

One of ordinary skill in the art will appreciate that all or part of step in the said method can instruct related hardware to finish by program, described program can be stored in the computer-readable recording medium, as read-only memory, disk or CD etc.Alternatively, all or part of step of above-described embodiment also can use one or more integrated circuits to realize.Correspondingly, each the module/unit in above-described embodiment can adopt the form of hardware to realize, also can adopt the form of software function module to realize.The present invention is not restricted to the combination of the hardware and software of any particular form.

Claims (7)

1. based on the frequency spectrum sensing method of the test of fitness of fot of normalization characteristic value, it is characterized in that, comprise the steps:

(1) receives the wireless signal for the treatment of on the perception frequency range;

(2) carry out sampling filter to received signal, calculate the covariance matrix of signal, be designated as R, its dimension is L * L;

(3) characteristic value of calculating covariance matrix is decomposed, the characteristic value that obtains sorting, and described characteristic value is expressed as σ from small to large ₁, σ ₂..., σ _L

(4) calculate the normalization characteristic value, namely characteristic value divided by all characteristic values and, the result is expressed as

(5) according to the cumulative distribution function F (x) of the normalization characteristic value of noise, calculate the cumulative probability of normalization characteristic value:

(6) calculate judgment variables T according to the test of fitness of fot, when T greater than predefined thresholding, then judging has authorization signal to exist on this frequency spectrum, when T less than predefined thresholding, then judging does not have authorization signal, i.e. this frequency spectrum free time.

2. the frequency spectrum sensing method of the test of fitness of fot based on the normalization characteristic value according to claim 1, it is characterized in that, the cumulative distribution function F (x) of the normalization characteristic value of described noise calculates by theory, perhaps by emulation F (x) is made into the form of form, tabling look-up obtains the cumulative probability of normalization characteristic value.

3. the frequency spectrum sensing method of the test of fitness of fot based on the normalization characteristic value according to claim 1, it is characterized in that: described judgment variables T adopts the Anderson-Darling check,

$T=-\frac{\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}(2l-1)\left\{1n\right[F\left({\widetilde{\mathrm{\&sigma;}}}_l\right)+1n[1-F\left({\widetilde{\mathrm{\&sigma;}}}_{L+1-l}\right)]\}}{L}-L$

Wherein, ln () expression natural logrithm function.

4. according to the frequency spectrum sensing method of each described test of fitness of fot based on the normalization characteristic value of claim 1-3, it is characterized in that: described method is applicable to the frequency spectrum perception of single antenna and multiaerial system.

5. according to the frequency spectrum sensing method of each described test of fitness of fot based on the normalization characteristic value of claim 1-3, it is characterized in that: described method is applicable to the cooperation perception of multinode.

6. based on the frequency spectrum sensing device of the test of fitness of fot of normalization characteristic value, comprising: the normalization computing module of wireless signal samples and filtration module, covariance matrix computing module, characteristic value decomposing module, characteristic value, the cumulative probability computing module of normalization characteristic value, judgment variables computing module and judging module; It is characterized in that,

Described wireless signal samples and filtration module are for the wireless signal that obtains institute's perception frequency range;

Described covariance matrix computing module is used for calculating the covariance matrix for the treatment of perceptual signal;

Described characteristic value decomposing module, the characteristic value decomposition that is used for calculating the covariance matrix for the treatment of perceptual signal;

The normalization computing module of described characteristic value, with characteristic value divided by all characteristic values and, obtain the normalization characteristic value;

The cumulative probability computing module of described normalization characteristic value calculates the cumulative probability of each normalization characteristic value correspondence;

Described judgment variables computing module calculates inspected number according to the cumulative probability of normalization characteristic value;

Described judging module comprises comparator, is used for relatively judgment variables and thresholding.

7. the frequency spectrum sensing device of the test of fitness of fot based on the normalization characteristic value according to claim 6, it is characterized in that: described judgment variables computing module adopts Anderson-Darling test of fitness of fot method.

Images