CN103973380B - The feedback stacks energy detection method that a kind of user of solution arrives at random - Google Patents

Abstract

The invention discloses the feedback stacks energy detection method that a kind of user of solution arrives at random, comprise step: target setting false alarm probability t detection time of feedback stacks energy detection algorithm _sens, and the sampling frequency f of feedback stacks energy detection algorithm _s; Calculate decision threshold η ˊ; Calculate the statistical decision amount Y of feedback stacks energy detection algorithm; If Y ≥ η is ˊ, adjudicate as H ₁user (primary user or secondary user) is there is in channel; If Y & is lt; η ˊ, adjudicates as H ₀user (primary user or secondary user) is there is not in channel; The present invention is by being added on the instantaneous energy of sense cycle first half sampled point by the instantaneous energy value of energy detection algorithm sense cycle latter half sampled point, under the prerequisite not extending detection time, improve the energy proportion of subscriber signal in energy detection algorithm statistical decision amount.The detection perform of cognitive user can be improved, reduce the probability that in cognitive radio system, user data collides, thus improve SU data throughout.

Description

The feedback stacks energy detection method that a kind of user of solution arrives at random

Technical field

The present invention relates to energy detection technique in cognitive radio technology, particularly relate to the feedback stacks energy detection method that a kind of user of solution arrives at random.

Background technology

Cognitive radio (CognitiveRadio, CR), as a kind of dynamic frequency spectrum huge profit technology, can improve the availability of frequency spectrum, effectively utilize rare frequency spectrum resource.In CR technology, energy measuring (EnergyDetection, the ED) advantages such as simple and complexity is low with it, are widely used.Current great majority are all based on authorized user/primary user (PrimaryUser about the research that idle frequency spectrum detects, PU) communications status carries out under detection-phase does not change this hypothesis, namely suppose within the whole detection period, PU or existence, or do not exist.But in actual CR environment, the active state of PU probably changes between detection period, especially under low signal-to-noise ratio (Signal-to-NoiseRatio, SNR) scene, improve detection perform by extending detection time.

In consideration PU traffic carrying capacity to cognitive user/time user (SecondaryUser, SU) when detection perform affects, when SU performs frequency spectrum detection in the scene that PU traffic carrying capacity is larger, in the detection of SU, PU may occur suddenly, now the sense cycle of SU has just become first half only containing white Gaussian noise (AdditiveWhiteGaussianNoise, AWGN) signal, and at latter half containing the mixed signal [TangL having plenty of PU signal and AWGN signal, ChenY, HinesEL, etal.Effectofprimaryusertrafficonsensing-throughputtrade offforcognitiveradios [J] .IEEETransactionsonWirelessCommunications, 2011, 10 (4): 1063 – 1068.], as Fig. 1.In acentric distributed cognition wireless network scenario, when SU traffic carrying capacity is larger, when SU performs frequency spectrum detection, the sense cycle of SU will be subject to the impact of other SU traffic carrying capacitys, namely within very little detection time, have multiple SU and detect same PU channel [XieXZ simultaneously, HuXF, YangHL, etal.CognitiveMACProtocolwiththeMinimumSamplingTimeandCr oss-LayerCooperationforLowSNREnvironments [J] .ChinaCommunications, 2013,10 (12): 125 – 138.], as Fig. 2.SU in Fig. 2 ₀at t ₁moment carries out perception, at t to PU frequency range ₃moment completes to be detected and detects that PU is in idle condition, so SU ₀at t ₃moment starts the data transmitting oneself, and in figure, part azury represents SU ₀data; But SU ₀start to detect soon, SU also arrives network, at t ₂moment starts to carry out perception to PU frequency range, and this just causes in the sense cycle of SU, at [t ₂, t ₃] in the time period only containing AWGN signal, at [t ₃, t ₄] in the time period containing SU ₀the mixed signal of signal and AWGN signal.

Be not difficult to find out from Fig. 1 and Fig. 2, when execution frequency spectrum detection, there is not subscriber signal before this in current SU in sense cycle, but after being through the sampling of a period of time, detected frequency range contains subscriber signal (may be PU also may be SU).Can obtain the sense cycle sampled point model of SU as shown in Figure 3, corresponding ED statistical decision amount is:

$Y={\mathrm{\Σ}}_{i=1}^{I-a}{n}_i^2+{\mathrm{\Σ}}_{i=I-a+1}^I{(s_i+n_i)}^2---\left(1\right)$

Wherein, s _ifor subscriber signal sampling in detected frequency range, n _ifor noise samples.A represents the sampled point number containing subscriber signal at SU sense cycle latter half, and 0<a≤I, I are the total number of samples of ED, are numerically the product T of ED detection time and sample frequency _sensf _s.A=f can be obtained _sτ, 0< τ≤T _sens, τ=t ₄-t ₃for the detected user data sampling time.If in CRN, the detection time of SU is identical, so τ=t ₄-t ₃=t ₂-t ₁.When I is larger, according to central-limit theorem, can calculate under Fig. 3 model by formula (1), it is I+a γ that the decision statistics of SU takes Y from average, and variance is the Gaussian Profile of 2I+4a γ.So the detection probability of SU is:

$P_{d\_\mathrm{SU}}=P_r\{Y>\mathrm{\&eta;}\}=Q\left(\frac{\mathrm{\&eta;}-I-\mathrm{a\&gamma;}}{\sqrt{2I+4\mathrm{a\&gamma;}}}\right)---\left(2\right)$

η is the decision threshold of traditional E D, and γ is the SNR that SU place receives subscriber signal.

Summary of the invention

For above deficiency of the prior art, a kind of user of reduction is the object of the present invention is to provide to arrive impact on energy measuring performance at random, when not needing to extend detection time, improving detection perform, improving the feedback stacks energy detection method of the data throughout of SU simultaneously.Technical scheme of the present invention is as follows: the feedback stacks energy detection method that a kind of user of solution arrives at random, and it comprises the following steps:

101, the parameter of Initialize installation feedback stacks energy detection algorithm, comprising: target false alarm probability t detection time of feedback stacks energy detection algorithm _sens, and the sample frequency f of feedback stacks energy detection algorithm _s, draw the sampled point number I=T that feedback stacks energy detection algorithm is total _sensf _s;

102, the target false alarm probability of setting in step 101 is utilized with total sampled point number I, according to formula calculate the decision threshold of feedback stacks energy detection algorithm, wherein Q ^-1() is the inverse function of standard normal cumulative distribution function;

103, at T detection time _sensin, the signal in channel is sampled, obtains I sampled point, be designated as y _i, wherein sampled point sequence number i be 1,2 ... I/2, I/2+1 ..., I, calculate the instantaneous energy of each sampled point by rear half period also namely sampled point sequence number be that the sampled point instantaneous energy of I/2 ~ I is added to front half period, also namely sampled point sequence number is on the sampled point instantaneous energy of 1 ~ I/2, and the statistical decision amount Y expression formula of feedback stacks energy detection algorithm is: wherein it is the instantaneous energy of i-th sampled point;

104, the decision threshold η ' that the energy statistics judgement amount Y obtained in step 103 and step 102 obtain is compared, if Y >=η ', then be judged as in channel, there is primary user or secondary user, if Y< is η ', then be judged as in channel, there is not primary user and time user, complete the detection to channel.

When the energy statistics judgement amount Y in step 103 is at sense cycle T _senswhen inside there is not primary user or secondary subscriber signal, then wherein n _ifor noise samples, the theoretical false alarm probability that can obtain feedback stacks energy detection algorithm is consider the emergent situation of user in sense cycle when there is subscriber signal, energy statistics judgement amount Y is simultaneously:

$Y=\left\{\begin{array}{cc}\underset{i=1}{\overset{I/2}{\mathrm{\&Sigma;}}}{n}_i^2+2\underset{i=I/2+1}{\overset{I-a}{\mathrm{\&Sigma;}}}{n}_i^2+2\underset{i=I-a+1}{\overset{I}{\mathrm{\&Sigma;}}}{(s_i+n_i)}^2& 0<a<I/2\\ \underset{i=1}{\overset{I-a}{\mathrm{\&Sigma;}}}{n}_i^2+2\underset{i=I/2+1}{\overset{I}{\mathrm{\&Sigma;}}}{(s_i+n_i)}^2+\underset{i=I-a+1}{\overset{I/2}{\mathrm{\&Sigma;}}}{(s_i+n_i)}^2& I/2\&le;a\&le;I\end{array}\right.,$ Wherein s _ifor subscriber signal sampling in detected frequency range, n _ifor noise samples, a represents the sampled point number including subscriber signal at sense cycle latter half, 0<a≤I, and the detection probability obtaining feedback stacks energy detection algorithm is

$P_d=\left\{\begin{array}{cc}Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+2\mathrm{a\&gamma;})}{\sqrt{5I+16\mathrm{a\&gamma;}}}\right)& 0<a<I/2\\ Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+\mathrm{a\&gamma;}+0.5\mathrm{I\&gamma;})}{\sqrt{5I+4\mathrm{a\&gamma;}+6\mathrm{I\&gamma;}}}\right)& I/2\&le;a\&le;I\end{array}\right.,$ γ is subscriber signal signal to noise ratio.

The performance evaluation of feedback stacks energy detection algorithm:

The present invention proposes a kind of feedback stacks energy detection method, by the instantaneous energy value of sense cycle latter half sampled point is added on the instantaneous energy of sense cycle first half sampled point, thus under the prerequisite not extending detection time, improve the energy proportion of subscriber signal in the statistical decision amount of feedback stacks energy detection algorithm.Emulation shows, the feedback stacks case of energy detection schemes adopting the present invention to propose carries out frequency spectrum detection, can improve the detection perform of SU, reduces the probability that in CR system, user data collides, thus improves user data-throughput

Accompanying drawing explanation

Fig. 1 is the impact of SU sense cycle by PU traffic carrying capacity;

The multiple SU of Fig. 2 detects same PU channel model;

Fig. 3 arrives at random by user to be affected, SU sense cycle sampled point model;

The Cleaning Principle of Fig. 4 feedback stacks energy detection algorithm;

The overhaul flow chart of Fig. 5 feedback stacks energy detection algorithm;

The detection probability of Fig. 6 tetra-kinds of ED compares, SNR=-5dB;

The detection probability of Fig. 7 tetra-kinds of ED compares, SNR=-12dB;

Fig. 8 under same detection performance and different user traffic case, data collision probability with change;

Fig. 9 is under same detection performance and different user traffic case, and SU data throughout is along with the change of SU frame length;

Under Figure 10 different user traffic case SU data throughout along with change.

Embodiment

The invention will be further elaborated to provide the embodiment of an indefiniteness below in conjunction with accompanying drawing.

Shown in Fig. 1-Figure 10, the feedback stacks energy detection method that a kind of user of solution arrives at random, it comprises the following steps:

101, the parameter of Initialize installation feedback stacks energy detection algorithm, comprising: target false alarm probability t detection time of feedback stacks energy detection algorithm _sens, and the sample frequency f of feedback stacks energy detection algorithm _s, draw the sampled point number I=T that feedback stacks energy detection algorithm is total _sensf _s;

102, the target false alarm probability of setting in step 101 is utilized with total sampled point number I, according to formula calculate the decision threshold of feedback stacks energy detection algorithm, wherein Q ^-1() is the inverse function of standard normal cumulative distribution function;

103, at T detection time _sensin, the signal in channel is sampled, obtains I sampled point, be designated as y _i, wherein sampled point sequence number i be 1,2 ... I/2, I/2+1 ..., I, calculate the instantaneous energy of each sampled point by rear half period also namely sampled point sequence number be that the sampled point instantaneous energy of I/2 ~ I is added to front half period, also namely sampled point sequence number is on the sampled point instantaneous energy of 1 ~ I/2, and the statistical decision amount Y expression formula of feedback stacks energy detection algorithm is: wherein it is the instantaneous energy of i-th sampled point;

104, the decision threshold η ' that the energy statistics judgement amount Y obtained in step 103 and step 102 obtain is compared, if Y >=η ', then be judged as in channel, there is primary user or secondary user, if Y< is η ', then be judged as in channel, there is not primary user and time user, complete the detection to channel.

Preferably, when the energy statistics judgement amount Y in step 103 is at sense cycle T _senswhen inside there is not primary user or secondary subscriber signal, then wherein n _ifor noise samples, the theoretical false alarm probability that can obtain feedback stacks energy detection algorithm is consider the emergent situation of user in sense cycle when there is subscriber signal, energy statistics judgement amount Y is simultaneously: $Y=\left\{\begin{array}{cc}\underset{i=1}{\overset{I/2}{\mathrm{\&Sigma;}}}{n}_i^2+2\underset{i=I/2+1}{\overset{I-a}{\mathrm{\&Sigma;}}}{n}_i^2+2\underset{i=I-a+1}{\overset{I}{\mathrm{\&Sigma;}}}{(s_i+n_i)}^2& 0<a<I/2\\ \underset{i=1}{\overset{I-a}{\mathrm{\&Sigma;}}}{n}_i^2+2\underset{i=I/2+1}{\overset{I}{\mathrm{\&Sigma;}}}{(s_i+n_i)}^2+\underset{i=I-a+1}{\overset{I/2}{\mathrm{\&Sigma;}}}{(s_i+n_i)}^2& I/2\&le;a\&le;I\end{array}\right.,$ Wherein s _ifor subscriber signal sampling in detected frequency range, n _ifor noise samples, a represents the sampled point number including subscriber signal at sense cycle latter half, 0<a≤I, and the detection probability obtaining feedback stacks energy detection algorithm is $P_d=\left\{\begin{array}{cc}Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+2\mathrm{a\&gamma;})}{\sqrt{5I+16\mathrm{a\&gamma;}}}\right)& 0<a<I/2\\ Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+\mathrm{a\&gamma;}+0.5\mathrm{I\&gamma;})}{\sqrt{5I+4\mathrm{a\&gamma;}+6\mathrm{I\&gamma;}}}\right)& I/2\&le;a\&le;I\end{array}\right.,$ γ is subscriber signal signal to noise ratio.

1. feedback stacks energy detection algorithm false alarm probability and detection probability analysis

When there is not subscriber signal in sense cycle, use H ₀represent, the statistical decision amount of feedback stacks energy detection algorithm is:

$Y=\underset{i=1}{\overset{I/2}{\mathrm{\&Sigma;}}}(n_i^2+n_{I-i+1}^2)+\underset{i=I/2+1}{\overset{I}{\mathrm{\&Sigma;}}}{n}_i^2---\left(3\right)$

Included subscriber signal when detection time, consider the emergent scene of user, the detection model namely shown in Fig. 3, uses H simultaneously ₁represent, the statistical decision amount of feedback stacks energy detection algorithm is:

$Y=\left\{\begin{array}{cc}\underset{i=1}{\overset{I/2}{\mathrm{\&Sigma;}}}{n}_i^2+2\underset{i=I/2+1}{\overset{I-a}{\mathrm{\&Sigma;}}}{n}_i^2+2\underset{i=I-a+1}{\overset{I}{\mathrm{\&Sigma;}}}{(s_i+n_i)}^2& 0<a<I/2\\ \underset{i=1}{\overset{I-a}{\mathrm{\&Sigma;}}}{n}_i^2+2\underset{i=I/2+1}{\overset{I}{\mathrm{\&Sigma;}}}{(s_i+n_i)}^2+\underset{i=I-a+1}{\overset{I/2}{\mathrm{\&Sigma;}}}{(s_i+n_i)}^2& I/2\&le;a\&le;I\end{array}\right.---\left(4\right)$

When I is larger, the statistical decision amount that can be obtained feedback stacks energy detection algorithm by central-limit theorem is similar to Normal Distribution, and corresponding false alarm probability and detection probability are respectively:

$P_f=P_r(Y>{\mathrm{\&eta;}}^{\&prime;}|H_0)=Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-1.5I}{\sqrt{5I}}\right)---\left(5\right)$

$\begin{array}{c}{P}_d=P_r(Y>{\mathrm{\&eta;}}^{\&prime;}|H_1)=\left\{\begin{array}{cc}Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+2\mathrm{a\&gamma;})}{\sqrt{5I+16\mathrm{a\&gamma;}}}\right)& 0<a<I/2\\ Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+\mathrm{a\&gamma;}+0.5\mathrm{I\&gamma;})}{\sqrt{5I+4\mathrm{a\&gamma;}+6\mathrm{I\&gamma;}}}\right)& I/2\&le;a\&le;I\end{array}\right.\\ =\left\{\begin{array}{cc}Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+2f_s\mathrm{\&tau;\&gamma;})}{\sqrt{5I+16f_s\mathrm{\&tau;\&gamma;}}}\right)& 0<\mathrm{\&tau;}<T_{\mathrm{sens}}/2\\ Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+f_s\mathrm{\&tau;\&gamma;}+0.5\mathrm{I\&gamma;})}{\sqrt{5I+4f_s\mathrm{\&tau;\&gamma;}+6\mathrm{I\&gamma;}}}\right)& T_{\mathrm{sens}}/2\&le;\mathrm{\&tau;}\&le;\end{array}\right.\end{array}---\left(6\right)$

Wherein Q () is standard just too cumulative distribution function, according to Neyman-Pearson criterion, adopt constant false alarm probability [WangP, LiHB, andHimedB.AParametricMovingTargetDetectorforDistributedM IMORadarinNon-HomogeneousEnvironment [J] .IEEETransactionsonSignalProcessing, 2013,61 (9): 2282 – 2294.], the decision threshold of feedback stacks energy detection algorithm can be obtained by formula (5) ${\mathrm{\&eta;}}^{\&prime;}=\sqrt{5I}{Q}^{-1}\left(P_f^{\mathrm{DES}}\right)+1.5I.$

2. feedback stacks energy detection algorithm data collision and throughput analysis

Analyze the data collision probability of feedback stacks energy detection algorithm, first its minimal sampling time (theMinimumSamplingTime will be calculated, MST) [XieXZ, HuXF, YangHL, etal.CognitiveMACProtocolwiththeMinimumSamplingTimeandCr oss-LayerCooperationforLowSNREnvironments [J] .ChinaCommunications, 2013,10 (12): 125 – 138.], namely calculate satisfied following optimization problem:

$a_0=\underset{a}{\arg }\left\{\min \left(a\right)\right|P_d\&GreaterEqual;P_d^{\mathrm{DES}}\}---\left(7\right)$

Wherein for CR target detection probability.The minimum a that can be met formula (7) is:

$a_0=\frac{1}{2{\mathrm{\&gamma;}}^2}\left\{\begin{array}{c}-\{3\mathrm{I\&gamma;}+I{\mathrm{\&gamma;}}^2-2{\mathrm{\&eta;}}^{\&prime;}\mathrm{\&gamma;}-4\mathrm{\&gamma;}{\left[Q^{-1}\left(P_d^{\mathrm{DES}}\right)\right]}^2\}+\\ \sqrt{\begin{array}{c}{\{3\mathrm{I\&gamma;}+I{\mathrm{\&gamma;}}^2-2{\mathrm{\&eta;}}^{\&prime;}\mathrm{\&gamma;}-4\mathrm{\&gamma;}{\left[Q^{-1}\left(P_d^{\mathrm{DES}}\right)\right]}^2\}}^2-\\ {4\mathrm{\&gamma;}}^2\{{({\mathrm{\&eta;}}^{\&prime;}-1.5I-1.5\mathrm{I\&gamma;})}^2-(5I+6\mathrm{I\&gamma;}){\left[Q^{-1}\left(P_d^{\mathrm{DES}}\right)\right]}^2\}\end{array}}\end{array}\right\}---\left(8\right)$

The MST that can obtain feedback stacks energy detection algorithm is:

τ ₀'＝a ₀/f _s(9)

Suppose that SU arrives network compliant Poisson distribution, between so each user, arrive time interval obeys index distribution f (the τ)=λ e of network ^{-λ τ}, τ >0, wherein λ is subscriber traffic intensity or is referred to as arrival rate.Based on MST, in the scene that subscriber traffic is larger, the average probability that SU and user data transmission collide is:

$\begin{array}{c}{P}_{\mathrm{col}}={\&Integral;}_0^{{{\mathrm{\&tau;}}_0}^{\&prime;}}[1-P_d\left(\mathrm{\&tau;}\right)]f\left(\mathrm{\&tau;}\right)\mathrm{d\&tau;}\\ ={\&Integral;}_0^{T_{\mathrm{sesn}/2}}[1-Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+2f_s\mathrm{\&tau;\&gamma;})}{\sqrt{5I+16f_s\mathrm{\&tau;\&gamma;}}}\right)]{\mathrm{\&lambda;}e}^{-\mathrm{\&lambda;\&tau;}}\mathrm{d\&tau;}\\ +{\&Integral;}_{T_{\mathrm{sean}/2}}^{{{\mathrm{\&tau;}}_0}^{\&prime;}}[1-Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}(1.5I+f_s\mathrm{\&tau;\&gamma;}+0.5\mathrm{I\&gamma;})}{\sqrt{5I+4f_s\mathrm{\&tau;\&gamma;}+6\mathrm{I\&gamma;}}}\right)]{\mathrm{\&lambda;e}}^{-\mathrm{\&lambda;\&tau;}}\mathrm{d\&tau;}\end{array}---\left(10\right)$

The corresponding SU normalization data throughput that can obtain is:

$R(T_{\mathrm{frame}},P_{\mathrm{col}})=\frac{T_{\mathrm{frame}}-T_{\mathrm{sens}}}{T_{\mathrm{frame}}}\&CenterDot;[1-P_{\mathrm{col}}]$

$=\frac{T_{\mathrm{frame}}-T_{\mathrm{sens}}}{T_{\mathrm{frame}}}\&CenterDot;\left[\begin{array}{c}1-{\&Integral;}_0^{T_{\mathrm{sesn}/2}}[1-Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+2f_s\mathrm{\&tau;\&gamma;})}{\sqrt{5I+16f_s\mathrm{\&tau;\&gamma;}}}\right)]{\mathrm{\&lambda;e}}^{-\mathrm{\&lambda;\&tau;}}\mathrm{d\&tau;}\\ -{\&Integral;}_{T_{\mathrm{sesn}/2}}^{{{\mathrm{\&tau;}}_0}^{\&prime;}}[1-Q\frac{{\mathrm{\&eta;}}^{\&prime;}(1.5I+f_s\mathrm{\&tau;\&gamma;}+0.5\mathrm{I\&gamma;})}{\sqrt{5I+4f_s\mathrm{\&tau;\&gamma;}+6\mathrm{I\&gamma;}}}]{\mathrm{\&lambda;e}}^{-\mathrm{\&lambda;\&tau;}}\mathrm{d\&tau;}\end{array}\right]---\left(11\right)$

With reference to the detailed overhaul flow chart giving feedback stacks energy detection algorithm in figure 5.

If subscriber signal adopts BPSK modulation, be operated in television spectrum frequency range, carrier frequency is 500MHz, sample frequency f _sfor 6MHz.The target false alarm probability of CR be 0.1.User arrives network compliant Poisson distribution, utilize MATLAB to carry out emulation and compare four kinds of energy detection algorithms, comprise traditional E D, document [BeaulieuNC, ChenY.Improvedenergydetectorsforcognitiveradioswithrando mlyarrivingordepartingprimaryusers [J] .IEEESignalProcessLetters, 2010, 17 (10): 867 – 870.] Beaulieu propose improvement ED, document [ChenY.ImprovedenergydetectorforrandomsignalsinGaussianno ise [J] .IEEETransactionsonWirelessCommunications, 2010, 9 (2): 558 – 563.] Chen propose improvement ED and feedback stacks energy detection algorithm in this paper (FA-ED).In analogous diagram, " sim " and " analy " represents simulation value and theory analysis value respectively.

Select the testing environment that two kinds different: SNR=-5dB and-12dB, T detection time of corresponding energy detection algorithm _sensbe respectively: 0.05 millisecond (ms) and 1ms.Fig. 6 and Fig. 7 gives the relation of the time τ that detection probability exists within the detection time of SU along with detected subscriber signal.As can be seen from the figure, the theoretical value (calculated value of formula 6) of the FA-ED that the present invention proposes and actual emulation value are coincide, and demonstrate the correctness of FA-ED detection algorithm theory analysis.In the scene that SNR is higher, as SNR=-5dB in Fig. 6, the detection perform of the ED that Beaulieu proposes is slightly better than the detection perform of traditional E D, but in the scene that SNR is lower, as SNR=-12dB in Fig. 7, improvement in performance is undesirable, this and document [BeaulieuNC, ChenY.Improvedenergydetectorsforcognitiveradioswithrando mlyarrivingordepartingprimaryusers [J] .IEEESignalProcessLetters, 2010,17 (10): 867 – 870.] in simulation result coincide; And document [ChenY.ImprovedenergydetectorforrandomsignalsinGaussianno ise [J] .IEEETransactionsonWirelessCommunications, 2010,9 (2): 558 – 563.] detection perform of detection scheme under new model that propose be slightly poorer than traditional E D.In these two kinds of SNR scenes, the detection perform of the FA-ED that the present invention proposes obviously will be better than other three kinds of detection schemes.

Set the simulating scenes that different SU traffic intensity λ is respectively 10,50,100, SNR=-20dB, the data collision probability that Fig. 8 gives traditional E D and FA-ED compares.Can very clearly find out from Fig. 8, when multiple SU detects same channel, along with the increase of SU traffic intensity, the collision probability of SU also improves thereupon, and the initial idea that this and the present invention propose this new model is consistent; Along with improve, data collision probability is corresponding raising also; And the probability that the user data that the FA-ED scheme adopted causes collides is less than the data collision probability adopting traditional E D to cause; And subscriber traffic intensity is larger, the performance gain of acquisition is larger.On the basis of Fig. 8, choose specific in order to meet setting T detection time _sensmake the P of SU _dcan reach 0.98, under the scene of SNR=-20dB, compare under the scene of different business amount intensity, the data throughout of SU obtains Fig. 9 with the change of SU data frame length.As can be seen from Figure 9, elongated along with SU data frame length, the data throughout of SU improved constantly before this, worked as T _sensduring>=400ms, the throughput of SU is tending towards constant, and reach capacity state.Along with the change of traffic intensity is large, the convergency value of SU data throughout diminishes, and reason is that the probability of SU data collision becomes large, causes data throughout to decline.But, under specific traffic carrying capacity scene, adopt FA-ED that the data throughout of SU can be made to be greater than the data throughout using traditional E D to obtain; And subscriber traffic intensity is larger, the performance gain of acquisition is more obvious.

The data frame length of setting SU is that 400ms, Figure 10 give under different SU traffic intensity scene, and the data throughout of SU is along with the change of target detection probability.As can be seen from the figure, when multiple SU detects with during with PU frequency range simultaneously, no matter adopt which kind of ED to carry out frequency spectrum detection, the data throughout of SU along with the change of traffic intensity large and decline, concrete reason be because along with improve, data collision probability is mentioned, thus the data throughout causing SU to obtain declines.The data throughout that adopt FA-ED detection algorithm to carry out data throughout that frequency spectrum detection SU can obtain obtains than traditional E D is large; In the scene that SU traffic intensity is larger, FA-ED can significantly improve the data throughout of SU.

These embodiments are interpreted as only being not used in for illustration of the present invention limiting the scope of the invention above.After the content of reading record of the present invention, technical staff can make various changes or modifications the present invention, and these equivalence changes and modification fall into the inventive method claim limited range equally.

Claims (2)

1. solve the feedback stacks energy detection method that user arrives at random, it is characterized in that comprising the following steps:

101, the parameter of Initialize installation feedback stacks energy detection algorithm, comprising: target false alarm probability t detection time of feedback stacks energy detection algorithm _sens, and the sample frequency f of feedback stacks energy detection algorithm _s, draw the sampled point number I=T that feedback stacks energy detection algorithm is total _sensf _s;

102, the target false alarm probability of setting in step 101 is utilized with total sampled point number I, according to formula calculate the decision threshold of feedback stacks energy detection algorithm, wherein Q ^-1() is the inverse function of standard normal cumulative distribution function;

103, at T detection time _sensin, the signal in channel is sampled, obtains I sampled point, be designated as y _i, wherein sampled point sequence number i be 1,2 ... I/2, I/2+1 ..., I, calculate the instantaneous energy of each sampled point by rear half period also namely sampled point sequence number be that the sampled point instantaneous energy of I/2+1 ~ I is added to front half period, also namely sampled point sequence number is on the sampled point instantaneous energy of 1 ~ I/2, and the statistical decision amount Y expression formula of feedback stacks energy detection algorithm is: wherein it is the instantaneous energy of i-th sampled point;

104, the decision threshold η ' that the energy statistics judgement amount Y obtained in step 103 and step 102 obtain is compared, if Y >=η ', then be judged as in channel, there is primary user or secondary user, if Y < is η ', then be judged as in channel, there is not primary user and time user, complete the detection to channel.

2. the feedback stacks energy detection method solving user and arrive at random according to claim 1, is characterized in that: when the energy statistics judgement amount Y in step 103 is at sense cycle T _senswhen inside there is not primary user or secondary subscriber signal, then wherein n _ifor noise samples, the theoretical false alarm probability that can obtain feedback stacks energy detection algorithm is consider the emergent situation of user in sense cycle when there is subscriber signal, energy statistics judgement amount Y is simultaneously: $Y=\left\{\begin{array}{cc}\underset{i=1}{\overset{I/2}{\&Sigma;}}{n}_i^2+2\underset{i=I/2+1}{\overset{I-a}{\&Sigma;}}{n}_i^2+2\underset{i=I-a+1}{\overset{I}{\&Sigma;}}{\left(s_i+n_i\right)}^2& 0<a<I/2\\ \underset{i=1}{\overset{I-a}{\&Sigma;}}{n}_i^2+2\underset{i=I/2+1}{\overset{I}{\&Sigma;}}{\left(s_i+n_i\right)}^2+\underset{i=I-a+1}{\overset{I/2}{\&Sigma;}}{\left(s_i+n_i\right)}^2& I/2\&le;a\&le;I\end{array}\right.,$ Wherein s _ifor subscriber signal sampling in detected frequency range, n _ifor noise samples, a represents the sampled point number including subscriber signal at sense cycle latter half, 0 < a≤I, and the detection probability obtaining feedback stacks energy detection algorithm is $P_d=\left\{\begin{array}{cc}Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+2a\mathrm{\&gamma;})}{\sqrt{5I+16a\mathrm{\&gamma;}}}\right)& 0<a<I/2\\ Q\left(\frac{{\mathrm{\&eta;}}^{\&prime;}-(1.5I+a\mathrm{\&gamma;}+0.5I\mathrm{\&gamma;})}{\sqrt{5I+4a\mathrm{\&gamma;}+6I\mathrm{\&gamma;}}}\right)& I/2\&le;a\&le;I\end{array}\right.,$ γ is subscriber signal signal to noise ratio.