CN104780006A - Frequency spectrum detector soft fusion method based on minimum error probability rule - Google Patents

Abstract

The invention discloses a frequency spectrum detector soft fusion method based on a minimum error probability rule. The method comprises the following steps: computing detection statistics by using an energy detection method at a secondary user terminal; and designing an optimal threshold value and a weight vector based on the minimum error probability rule of Bayesian theory at a fusion center terminal of a frequency spectrum detector. A cooperative frequency spectrum sensing method in the invention is set as a linear weighted cooperative frequency spectrum sensing soft fusion method. A plurality of secondary users are placed in a cognitive wireless network to respectively sense the energy detection statistics of a primary user. Each secondary user is used for transmitting the energy detection statistics to the fusion center terminal through a report channel. The fusion center terminal is used for computing the final detection statistics value according to the designed detector weight vector, and then judging whether the primary user is existent after comparing the final detection statistics value with the optimal threshold value. Through the adoption of the method disclosed by the invention, the global error probability of the frequency spectrum can be reduced, the utilization rate of the frequency spectrum resource is effectively improved, and the detector performance is further improved.

Description

The soft fusion method of frequency spectrum detector based on minimum total error probability criterion

Technical field

The invention belongs to radio communication and input field, particularly relate to the soft fusion method of frequency spectrum detector of cognition wireless network.

Background technology

Cognitive radio (Cognitive Radio, CR) is the new ideas of wireless communication field.The unemployed frequency spectrum resource occurred in spatial domain, time domain and frequency domain is called as " frequency spectrum cavity-pocket ".CR core concept makes Wireless Telecom Equipment have the ability of discovery " frequency spectrum cavity-pocket " these frequency spectrum cavity-pockets of Appropriate application exactly.Secondary user's (Secondary User, SU) can carry out continuous print monitoring to target frequency bands, finds out not by the mandate frequency range that primary user (Primary User, PU) takies, so unauthorized user just can access this frequency range.But when primary user needs to use this frequency range, secondary user's (cognitive user) must exit this frequency range at the appointed time.

Cognition wireless network is a kind of Novel Communication network, from discovery idle frequency spectrum to Appropriate application frequency spectrum, achieves the function not available for conventional communication networks.Cognitive radio technology comes from software and radio technique, focuses on the utilization again to distributing frequency spectrum resource.By dynamic spectrum detection, frequency spectrum access and spectrum management, secondary user's can realize the abundant utilization to frequency spectrum resource.Its groundwork flow process is: secondary user's continues to detect the frequency spectrum resource of having authorized, and ensure the communication not between interfere with primary users, namely ensure that primary user preferentially uses, and when its transmission performance is not impaired, adjust transceiver adaptively to the enterprising Serial Communication of current detection idle frequency spectrum.The optimum ideals of this frequency spectrum resource makes secondary user's can not only utilize the usable spectrum detected, and this secondary user's can also be switched to new frequency spectrum, reduces interference and obtains better signal to noise ratio, thus enhancing the capacity of user.Generally speaking, cognitive radio networks effectively can reduce the radio communication restriction caused by spectrum shortage, improves the availability of frequency spectrum.

Summary of the invention

Goal of the invention: the invention provides the soft fusion method of a kind of cognition wireless network neutral line frequency spectrum detection, to reach the effect of perception idle frequency spectrum efficiently and accurately.

Technical scheme: the soft fusion method of a kind of frequency spectrum detector based on minimum total error probability criterion, comprises the steps:

1) primary user is set to the domestic consumer of primary user's network, according to concrete network environment determination modulation system and transmitting power;

2) channel perception in cognition wireless network, comprises channel fading and additive noise information;

3) M secondary user's is distributed in the diverse location of cognition wireless network, carries out the energy measuring in limited detection time T respectively; M secondary user's is calculated energy measuring statistic respectively;

4) reporting channel of supposition cognition wireless network is additive white Gaussian noise channel, and M secondary user's sends M energy measuring statistic by reporting channel respectively to fusion center end;

5) fusion center termination receives M energy measuring statistic data { y _i; Adopt minimum total error probability criterion, the linear weighted function coefficient { w of design fusion center end _iand optimum gate limit value γ _c;

6) fusion center end adopts the mode of linear weighted function to calculate global detection statistic y _c;

7) fusion center end is by detection statistic y _cwith optimum gate limit value γ _ccontrast; If y _c> γ _c, then adjudicate primary user and exist; If y _c< γ _c, then adjudicate primary user and do not exist.

Described step 1) in determine that primary user's signal madulation mode is QPSK signal, transmitting power according in IEEE802.22 standard signal to noise ratio require arrange.

Described step 2) channel perception and described step 4) in reporting channel be first set to additive white Gaussian noise channel; Then network analysis is carried out to channel fading profiles.

Described step 3) be specially:

A) setting observation time is T, and spectrum width is W, and the discrete detection of local frequency spectrum detection is counted as N=2TW;

B) owing to adopting soft merging mode, secondary user's energy measuring only calculates energy measuring statistic, does not need to realize judgement at secondary user's end;

C) energy measuring statistic obeys card side's distribution that the degree of freedom is 2TW; When detecting points N and being enough large, according to central-limit theorem, energy measuring statistic obeys approximate Gaussian distribution.

Described step 5) in specific as follows:

A) linear weighting method w is set _opt, target makes detector performance optimization; Concrete weight w _optcan according to practical communication environmental selection.

B) according to minimum global error canon of probability, optimum gate limit value γ is set _c; This optimum gate limit value can detect Derivation of Mathematical Model by collaboration frequency spectrum and draw.

Described step 6) in global detection normalized set method be linear weighted function mode, realize soft merging at fusion center end, be wherein y _ifor energy measuring statistic data, w _ifor the linear weighted function coefficient of fusion center end.

Described step 7) in judgement be frequency spectrum detecting result in frame detection time.

Beneficial effect: compared with prior art, the present invention is based on minimum total error probability criterion, considers the prior probability whether primary user exists, and design detector global optimum thresholding, can reduce frequency spectrum detector global error probability.The detector of the present invention's design can reduce the probability of erroneous judgement, effectively improves the utilance of frequency spectrum resource.This detector design is based on the mode of the soft merging of linear weighted function, and detector amount of calculation mainly concentrates on fusion center end, can reduce secondary user's operand, is conducive to the Project Realization of detector.Fusion center end mainly realizes the determination and global decision etc. of optimum threshold scheme, best initial weights, can further improve detector performance.

Accompanying drawing explanation

Fig. 1 is collaboration frequency spectrum detection figure in cognition wireless network of the present invention;

Fig. 2 is Mathematical Modeling figure of the present invention;

Fig. 3 is the collaboration frequency spectrum detector design figure that the present invention is based on minimum error probability;

Fig. 4 is design of Simulation figure of the present invention;

Fig. 5 is the curve chart of detector performance parameter of the present invention under different prior probability;

Fig. 6 is the receiver characteristic curve diagram under the different prior probability condition of the present invention;

Fig. 7 is collaboration frequency spectrum detector of the present invention different soft combination method receiver performance comparison diagram;

Fig. 8 is minimum error probability detector of the present invention and N-P criterion detector performance comparison diagram.

Embodiment

The present invention mainly designs a kind of collaboration frequency spectrum based on minimum total error probability criterion and detects soft blending algorithm.Conceptual design thinking is: at secondary user's end, adopts energy detection method to calculate detection statistic; At frequency spectrum detector fusion center end, based on the minimum total error probability criterion of bayesian theory, devise optimum threshold value and weighted vector.In the present invention, cooperative frequency spectrum sensing method is set as the soft fusion method of a kind of collaborative spectrum sensing of linear weighted function.Multiple secondary user's is placed, respectively the energy measuring statistic of perception primary user in cognition wireless network.Energy measuring statistic is transferred to fusion center end by reporting channel by each secondary user's respectively.Whether fusion center calculates final detection statistics value according to the detector weighted vector designed, after comparing, adjudicate this primary user and exist with optimum gate limit value.Frequency spectrum detector performance in the present invention is weighed by the statistic of detection probability, false alarm probability and false dismissal probability.Below in conjunction with accompanying drawing, there are following specific descriptions to the present invention.

(1) cognition wireless network collaborative spectrum sensing

What the present invention mainly studied is prioritization scheme that in cognitive radio, collaboration frequency spectrum detects, its collaboration frequency spectrum detection model under practical circumstances, as shown in Figure 1.Collaboration frequency spectrum detection technique proposes on single user frequency spectrum detection basis, object effectively eliminates multipath fading, solve the problems such as the concealed terminal in single user frequency spectrum detection, thus improve the detection probability under complicated wireless communications environment, reduce false alarm probability.Collaboration frequency spectrum testing process generally includes local detection, and Detection Information transmission and Detection Information merge three phases.First, each secondary user's detects frequency spectrum independently; Subsequently, testing result information is sent to fusion center by all secondary user's; Finally, Detection Information merges by fusion center, realizes judgement simultaneously, determines in observed frequency range, whether have primary user to exist.

(2) Mathematical Modeling

Collaborative spectrum sensing system as shown in Figure 2.The mathematical description of collaborative spectrum sensing is as follows:

Suppose a cognition wireless network realizing collaborative spectrum sensing, modulation signal is the authorized user of primary user's network, M to be in the fusion center composition that the separate cognitive user of different signal to noise ratio environment and carry out linear weighted function fusion, channel circumstance is additive white Gaussian noise (the Additive White Gaussian Noise not carrying out filtering, AWGN) channel, what each cognitive user was sampled to authorized user counts as N.

Assumed condition (for the i-th cognitive user) is as follows:

1. authorized user sends signal is s (k); 2. there is situation and be in authorized user there is not situation is 3. channel perception gain h _i; 4. channel perception noise is additive white Gaussian noise v _i(k); 5. energy detector input end signal x _i(k); 6. energy detector output end signal u _i(k); 7. reporting channel noise is additive white Gaussian noise n _i; 8. linear weighted function device input end signal y _i; 9. weight w is distributed _i.

For each secondary user's, the second order hypothesis testing formulas on kth sampled point is:

Wherein s (k) represents the signal that authorized user sends, x _ik () is the signal that i-th cognitive user receives, h _ibe channel gain, be constant in each sense cycle.Noise v _i(k) to be average be zero additive white Gaussian noise, namely meet distribution.Noise { v herein _i(k) } be noise-aware, the vector of the noise-aware variance composition of M cognitive user is set to

With reference to the accompanying drawings 2, the Received signal strength energy value of each cognitive user in a sense cycle is u _i, namely then { u _i{ y is obtained by carrying out adding up with the noise of reporting channel _i, be y=u+n with matrix representation.Wherein reporting channel noise { n _ibe average to be zero variance be gaussian variable.Reporting channel noise is also assumed to awgn channel, and reporting channel noise variance vector is { y _iat fusion center with the final detection statistic y of certain linear weighted function generate rule one _c:

$y_C=\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{y}_i=w^{T}y---\left(2\right)$

Wherein $w\stackrel{\mathrm{\&Delta;}}{=}{[{\mathrm{\&omega;}}_1,{\mathrm{\&omega;}}_2,...,{\mathrm{\&omega;}}_M]}^T,{\mathrm{\&omega;}}_i\&GreaterEqual;0.$

Because u _ithe quadratic sum of N number of Gaussian random variable, thus when authorize frequency range can with ( situation) time, obey the distribution of center card side, N rank, otherwise it obeys the non-central card side distribution of N rank, and parameter is η _i.

$\frac{u_i}{{\mathrm{\&sigma;}}_i^2}~\left\{\begin{array}{cc}{\mathrm{\&chi;}}_N^2& H_0\\ {\mathrm{\&chi;}}_N^2\left({\mathrm{\&eta;}}_i\right)& H_1\end{array}\right.---\left(3\right)$

Wherein, η _iit is the signal to noise ratio of i-th cognitive user environment e _sthe sequence energy of primary user's signal in one-period,

According to central-limit theorem, if when namely N is enough large, detection statistic { u _iaverage and variance respectively as follows:

Detection statistic such as formula (2) represents.Detection statistic and threshold value compare by collaboration frequency spectrum detector, thus draw the judgement of authorizing frequency range whether idle, and namely whether primary user's signal exists.

Single cognitive user frequency spectrum perception is easily subject to the impact of channel fading or shadow effect, therefore needs to adopt multiple user collaboration jointly to complete the validity detecting and improve detector.{ u _iobtain through noisy communication channel

{ y _i, and Ey _i=Eu _i, { y _ivariance be

Because { y _igaussian random variable, the y that its linear combining generates _calso be Gaussian Profile, its average and variance are distinguished as follows:

Wherein, m channel gain square vector.

$\begin{array}{c}\mathrm{Var}\left(y_C\right)=E{(y_C-\stackrel{\&OverBar;}{y_C})}^2\\ =w^{T}E\left[(y-\stackrel{\&OverBar;}{y}){(y-\stackrel{\&OverBar;}{y})}^T\right]w\end{array}---\left(9\right)$

For different supposed premises, variance is respectively

Wherein with be respectively with covariance matrix under supposing,

Detect as single cognitive user, final step is compared detection statistic and thresholding, obtains conclusive judgement,

Suppose with represent that this channel is at idle or seizure condition respectively.α and β represents average holding time and the mean down time of channel respectively, and then channel idle is defined as respectively with the probability taken:

$P\left(H_0\right)=\frac{\mathrm{\&beta;}}{\mathrm{\&alpha;}+\mathrm{\&beta;}},P\left(H_1\right)=\frac{\mathrm{\&alpha;}}{\mathrm{\&alpha;}+\mathrm{\&beta;}}---\left(15\right)$

Global error probability is expressed as:

(3) based on the collaboration frequency spectrum detector design scheme of minimum error probability

Fig. 3 describes the collaboration frequency spectrum detector design based on minimum total error probability criterion.First, primary user's signal is the signal of primary user's network proper communication, and its communications takies certain frequency range.Secondary user's in cognition wireless network detects this primary user in real time and whether is taking this frequency range sometime, realizes frequency spectrum detection task.So M secondary user's is by the signal energy value of channel perception perception primary user signal after channel perception decline with noise processed; Define the detection statistic that this energy value is each secondary user's.The present invention adopts the mode of soft merging to realize frequency spectrum detection, and therefore secondary user's end does not realize judgement.Then, M energy measuring statistic is transferred to fusion center through reporting channel.After fusion center termination receives these data, the weights determination mode of the soft merging of select linear, then based on minimum total error probability criterion devise optimum thresholding.Fusion center calculates global detection statistic, this global statistics is compared with optimum thresholding, and whether judgement primary user exist, and namely whether takies this frequency range.

1. primary user's signal

The primary user's signal supposing cognition wireless network is stable modulation signal, and transmitting power is enough large; Modulation system is according to the setting of primary user's network communications standards, and be assumed to QPSK signal here, described QPSK (Quadrature Phase Shift Keying) signal is Quadrature Phase Shift Keying signal.

2. detector is energy measuring mode

Suppose that the bandwidth of ideal bandpass filter is W, so be N=2TW at the number of samples of perception cycle T, can being expressed as of the energy measuring of i-th secondary user's:

$u_i=\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{\left|x_i\left(k\right)\right|}^2,i=\mathrm{1,2},...,M---\left(17\right)$

From formula (3), this energy detector statistic is the distribution of card side.When sampled point N is enough large, this detection statistic is approximate Gaussian distribution.

3. channel design

This collaboration frequency spectrum detector comprises channel perception and reporting channel.Wherein, channel perception is the energy value that each secondary user's catches primary user's signal, is set to awgn channel; Reporting channel is, after each secondary user's obtains energy measuring amount, statistic is sent to fusion center by reporting channel, is set as awgn channel.

4. signal to noise ratio design

According to IEEE 802.22 standard, ensure that signal to noise ratio is for being greater than-11dB when detection time is 0.2ms; When detection time is 1ms, signal to noise ratio is for being greater than-15dB; When detection time is 5ms, signal to noise ratio is for being greater than-18dB;

5. threshold scheme method:

The present invention is based on minimum global error canon of probability, global error probability expression can be obtained according to formula (16), so

Order for obtain minimum error probability, optimal threshold γ _c.Derive according to formula (8), (10) and (11)

${\mathrm{\&gamma;}}_c=\frac{{(\mathrm{N\&sigma;}+E_{s}g)}^T{\mathrm{\&Sigma;}}_{H_0}w-N{\mathrm{\&sigma;}}^T{\mathrm{\&Sigma;}}_{H_1}w-2\sqrt{\mathrm{\&Psi;}}}{w^T{\mathrm{\&Sigma;}}_{H_0}w-w^T{\mathrm{\&Sigma;}}_{H_1}w}---\left(19\right)$

Wherein

$\begin{array}{c}\mathrm{\&Psi;}={[N{\mathrm{\&sigma;}}^T{\mathrm{\&Sigma;}}_{H_1}w-{(\mathrm{N\&sigma;}+E_{s}g)}^T{\mathrm{\&Sigma;}}_{H_0}w]}^2-(w^T{\mathrm{\&Sigma;}}_{H_0}w-w^T{\mathrm{\&Sigma;}}_{H_1}w)\\ \&CenterDot;\{w^T{\mathrm{\&Sigma;}}_{H_0}w\&CenterDot;{\left(N{\mathrm{\&sigma;}}^{T}w\right)}^2-w^T{\mathrm{\&Sigma;}}_{H_1}w\&CenterDot;{\left[{(\mathrm{N\&sigma;}+E_{s}g)}^{T}w\right]}^2+2w^T{\mathrm{\&Sigma;}}_{H_0}{\mathrm{\&Sigma;}}_{H_1}w\&CenterDot;\ln [\frac{\mathrm{\&beta;}\&CenterDot;\sqrt{w^T{\mathrm{\&Sigma;}}_{H_1}w}}{\mathrm{\&alpha;}\&CenterDot;\sqrt{w^T{\mathrm{\&Sigma;}}_{H_0}w}}\left]\right\}\end{array}---\left(20\right)$

6. linear weighted function collaboration frequency spectrum detects weight number combining method

First present design devises optimum thresholding according to minimum total error probability criterion.Then according to the average signal-to-noise ratio of different secondary user's, adopt the weights soft combination method that three kinds different, weight number combining method (the Equal Gain Combining such as to be respectively, EGC), maximum-ratio combing method (Maximal Ratio Combing, MRC) and revise offset coefficient method (Modified Deflection Coefficient, MDC).Three kinds of soft merging detector coefficient defining method are as follows:

$w_{\mathrm{EGC},i}=\frac{1}{\sqrt{M}},1\&le;i\&le;M---\left(21\right)$

$w_{\mathrm{MRC},i}=\frac{{\mathrm{\&gamma;}}_i}{{\mathrm{\&Sigma;}}_{j=1}^M{\mathrm{\&gamma;}}_j},1\&le;i\&le;M---\left(22\right)$

$w_{\mathrm{MDC},i}=\frac{\frac{{\mathrm{\&gamma;}}_i}{(1+2{\mathrm{\&gamma;}}_i)}}{{\mathrm{\&Sigma;}}_{j=1}^M\frac{{\mathrm{\&gamma;}}_j}{(1+2{\mathrm{\&gamma;}}_j)}},1\&le;i\&le;M---\left(23\right)$

7. detector judgement

Derive according to formula (19) and draw optimum decision thresholding γ based on minimum error probability _c, be defined as the method method collaboration frequency spectrum detection fusion center-side decision threshold.According to linear weighted function soft fusion best initial weights vector w and decision threshold γ _c; Obtain the detection statistic y of linear weighted function in conjunction with decision-making of each secondary user's statistic _c, by this statistic y _cwith decision threshold γ _cjudgement is compared, and whether decision-making primary user takies this section of frequency spectrum.

8. detector performance index

The detection perform of linear weighted function collaborative spectrum sensing framework can according to its false alarm probability P _fwith detection probability P _dembody.

$P_f=Q\left[\frac{\mathrm{\&gamma;}-N{\mathrm{\&sigma;}}^{T}w}{\sqrt{w^T{\mathrm{\&Sigma;}}_{H_0}w}}\right]---\left(24\right)$

$P_d=Q\left[\frac{\mathrm{\&gamma;}-{(\mathrm{N\&sigma;}+E_{s}g)}^{T}w}{\sqrt{w^T{\mathrm{\&Sigma;}}_{H_1}w}}\right]---\left(25\right)$

P _m=1-P _d(26) wherein false alarm probability is the probability of system generation false-alarm, and false-alarm is exactly the existence that judgement has authorized user when frequency spectrum is idle, and false alarm probability is the smaller the better when other conditions are identical.Detection probability the probability that authorized user exists detected, false dismissal probability and detection probability complementation.The larger explanation detector performance of detection probability is better.

(4) the Realization of Simulation step

Collaboration frequency spectrum detection simulation can realize by following thinking: first authorization user signal produces random number, then modulates random number, produces the random number of 0 or 1; By there being the channel perception of additive white Gaussian noise, M secondary user's receives the energy of its primary user respectively, obtains detection statistic u through energy measuring _i; This detection statistic is through again by obtaining y after reporting channel; By fusion center, after being combined with weighted value ω, obtain global statistics y _c; Last global statistics y _ccompare with threshold gamma, make judgement.

Design of Simulation has needed the setting to environment of cognitive radio network, roughly comprises as follows: received signal to noise ratio η _i, secondary user's number, Monte Carlo simulation number of times, channel fading, primary user's signal energy, interchannel noise etc.

This simulating scheme frequency spectrum perception algorithm realization as shown in Figure 4, describes the algorithm specific implementation simulation flow of collaboration frequency spectrum detector.

As shown in Figure 4, the key step based on collaborative spectrum sensing design is as follows:

Step 1: be set major parameter, as time T, channel width W, secondary user's number M, signal to noise ratio η _iand Monte Carlo simulation number of times etc.

Step 2: produce primary user's signal, adopts QPSK modulation system, arranges primary user's signal transmission power;

Step 3: channel perception is set to awgn channel;

Step 4: initialization false alarm probability, at false alarm probability P _fwhen certain, the initialization of detection probability counting, starts Monte Carlo simulation;

Step 5:M secondary user's receives the transmission power level of primary user's signal;

Step 6: calculate M energy measuring statistic

Step 7: reporting channel is set to awgn channel;

Step 8: M energy measuring statistic is sent to fusion center end by reporting channel;

Step 9: fusion center termination receives detection statistic data y=u+n;

Step 10: fusion center end determination weights method, is respectively ω ^* _{opt, MRC}and and normalization obtains ω _{opt, DC}, ω _{opt, MRC}and ω _{opt, MDC}.Calculate global statistics

Step 11: according to minimum total error probability criterion, determines optimum gate limit value γ by formula (19) _c;

Step 12: by global statistics y _cwith optimum thresholding γ _crelatively; Work as y _cbe greater than thresholding γ _ctime, rolling counters forward, represents that judgement primary user exists; Work as y _cbe less than thresholding γ _ctime, counter remains unchanged, and represents that judgement primary user does not exist and works as.Detection probability P is obtained finally by Monte Carlo simulation circle statistics _dvalue.

Step 13: terminate Monte Carlo simulation;

Step 14: reset false alarm probability P _fvalue, repeat the process of step 5 to step 13, statistics draws different false alarm probability P _funder detection probability P _dvalue.

Step 15: the receiver performance curve chart drawing collaboration frequency spectrum detector, abscissa is P _f, ordinate is P _d.

(5) simulation results show

The soft merging simulation process of cognition wireless network frequency spectrum perception and the main basis of model describe realization above.

Simulation parameter is set to: sampling number N=20, and modulation system is QPSK, cognitive user number M=8, and Monte Carlo simulation number of times is 10 ⁵, the secondary user's received signal to noise ratio of channel perception is [-7.9 ,-3,2 ,-2.5,3 ,-9.2,5 ,-5.9] dB.

First verify in different prior probability situation, the performance difference of analyzing and testing device system.As shown in Figure 5.Wherein Fig. 5 (a) is the threshold value-probability curve diagram under the reasonable condition of signal to noise ratio, and Fig. 5 (b) is the thresholding-probability curve diagram under the poor condition of signal to noise ratio condition.As shown in Figure 5, prior probability is worked as when becoming large gradually, the minimum value of global error probability is reducing gradually, and namely systematic function becomes excellent gradually.So false alarm probability reduces when decision threshold increases.Namely, when decision-making value is lower, be easy to occur incorrect existence main users being detected.Along with the increase of decision threshold, false alarm probability reduces gradually, and drops to a less value at optimum thresholding.The change that global error probability is arranged with thresholding and adjusting.Minimum global error probability can be obtained at optimum thresholding place.

Fig. 6 describes the receiver feature complementary curve in above-mentioned situation.As shown in Figure 6, under identical signal to noise ratio condition, detector performance difference is little.When signal to noise ratio improves, the detection perform that receiver performance curve embodies improves a lot.

Fig. 7 describes the performance comparison of collaboration frequency spectrum detector under different weights method of the present invention's design.As shown in Figure 7, in three kinds of methods (EGC, MRC and MDC), the detector performance effect of MDC method is best.When false alarm probability is lower, the detector performance of MRC method is better than EGC method; But when false alarm probability is greater than 0.2, EGC method is better than MRC method.

Fig. 8 describe the present invention design minimum total error probability criterion detector and N-P criterion detector performance contrast.This emulation hypothesis prior probability is P (H ₀)=0.1 and P (H ₁)=0.9.As shown in Figure 8, minimum total error probability criterion detector performance is slightly better than N-P criterion detector, but its amount of calculation is also a little more than N-P criterion detector.In addition, detector performance also depends on the received signal to noise ratio of each secondary user's end, when received signal to noise ratio is larger, the performance of this collaboration frequency spectrum detector be better than signal to noise ratio less time detection perform.I.e. identical false alarm probability P _funder condition, false dismissal probability P _mless.

Claims (7)

1., based on the soft fusion method of frequency spectrum detector of minimum total error probability criterion, it is characterized in that, comprise the steps:

1) primary user is set to the domestic consumer of primary user's network, according to concrete network environment determination modulation system and transmitting power;

2) channel perception in cognition wireless network, comprises channel fading and additive noise information;

3) M secondary user's is distributed in the diverse location of cognition wireless network, carries out the energy measuring in limited detection time T respectively; M secondary user's is calculated energy measuring statistic respectively;

4) reporting channel of supposition cognition wireless network is additive white Gaussian noise channel, and M secondary user's sends M energy measuring statistic by reporting channel respectively to fusion center end;

5) fusion center termination receives M energy measuring statistic data { y _i; Adopt minimum total error probability criterion, the linear weighted function coefficient { w of design fusion center end _iand optimum gate limit value γ _c;

6) fusion center end adopts the mode of linear weighted function to calculate global detection statistic y _c;

7) fusion center end is by detection statistic y _cwith optimum gate limit value γ _ccontrast; If y _c> γ _c, then adjudicate primary user and exist; If y _c< γ _c, then adjudicate primary user and do not exist.

2. the soft fusion method of the frequency spectrum detector based on minimum total error probability criterion according to claim 1, it is characterized in that, described step 1) in determine that primary user's signal madulation mode is QPSK signal, transmitting power according in IEEE 802.22 standard signal to noise ratio require arrange.

3. the soft fusion method of the frequency spectrum detector based on minimum total error probability criterion according to claim 1, is characterized in that, described step 2) channel perception and described step 4) in reporting channel be first set to additive white Gaussian noise channel; Then network analysis is carried out to channel fading profiles.

4. the soft fusion method of the frequency spectrum detector based on minimum total error probability criterion according to claim 1, is characterized in that, described step 3) be specially:

A) setting observation time is T, and spectrum width is W, and the discrete detection of local frequency spectrum detection is counted as N=2TW;

B) owing to adopting soft merging mode, secondary user's energy measuring only calculates energy measuring statistic, does not need to realize judgement at secondary user's end;

C) energy measuring statistic obeys card side's distribution that the degree of freedom is 2TW; When detecting points N and being enough large, according to central-limit theorem, energy measuring statistic obeys approximate Gaussian distribution.

5. the soft fusion method of the frequency spectrum detector based on minimum total error probability criterion according to claim 1, is characterized in that, described step 5) in specific as follows:

A) linear weighting method w is set _opt, target makes detector performance optimization; Concrete weight w _optcan according to practical communication environmental selection.

B) according to minimum global error canon of probability, optimum gate limit value γ is set _c; This optimum gate limit value can detect Derivation of Mathematical Model by collaboration frequency spectrum and draw.

6. the soft fusion method of the frequency spectrum detector based on minimum total error probability criterion according to claim 1, is characterized in that, described step 6) in global detection normalized set method be linear weighted function mode, realize soft merging at fusion center end, be wherein y _ifor energy measuring statistic data, w _ifor the linear weighted function coefficient of fusion center end.

7. the soft fusion method of the frequency spectrum detector based on minimum total error probability criterion according to claim 1, is characterized in that, described step 7) in judgement be frequency spectrum detecting result in frame detection time.