CN103684635B - A kind of method and system of secondary user's detection cognitive radio frequency spectrum - Google Patents

Abstract

The invention provides the method and system of a kind of secondary user's detection cognitive radio frequency spectrum, described method comprises: step 101) according to the time dependent diversity of Frequency Hopping Signal power spectrum, calculate secondary user's within certain period, receive the actual power spectrum of data and the difference of average power spectra, thus offset fixed frequency jamming power spectrum and retain Frequency Hopping Signal power spectrum；Step 102) the Frequency Hopping Signal power spectrum retained is carried out inverse Fourier transform is obtained auto-correlation function, then use the detection of time domain auto-correlation algorithm containing frequency range during burst fixed frequency interference.Determine the situation of frequency interference owing to the time domain auto-correlation algorithm of prior art does not accounts for happening suddenly, i.e. only considered the single signal of superposition white Gaussian noise, thus accuracy of detection is the lowest.And the time domain Autocorrelation Detection method improved in the present invention considers the burst fixed frequency interference impact on testing result, so that accuracy of detection is higher.

Description

A kind of method and system of secondary user's detection cognitive radio frequency spectrum

Technical field

The present invention relates to cognitive radio frequency spectrum perception field, be specifically related to a kind of secondary user's detection cognitive radio The method and system of frequency spectrum.

Background technology

Along with developing rapidly of wireless communication technology, frequency spectrum resource becomes more and more nervous.Especially with wireless office Territory net (WLAN) technology, the development of wireless personal area network (WPAN) technology, increasing people passes through These technology access the Internet wirelessly.These network technologies are used mostly unauthorized frequency range (UFB) Work.Due to the fast development of WLAN, WPAN radio communication service, it is unauthorized that these networks are worked Frequency range is the most saturated.And other communication service (such as visual broadcast service etc.) needs communication network to provide Certain protection, makes them from the interference of other communication services.In order to provide good protection, frequency management portion The door specific mandate frequency range (LFB) of specific assigned uses for specific communication service.Compared with authorizing frequency range, The frequency spectrum resource of unauthorized frequency range to lack a lot (most frequency spectrum resource is all used to do mandate frequency range and uses). And the utilization rate of a considerable amount of mandate frequency spectrum resource is the lowest.Then the fact has been occurred as soon as: some portion Point frequency spectrum resource relatively fewer but portfolio of carrying it on is very big, and the frequency spectrum resource that other has authorized is sharp The lowest by rate.Therefore, it can reach a conclusion that the frequency spectrum resource allocation method based on current have quite The utilization rate of a part of frequency spectrum resource is the lowest.

The problem deficient in order to solve frequency spectrum resource, basic ideas are just to try to improve the utilization rate of existing frequency spectrum.For This, there has been proposed the concept of cognitive radio.The basic point of departure of cognitive radio is exactly: in order to improve frequency spectrum Utilization rate, the Wireless Telecom Equipment (i.e. secondary user's) with cognitive function " can be waited for an opportunity according to certain (Opportunistic Way) " mode be operated in the frequency range authorized.Certainly, this must set up and award Weigh the useless or the most little communication service of frequency range in the case of activity.This go out in spatial domain, time domain and frequency domain The existing frequency spectrum resource that can be utilized is referred to as " frequency spectrum cavity-pocket ".The core concept of cognitive radio makes wireless exactly Communication equipment has discovery " frequency spectrum cavity-pocket " the ability of Appropriate application.

When unauthorized communication user (i.e. secondary user's) uses, by the way of " using ", the frequency spectrum resource authorized, Must assure that its communication does not interferes with the communication of other authorized users (i.e. primary user).This point to be accomplished, Unauthorized user must use " frequency spectrum cavity-pocket " found according to certain rule.

In prior art when primary user uses frequency-hopping mode to communicate, the secondary user of cognitive radio generally uses time domain certainly Correlation detection carries out frequency spectrum detection, if at this moment the Frequency Hopping Signal of primary user go here and there into burst determine frequency interference can cause inspection Dendrometry loses.And the resolution policy not being correlated with currently for this technical problem.

Summary of the invention

It is an object of the invention to, for overcoming the problems referred to above, thus provide a kind of secondary user's detection cognitive radio The method and system of frequency spectrum.

For achieving the above object, the method that the invention provides a kind of secondary user's detection cognitive radio frequency spectrum, institute The method of stating comprises:

Step 101) according to the time dependent diversity of Frequency Hopping Signal power spectrum, calculate secondary user's when certain section The interior actual power spectrum receiving data and the difference of average power spectra, thus offset and determine frequency jamming power spectrum and retain Frequency Hopping Signal power spectrum；

Step 102) the Frequency Hopping Signal power spectrum retained is carried out inverse Fourier transform is obtained auto-correlation function, then Use the detection of time domain auto-correlation algorithm containing frequency range during burst fixed frequency interference.

Above-mentioned steps 101) comprise further:

Step 101-1) intercept the non-stationary signal that secondary user's received within certain period, and the non-stationary intercepted Signal is the discrete data sequences of a length of N, is embodied as:

$x_N\left(n\right)=\underset{i=1}{\overset{a}{\mathrm{\Σ}}}{S}_{\mathrm{Fi}}\left(n\right)+\underset{j=1}{\overset{b}{\mathrm{\Σ}}}{S}_{\mathrm{Hj}}\left(n\right)+n\left(n\right)$

Wherein, S_FiN () represents that a fixed frequency disturbs, i=1,2 ..., a；S_HjN () represents b Frequency Hopping Signal, j= 1,2,…,b；N (n) represents white Gaussian noise；Function x_NN the span of parameter n of () is: n=1,2 ..., N；

Step 101-2) non-stationary signal of above-mentioned a length of N is divided into L section stationary signal, and each section is put down The a length of M of steady signal, each section with x_MKRepresent；

Step 101-3) use equation below to calculate each section of stationary signal x_MKPower spectrum P_xMK(n′)(ω):

$P_{\mathrm{xMK}\left(n^{\′}\right)}\left(\mathrm{\ω}\right)=\frac{1}{M}{\left|{\mathrm{\Σ}}_{n^{\′}=0}^{M-1}{x}_{\mathrm{MK}}\left(n^{\′}\right)e^{-\mathrm{j\ω}{n}^{\′}}\right|}^2;$

Wherein, the span of n ' is: n '=0,1,2 ..., (M-1)；K=1,2,3,…,L；

Step 101-3) according to the meansigma methods of the power spectrum of the calculated all each section of stationary signals of equation below:

${\stackrel{\‾}{P}}_{x_{N\left(n\right)}}\left(\mathrm{\ω}\right)=\frac{1}{L}\underset{K=1}{\overset{L}{\mathrm{\Σ}}}{P}_{\mathrm{xMK}\left(n^{\′}\right)}\left(\mathrm{\ω}\right)=\frac{1}{\mathrm{ML}}\underset{K=1}{\overset{L}{\mathrm{\Σ}}}{\left|\underset{n^{\′}=0}{\overset{M-1}{\mathrm{\Σ}}}{x}_{\mathrm{MK}}\left(n^{\′}\right)e^{-\mathrm{j\ω}{n}^{\′}}\right|}^2$

Step 101-4) according to equation below, to step 101-1) secondary user's intercept whole section of non-stationary signal enter Row Fourier transformation, and then obtain the power spectrum of institute's intercept signal；

$P_{\mathrm{xN}\left(n\right)}\left(\mathrm{\ω}\right)=\frac{1}{N}{\left|\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}{x}_N\left(n\right)e^{-\mathrm{j\ωn}}\right|}^2$

Step 101-5) the meansigma methods meter of power spectrum of power spectrum based on intercept signal and all each section of stationary signals Calculation offsets power spectrum, obtains retaining Frequency Hopping Signal power spectrum, and specific formula for calculation is as follows:

$P_{\mathrm{SUB}}\left(\mathrm{\ω}\right)=P_{\mathrm{xN}\left(n\right)}\left(\mathrm{\ω}\right)-{\stackrel{\‾}{P}x}_{N\left(n\right)}\left(\mathrm{\ω}\right).$

Above-mentioned steps 102) comprise further:

Step 102 1) to retaining Frequency Hopping Signal power spectrum P_SUB(ω) carry out inverse Fourier transform and obtain auto-correlation function R_X1(τ), concrete transformation for mula is as follows:

$R_{X1}\left(\mathrm{\τ}\right)={F^{-1}}_{\mathrm{\τ}}\left[P_{\mathrm{SUB}}\left(\mathrm{\ω}\right)\right]={\∫}_{-\∞}^{+\∞}{P}_{\mathrm{SUB}}\left(\mathrm{\ω}\right)e^{\mathrm{j\ω\τ}}\mathrm{d\ω}$

Wherein, auto-correlation function R_X1(τ) it is the auto-correlation function after determining frequency AF panel；

Step 102-2) based on determining the auto-correlation function after frequency AF panel, use time domain auto-correlation strategy to realize frequency The detection in spectrum cavity.

Above-mentioned steps 102-2) comprise further:

Step 102-2-1) it is calculated as follows two characteristic quantity E according to the auto-correlation function obtained₁And E₂；

$E_1=\frac{1}{T_H}{\∫}_0^{T_H}\left|R_{X1}\left(\mathrm{\τ}\right)\right|\mathrm{d\τ}$

$E_2=\frac{1}{T-T_H}{\∫}_{T_H}^T\left|R_{X1}\left(\mathrm{\τ}\right)\right|\mathrm{d\τ}$

Step 102-2-2) calculate above-mentioned two characteristic quantity ratio ρ；

Step 102-2-3) ρ obtained is compared size with the threshold value of setting, then deposit when ρ is more than threshold value At Frequency Hopping Signal, otherwise there is no Frequency Hopping Signal.

Such scheme is by estimating ρ₁Expectation and variance yields, and combine the probability density of normally distributed random variable Distribution situation and false-alarm probability determine suitable threshold value, described ρ₁For only having characteristic quantity during noise without Frequency Hopping Signal Ratio.

Additionally, the invention provides the system of a kind of secondary user's detection cognitive radio frequency spectrum, described system comprises:

Frequency Hopping Signal power spectrum extraction module, for according to the time dependent diversity of Frequency Hopping Signal power spectrum, meter Calculate secondary user's within certain period, receive the actual power spectrum of data and the difference of average power spectra, and then balance out Fixed frequency jamming power is composed and is retained Frequency Hopping Signal power spectrum；

Time domain auto-correlation function obtains and detection module, for the Frequency Hopping Signal power spectrum retained is taken Fourier's inversion Change and obtain auto-correlation function, and then use the detection of time domain auto-correlation algorithm containing frequency range during burst fixed frequency interference.

Above-mentioned Frequency Hopping Signal power spectrum extraction module comprises further:

Intercepting submodule, for intercepting the non-stationary signal that secondary user's received within certain period, and intercepting is non- Stationary signal is the discrete data sequences of a length of N, is embodied as:

$x_N\left(n\right)=\underset{i=1}{\overset{a}{\mathrm{\Σ}}}{S}_{\mathrm{Fi}}\left(n\right)+\underset{j=1}{\overset{b}{\mathrm{\Σ}}}{S}_{\mathrm{Hj}}\left(n\right)+n\left(n\right)$

Wherein, S_FiN () represents that a fixed frequency disturbs, i=1,2 ..., a；S_HjN () represents b Frequency Hopping Signal, j= 1,2,…,b；N (n) represents white Gaussian noise；Function x_NN the span of parameter n of () is: n=1,2 ..., N；

Segmentation submodule, for the non-stationary signal of a length of N is divided into L section stationary signal, and each section is put down The a length of M of steady signal, each section with x_MKRepresent；

Each section of stationary signal spectra calculation submodule, is used for using equation below to calculate each section of stationary signal x_MK Power spectrum P_xMK(m′)(ω):

$P_{\mathrm{xMK}\left(n^{\′}\right)}\left(\mathrm{\ω}\right)=\frac{1}{M}{\left|{\mathrm{\Σ}}_{n^{\′}=0}^{M-1}{x}_{\mathrm{MK}}\left(n^{\′}\right)e^{-\mathrm{j\ω}{n}^{\′}}\right|}^2;$

Wherein, the span of n ' is: n '=0,1,2 ..., (M-1)；K=1,2,3,…,L；

Mean power obtains submodule, for the power according to the calculated all each section of stationary signals of equation below The meansigma methods of spectrum:

${\stackrel{\‾}{P}}_{x_{N\left(n\right)}}\left(\mathrm{\ω}\right)\frac{1}{L}\underset{K=1}{\overset{L}{\mathrm{\Σ}}}{P}_{\mathrm{xMK}\left(n^{\′}\right)}\left(\mathrm{\ω}\right)=\frac{1}{\mathrm{ML}}\underset{K=1}{\overset{L}{\mathrm{\Σ}}}{\left|\underset{n^{\′}=0}{\overset{M-1}{\mathrm{\Σ}}}{x}_{\mathrm{MK}}\left(n^{\′}\right)e^{-\mathrm{j\ω}{n}^{\′}}\right|}^2$

Intercept signal power spectrum obtains submodule, for according to equation below, to step 101-1) secondary user's cuts The whole section of non-stationary signal taken carries out Fourier transformation, and then obtains the power spectrum of institute's intercept signal；

$P_{\mathrm{xN}\left(n\right)}\left(\mathrm{\ω}\right)=\frac{1}{N}{\left|\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}{x}_N\left(n\right)e^{-\mathrm{j\ωn}}\right|}^2$

Process submodule, for power spectrum average of power spectrum based on intercept signal and all each section of stationary signals Value calculating offsets power spectrum, obtains retaining Frequency Hopping Signal power spectrum, and specific formula for calculation is as follows:

$P_{\mathrm{SUB}}\left(\mathrm{\ω}\right)=P_{\mathrm{xN}\left(n\right)}\left(\mathrm{\ω}\right)-{\stackrel{\‾}{P}x}_{N\left(n\right)}\left(\mathrm{\ω}\right).$

Above-mentioned time domain auto-correlation function obtains and detection module comprises further:

Auto-correlation function obtains submodule, for retaining Frequency Hopping Signal power spectrum P_SUB(ω) Fourier's inversion is carried out Get auto-correlation function R in return_X1(τ), concrete transformation for mula is as follows:

$R_{X1}\left(\mathrm{\τ}\right)={F^{-1}}_{\mathrm{\τ}}\left[P_{\mathrm{SUB}}\left(\mathrm{\ω}\right)\right]={\∫}_{-\∞}^{+\∞}{P}_{\mathrm{SUB}}\left(\mathrm{\ω}\right)e^{\mathrm{j\ω\τ}}\mathrm{d\ω}$

Wherein, auto-correlation function R_X1(τ) it is the auto-correlation function after determining frequency AF panel；

Detection sub-module, for based on determining the auto-correlation function after frequency AF panel, uses time domain auto-correlation strategy real The detection of existing frequency spectrum cavity-pocket.

In a word, for the problems referred to above, the present invention, on the basis of fully analyzing time domain correlation method, utilizes power spectrum Opposition method improves time domain Autocorrelation Detection method, curbs and determines frequency interference and then improve frequency spectrum detection performance.

Compared with prior art, the present invention's it is a technical advantage that:

Determine the situation of frequency interference owing to the time domain auto-correlation algorithm of prior art does not accounts for happening suddenly, i.e. only considered folded Add the single signal of white Gaussian noise, thus accuracy of detection is the lowest.And the time domain Autocorrelation Detection improved in the present invention Method considers the burst fixed frequency interference impact on testing result, so that accuracy of detection is higher.

Accompanying drawing explanation

Fig. 1 is the flow chart of the method for the secondary user's detection cognitive radio frequency spectrum that the present invention provides；

Fig. 2 is to receive autocorrelation coefficient when signal contains only frequency hopping and noise and time delay；

Fig. 3 is to receive autocorrelation coefficient when signal contains frequency hopping, fixed frequency, noise and time delay；

Fig. 4 is the actual power spectrum of data intercept section；

Fig. 5 is to use the power spectrum after power spectrum opposition method suppression fixed frequency interference；

Fig. 6 is the power spectrum after suppressing fixed frequency interference to be carried out inverse Fourier transform seek auto-correlation function；

Performance comparision figure before and after Fig. 7 frequency AF panel surely.

Detailed description of the invention

With embodiment, the method for the invention is described in detail below in conjunction with the accompanying drawings.

The present invention uses power spectrum opposition method to improve time domain auto-correlation algorithm, and the fixed frequency of suppression burst disturbs.Power spectrum Opposition method is according to the time dependent diversity of Frequency Hopping Signal power spectrum, by reception data are carried out actual power spectrum With subtracting each other of average power spectra, balance out and determine frequency jamming power spectrum, retain Frequency Hopping Signal power spectrum.Pungent according to wiener Make by imperial order reason, the power spectrum retained is taken inverse Fourier transform and obtains auto-correlation function, and then use time domain auto-correlation to calculate Method goes detection containing frequency range during burst fixed frequency interference.Concrete principle is explained as follows:

The improvement principle of the time domain Autocorrelation Detection method that the present invention is based on is:

When assuming that receiving signal contains Frequency Hopping Signal, noise and fixed frequency interference, it is represented by

x(t)=S_H(t)+S_F(t)+n(t)(1)

Wherein, x (t) represents that time user receives signal.S_HT () represents that hop period is T_HFrequency Hopping Signal, during sampling Between T much larger than hop period (but being less than the frequency hop sequences cycle), thus ensure to sample the signal packet obtained containing multiple The information of Hopping frequencies.S_FT () represents the burst frequency interference surely that whole detection frequency range all exists.N (t) represents that average is 0, single-side belt power spectral density isBand lead to white Gaussian noise.The auto-correlation computation such as following formula of reception signal:

$\begin{array}{c}{R}_X\left(\mathrm{\τ}\right)=E\left[x\left(t\right)x(t+\mathrm{\τ})\right]\\ =R_{{\mathrm{SS}}_{\mathrm{HH}}}\left(\mathrm{\τ}\right)+R_{{\mathrm{SS}}_{\mathrm{FF}}}\left(\mathrm{\τ}\right)+R_{\mathrm{NN}}\left(\mathrm{\τ}\right)+R_{S_H{S}_F}\left(\mathrm{\τ}\right)+R_{S_F{S}_H}\left(\mathrm{\τ}\right)+R_{S_{H}N}\left(\mathrm{\τ}\right)\\ +R_{N{S}_H}\left(\mathrm{\τ}\right)+R_{S_{F}N}\left(\mathrm{\τ}\right)+R_{N{S}_F}\left(\mathrm{\τ}\right)\end{array}---\left(2\right)$

WhereinRepresent the Frequency Hopping Signal auto-correlation of primary user's transmission,Represent fixed frequency interference auto-correlation,Represent the cross-correlation between Frequency Hopping Signal, fixed frequency interference.Due to noise and frequency hopping with determine frequency Between cross-correlation can neglect to fall to disregarding.Then formula (2) is expressed as:

$R_X\left(\mathrm{\τ}\right)=R_{S{S}_{\mathrm{HH}}}\left(\mathrm{\τ}\right)+R_{S{S}_{\mathrm{FF}}}\left(\mathrm{\τ}\right)+R_{\mathrm{NN}}\left(\mathrm{\τ}\right)+R_{S_H{S}_F}\left(\mathrm{\τ}\right)+R_{S_F{S}_H}\left(\mathrm{\τ}\right)---\left(3\right)$

Owing to white Gaussian noise is incoherent in time domain, so the auto-correlation receiving signal depends on what frequency hopping was believed The auto-correlation that auto-correlation is disturbed with fixed frequency.Frequency Hopping Signal auto-correlationAt τ ＜ T_HTime, Frequency Hopping Signal is one It is relevant in jump space, its value non-zero.Owing to the frequency of Frequency Hopping Signal is not within adjacent several hop periods Identical, therefore at τ ＞ T_HTime, when i.e. frequency hopping number of times is more,Depending on frequency interference whole Therefore individual detection frequency range has the strongest autocorrelation, fixed frequency interference autocorrelation valueWork as primary user When containing only Frequency Hopping Signal with white Gaussian noise, according to receiving the auto-correlation of signal the most only a hop cycle time delay In the range of a bigger peak value occurs, and more mild and close to 0 this character in other time delay range, i.e. The detection to Frequency Hopping Signal existence can be realized.Now relation such as Fig. 2 institute of normalized autocorrelation coefficient and delay, τ Show, and when primary user contains Frequency Hopping Signal, fixed frequency interference and white Gaussian noise, disturb in whole inspection owing to determining frequency The relation of the autocorrelation that frequency measurement Duan Douyou is the strongest, its normalized autocorrelation coefficient and delay, τ is as shown in Figure 3.

When receiving signal without Frequency Hopping Signal, i.e. containing noise and fixed frequency interference, formula (3) is represented by

$R_X\left(\mathrm{\&tau;}\right)=R_{S{S}_{\mathrm{FF}}}\left(\mathrm{\&tau;}\right)+\mathrm{PR}\left(\mathrm{\&tau;}\right)|{\mathrm{\&sigma;}}^2=\mathrm{1,0}<\mathrm{\&tau;}<T---\left(4\right)$

When receiving signal packet containing Frequency Hopping Signal, i.e. containing Frequency Hopping Signal, noise and fixed frequency interference, formula (3) can represent For

$P_X\left(\mathrm{\&tau;}\right)=\left\{\begin{array}{cc}{R}_{{\mathrm{SS}}_{\mathrm{HH}}}\left(\mathrm{\&tau;}\right)+R_{{\mathrm{SS}}_{\mathrm{FF}}}\left(\mathrm{\&tau;}\right)+R_{S_H{S}_F}\left(\mathrm{\&tau;}\right)+R_{S_F{S}_H}\left(\mathrm{\&tau;}\right)+\mathrm{PR}\left(\mathrm{\&tau;}\right){|}_{{\mathrm{\&sigma;}}^2=1},& 0<\mathrm{\&tau;}<T_H\\ R_{{\mathrm{SS}}_{\mathrm{FF}}}\left(\mathrm{\&tau;}\right)+R_{S_H{S}_F}\left(\mathrm{\&tau;}\right)+R_{S_F{S}_H}\left(\mathrm{\&tau;}\right)+\mathrm{PR}\left(\mathrm{\&tau;}\right){|}_{{\mathrm{\&sigma;}}^2=1,}& T_H<\mathrm{\&tau;}<T\end{array}\right.---\left(5\right)$

Wherein P represents noise power, R (τ) | σ²=1 auto-correlation representing the white Gaussian noise that power is unit value.By Formula (4), (5) understand, even if containing Frequency Hopping Signal in detection frequency range, owing to the frequency of determining on each frequency disturbs self also Having good autocorrelation, therefore which type of value no matter delay, τ take, and the autocorrelation value receiving signal also will not become In 0.

Additionally, according to the auto-correlation function R of detection signal_X(τ) ratio calculating characteristic quantity and characteristic quantity is as follows:

$E_1=\frac{1}{T_H}{\&Integral;}_0^{T_H}\left|R_X\left(\mathrm{\&tau;}\right)\right|\mathrm{d\&tau;};E_2=\frac{1}{T-T_H}{\&Integral;}_{T_H}^T\left|R_X\left(\mathrm{\&tau;}\right)\right|\mathrm{d\&tau;}---\left(6\right)$

Time without Frequency Hopping Signal characteristic quantity ratio ρ₁It is expressed as

${\mathrm{\&rho;}}_1=\frac{E_1}{E_2}=\frac{(T-T_H){\&Integral;}_0^{T_H}|R_{{\mathrm{SS}}_{\mathrm{FF}}}\mathrm{\&tau;}+\mathrm{PR}\left(\mathrm{\&tau;}\right){|}_{{\mathrm{\&sigma;}}^2=1}|\mathrm{d\&tau;}}{T_H{\&Integral;}_{T_H}^T|R_{S{S}_{\mathrm{FF}}}\left(\mathrm{\&tau;}\right)+\mathrm{PR}\left(\mathrm{\&tau;}\right){|}_{{\mathrm{\&sigma;}}^2=1}|\mathrm{d\&tau;}}---\left(7\right)$

The ratio ρ of characteristic quantity when having Frequency Hopping Signal₂It is expressed as

${\mathrm{\&rho;}}_2=\frac{E_1}{E_2}=\frac{(T-T_H){\&Integral;}_0^{T_H}|R_{{\mathrm{SS}}_{\mathrm{HH}}}\left(\mathrm{\&tau;}\right)+R_{{\mathrm{SS}}_{\mathrm{FF}}}\left(\mathrm{\&tau;}\right)+R_{S_H{S}_F}\left(\mathrm{\&tau;}\right)+R_{S_F{S}_H}\left(\mathrm{\&tau;}\right)+\mathrm{PR}\left(\mathrm{\&tau;}\right){|}_{{\mathrm{\&sigma;}}^2=1}|\mathrm{d\&tau;}}{T_H{\&Integral;}_{T_H}^T|R_{S{S}_{\mathrm{FF}}}\left(\mathrm{\&tau;}\right)+R_{S_H{S}_F}\left(\mathrm{\&tau;}\right)+R_{S_F{S}_F}\left(\mathrm{\&tau;}\right)+\mathrm{PR}\left(\mathrm{\&tau;}\right){|}_{{\mathrm{\&sigma;}}^2=1}|\mathrm{d\&tau;}}---\left(8\right)$

In formula (7), owing to determining the existence that frequency disturbs, false-alarm probability can be caused to increase, time user so can be made to lose more The chances using frequency spectrum cavity-pocket, make the availability of frequency spectrum reduce more.Formula (8), exists owing to determining frequency interference, and its average is not There will be an obvious increment.

Shown in sum up, the improvement time domain Autocorrelation Detection method that the present invention provides, first to cognition wireless electrical receive signal Carry out actual power spectrum and mean power spectrum subtraction.Balance out and determine frequency interference, retain Frequency Hopping Signal.Next uses these The information retained carries out time domain Autocorrelation Detection, it is judged that with or without frequency spectrum cavity-pocket.

Embodiment

Assume that detecting frequency range is mainly made up of, owing to Frequency Hopping Signal belongs to non-flat Frequency Hopping Signal, noise and fixed frequency interference Steady signal.

First, one section of non-stationary signal is divided into several sections of stationary signals.

Assume the discrete series that non-stationary signal is a length of N, be expressed as:

$x_N\left(n\right)=\underset{i=1}{\overset{a}{\mathrm{\&Sigma;}}}{S}_{\mathrm{Fi}}\left(n\right)+\underset{j=1}{\overset{b}{\mathrm{\&Sigma;}}}{S}_{\mathrm{Hj}}\left(n\right)+n\left(n\right)---\left(9\right)$

Wherein, S_FiN () represents that a fixed frequency disturbs, i=1,2 ..., a；S_HjN () represents b Frequency Hopping Signal, j= 1,2,…,b；N (n) represents white Gaussian noise.Function x_NN the span of parameter n of () is: n=1,2 ..., N；

Non-stationary signal x by above-mentioned a length of N_NN () is divided into L section stationary signal, and each section of stationary signal A length of M, each section with x_MX represents；Respectively to each section of stationary signal x_MKTake FFT (Fast Fourier Transform) power spectrum power spectrum P is sought_xMK(n′)(ω):

$P_{\mathrm{xMK}\left(n^{\&prime;}\right)}\left(\mathrm{\&omega;}\right)=\frac{1}{M}{\left|{\mathrm{\&Sigma;}}_{n^{\&prime;}=0}^{M-1}{x}_{\mathrm{MK}}\left(n^{\&prime;}\right)e^{-\mathrm{j\&omega;}{n}^{\&prime;}}\right|}^2---\left(10\right);$

Wherein, the span of n ' is: n '=0,1,2 ... and, (M-1)；K=1,2,3,…,L；

Then, the power spectrum correspondence of every segment data is added and is averaged, the average power spectra of this data segment can be obtained, Specific formula for calculation is as follows:

${\stackrel{\&OverBar;}{P}}_{x_{N\left(n\right)}}\left(\mathrm{\&omega;}\right)=\frac{1}{L}\underset{K=1}{\overset{L}{\mathrm{\&Sigma;}}}{P}_{\mathrm{xMK}\left(n^{\&prime;}\right)}\left(\mathrm{\&omega;}\right)=\frac{1}{\mathrm{ML}}\underset{K=1}{\overset{L}{\mathrm{\&Sigma;}}}{\left|\underset{n^{\&prime;}=0}{\overset{M-1}{\mathrm{\&Sigma;}}}{x}_{\mathrm{MK}}\left(n^{\&prime;}\right)e^{-\mathrm{j\&omega;}{n}^{\&prime;}}\right|}^2---\left(11\right)$

Then, whole segment data is made FFT, calculate the power spectrum P of institute's intercept signal_xN(n)(ω), computing formula is as follows And with reference to figure (4):

$P_{\mathrm{xN}\left(n\right)}\left(\mathrm{\&omega;}\right)=\frac{1}{N}{\left|\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{x}_N\left(n\right)e^{-\mathrm{j\&omega;n}}\right|}^2---\left(12\right)$

Subsequently, use equation below to calculate and offset power spectrum P_SUB(ω), simultaneously as shown in figure (5), wherein 1, The frequency of 2 two occurrence frequency collisions the most corresponding.Compared with frequency corresponding with Fig. 4, after power spectrum offsets, The frequency hopping power spectrum of two frequency bins has reduced, but the impact on detection performance is little, and does not has determining of frequency hit Frequency interference is suppressed.

$P_{\mathrm{SUB}}\left(\mathrm{\&omega;}\right)=P_{\mathrm{xN}\left(n\right)}\left(\mathrm{\&omega;}\right)-{\stackrel{\&OverBar;}{P}x}_{N\left(n\right)}\left(\mathrm{\&omega;}\right)---\left(13\right)$

In a word, most fixed frequent spectrum information can be balanced out by computing above, retain most of frequency hopping frequency Spectrum information.

Finally, according to wiener-khintchine theorem, auto-correlation function and power spectrum are a pair Fourier transform.Right P_SUB(ω) carry out inverse Fourier transform and obtain auto-correlation function R_X1(τ), computing formula is following the most as shown in Figure 6.

$P_{\mathrm{SUB}}\left(\mathrm{\&omega;}\right)=F_{\mathrm{\&tau;}}\left[R_{X1}\left(\mathrm{\&tau;}\right)\right]={\&Integral;}_{-\&infin;}^{+\&infin;}{R}_{X1}\left(\mathrm{\&tau;}\right)e^{-\mathrm{j\&omega;\&tau;}}\mathrm{d\&omega;}---\left(14\right)$

$R_{X1}\left(\mathrm{\&tau;}\right)={F^{-1}}_{\mathrm{\&tau;}}\left[P_{\mathrm{SUB}}\left(\mathrm{\&omega;}\right)\right]={\&Integral;}_{-\&infin;}^{+\&infin;}{P}_{\mathrm{SUB}}\left(\mathrm{\&omega;}\right)e^{\mathrm{j\&omega;\&tau;}}\mathrm{d\&omega;}---\left(15\right)$

Wherein, R in above formula_X1(τ) represent the auto-correlation after fixed frequency AF panel, obtain determining frequency by formula (15) and disturb Auto-correlation after suppression, the time domain autocorrelation method that recycling is commonly used realizes the detection of frequency spectrum cavity-pocket.

At the same P of false-alarm probability_fIn the case of=0.05, to the time domain of existing time domain Autocorrelation Detection method and improvement from Correlation detection carries out Performance comparision, as shown in Figure 7.

It should be noted last that, above example is only in order to illustrate technical scheme and unrestricted.Although With reference to embodiment, the present invention is described in detail, it will be understood by those within the art that, to the present invention Technical scheme modify or equivalent, without departure from the spirit and scope of technical solution of the present invention, it is equal Should contain in the middle of scope of the presently claimed invention.

Claims (4)

1. a method for secondary user's detection cognitive radio frequency spectrum, described method comprises:

Step 101) according to the time dependent diversity of Frequency Hopping Signal power spectrum, calculate secondary user's certain period The actual power spectrum of interior reception data and the difference of average power spectra, thus offset fixed frequency jamming power spectrum and retain jumping Frequently power spectrum signal；

Step 102) the Frequency Hopping Signal power spectrum retained is carried out inverse Fourier transform is obtained auto-correlation function, then make With the detection of time domain auto-correlation algorithm containing frequency range during burst fixed frequency interference；

Wherein, described step 101) comprise further:

Step 101-1) intercept the non-stationary signal that secondary user's received within certain period, and the non-stationary letter intercepted Number it is the discrete data sequences of a length of N, is embodied as:

$x_N\left(n\right)=\underset{i=1}{\overset{a}{\&Sigma;}}{S}_{Fi}\left(n\right)+\underset{j=1}{\overset{b}{\&Sigma;}}{S}_{Hj}\left(n\right)+n\left(n\right)$

Wherein, S_FiN () represents that a fixed frequency disturbs, i=1,2 ..., a；S_HjN () represents b Frequency Hopping Signal, j= 1,2,…,b；N (n) represents white Gaussian noise；Function x_NN the span of parameter n of () is: n=0,1,2 ..., N- 1；

Step 101-2) non-stationary signal of above-mentioned a length of N is divided into L section stationary signal, and each section is put down The a length of M of steady signal, each section with x_MKRepresent；

Step 101-3) use equation below to calculate each section of stationary signal x_MKPower spectrum P_xMK(n′)(ω):

$P_{xMK\left(n^{\&prime;}\right)}\left(\mathrm{\&omega;}\right)=\frac{1}{M}|{\mathrm{\&Sigma;}}_{n^{\&prime;}=0}^{M-1}{x}_{MK}\left(n^{\&prime;}\right)e^{-{\mathrm{j\&omega;n}}^{\&prime;}}{|}^2;$

Wherein, the span of n ' is: n '=0,1,2 ..., (M-1)；K=1,2,3 ..., L；

Step 101-3) according to the meansigma methods of the power spectrum of the calculated all each section of stationary signals of equation below:

${\stackrel{\&OverBar;}{P}}_{x_{N\left(n\right)}}\left(\mathrm{\&omega;}\right)=\frac{1}{L}\underset{K=1}{\overset{L}{\&Sigma;}}{P}_{xMK\left(n^{\&prime;}\right)}\left(\mathrm{\&omega;}\right)=\frac{1}{ML}\underset{K=1}{\overset{L}{\&Sigma;}}|\underset{n^{\&prime;}=0}{\overset{M-1}{\&Sigma;}}{x}_{MK}\left(n^{\&prime;}\right)e^{-{\mathrm{j\&omega;n}}^{\&prime;}}{|}^2$

Step 101-4) according to equation below, to step 101-1) secondary user's intercept whole section of non-stationary signal enter Row Fourier transformation, and then obtain the power spectrum of institute's intercept signal；

$P_{xN\left(n\right)}\left(\mathrm{\&omega;}\right)=\frac{1}{N}|\underset{n=0}{\overset{N-1}{\&Sigma;}}{x}_N\left(n\right)e^{-j\mathrm{\&omega;}n}{|}^2$

Step 101-5) mean value calculation of power spectrum of power spectrum based on intercept signal and all each section of stationary signals Offseting power spectrum, obtain retaining Frequency Hopping Signal power spectrum, specific formula for calculation is as follows:

Described step 102) comprise further:

Step 102-1) to retaining Frequency Hopping Signal power spectrum P_SUB(ω) carry out inverse Fourier transform and obtain auto-correlation function R_X1(τ), concrete transformation for mula is as follows:

$R_{X1}\left(\mathrm{\&tau;}\right)={F^{-1}}_{\mathrm{\&tau;}}\&lsqb;P_{SUB}\left(\mathrm{\&omega;}\right)\&rsqb;={\&Integral;}_{-\mathrm{\&infin;}}^{+\mathrm{\&infin;}}{P}_{SUB}\left(\mathrm{\&omega;}\right)e^{j\mathrm{\&omega;}\mathrm{\&tau;}}d\mathrm{\&omega;}$

Wherein, auto-correlation function R_X1(τ) it is the auto-correlation function after determining frequency AF panel；

Step 102-2) based on determining the auto-correlation function after frequency AF panel, use time domain auto-correlation strategy to realize frequency spectrum The detection in cavity.

The method of secondary user's the most according to claim 1 detection cognitive radio frequency spectrum, it is characterised in that Described step 102-2) comprise further:

Step 102-2-1) it is calculated as follows two characteristic quantity E according to the auto-correlation function obtained₁And E₂；

$E_1=\frac{1}{T_H}{\&Integral;}_0^{T_H}\left|R_{X1}\left(\mathrm{\&tau;}\right)\right|d\mathrm{\&tau;}$

$E_2=\frac{1}{T-T_H}{\&Integral;}_{T_H}^T\left|R_{X1}\left(\mathrm{\&tau;}\right)\right|d\mathrm{\&tau;}$

Step 102-2-2) calculate above-mentioned two characteristic quantity ratio ρ；

Step 102-2-3) ρ obtained is compared size with the threshold value of setting, then exist when ρ is more than threshold value Frequency Hopping Signal, does not otherwise have Frequency Hopping Signal.

The method of secondary user's the most according to claim 2 detection cognitive radio frequency spectrum, it is characterised in that By estimating ρ₁Expectation and variance yields, and combine probability density distribution situation and the void of normally distributed random variable Alarm probability determines suitable threshold value, described ρ₁For only having characteristic quantity ratio during noise without Frequency Hopping Signal.

4. the system of a secondary user's detection cognitive radio frequency spectrum, it is characterised in that described system comprises:

Frequency Hopping Signal power spectrum extraction module, for according to the time dependent diversity of Frequency Hopping Signal power spectrum, calculates Secondary user's receives the actual power spectrum of data and the difference of average power spectra within certain period, and then balances out calmly Frequently jamming power is composed and is retained Frequency Hopping Signal power spectrum；

Time domain auto-correlation function obtains and detection module, for the Frequency Hopping Signal power spectrum retained is taken inverse Fourier transform Obtain auto-correlation function, and then use the detection of time domain auto-correlation algorithm containing frequency range during burst fixed frequency interference；

Wherein, described Frequency Hopping Signal power spectrum extraction module comprises further:

Intercepting submodule, for intercepting the non-stationary signal that secondary user's received within certain period, and intercepting is non- Stationary signal is the discrete data sequences of a length of N, is embodied as:

$x_N\left(n\right)=\underset{i=1}{\overset{a}{\&Sigma;}}{S}_{Fi}\left(n\right)+\underset{j=1}{\overset{b}{\&Sigma;}}{S}_{Hj}\left(n\right)+n\left(n\right)$

Wherein, S_FiN () represents that a fixed frequency disturbs, i=1,2 ..., a；S_HjN () represents b Frequency Hopping Signal, j= 1,2,…,b；N (n) represents white Gaussian noise；Function x_NN the span of parameter n of () is: n=0,1,2 ..., N- 1；

Segmentation submodule, for the non-stationary signal of a length of N is divided into L section stationary signal, and each section is put down The a length of M of steady signal, each section with x_MKRepresent；

Each section of stationary signal spectra calculation submodule, is used for using equation below to calculate each section of stationary signal x_MK's Power spectrum P_xMK(n′)(ω):

$P_{xMK\left(n^{\&prime;}\right)}\left(\mathrm{\&omega;}\right)=\frac{1}{M}|{\mathrm{\&Sigma;}}_{n^{\&prime;}=0}^{M-1}{x}_{MK}\left(n^{\&prime;}\right)e^{-{\mathrm{j\&omega;n}}^{\&prime;}}{|}^2;$

Wherein, the span of n ' is: n '=0,1,2 ..., (M-1)；K=1,2,3 ..., L；

Mean power obtains submodule, for the power according to the calculated all each section of stationary signals of equation below The meansigma methods of spectrum:

${\stackrel{\&OverBar;}{P}}_{x_{N\left(n\right)}}\left(\mathrm{\&omega;}\right)=\frac{1}{L}\underset{K=1}{\overset{L}{\&Sigma;}}{P}_{xMK\left(n^{\&prime;}\right)}\left(\mathrm{\&omega;}\right)=\frac{1}{ML}\underset{K=1}{\overset{L}{\&Sigma;}}|\underset{n^{\&prime;}=0}{\overset{M-1}{\&Sigma;}}{x}_{MK}\left(n^{\&prime;}\right)e^{-{\mathrm{j\&omega;n}}^{\&prime;}}{|}^2$

Intercept signal power spectrum obtains submodule, for according to equation below, the secondary user's to this intercepting submodule The whole section of non-stationary signal intercepted carries out Fourier transformation, and then obtains the power spectrum of institute's intercept signal；

$P_{xN\left(n\right)}\left(\mathrm{\&omega;}\right)=\frac{1}{N}|\underset{n=0}{\overset{N-1}{\&Sigma;}}{x}_N\left(n\right)e^{-j\mathrm{\&omega;}n}{|}^2$

Process submodule, for power spectrum average of power spectrum based on intercept signal and all each section of stationary signals Value calculating offsets power spectrum, obtains retaining Frequency Hopping Signal power spectrum, and specific formula for calculation is as follows:

$P_{SUB}\left(\mathrm{\&omega;}\right)=P_{xN\left(n\right)}\left(\mathrm{\&omega;}\right)-\stackrel{\&OverBar;}{P}{x}_{N\left(n\right)}\left(\mathrm{\&omega;}\right);$

Described time domain auto-correlation function obtains and detection module comprises further:

Auto-correlation function obtains submodule, for retaining Frequency Hopping Signal power spectrum P_SUB(ω) inverse Fourier transform is carried out Obtain auto-correlation function R_X1(τ), concrete transformation for mula is as follows:

$R_{X1}\left(\mathrm{\&tau;}\right)={F^{-1}}_{\mathrm{\&tau;}}\&lsqb;P_{SUB}\left(\mathrm{\&omega;}\right)\&rsqb;={\&Integral;}_{-\mathrm{\&infin;}}^{+\mathrm{\&infin;}}{P}_{SUB}\left(\mathrm{\&omega;}\right)e^{j\mathrm{\&omega;}\mathrm{\&tau;}}d\mathrm{\&omega;}$

Wherein, auto-correlation function R_X1(τ) it is the auto-correlation function after determining frequency AF panel；

Detection sub-module, for based on determining the auto-correlation function after frequency AF panel, uses time domain auto-correlation strategy real The detection of existing frequency spectrum cavity-pocket.