CN103929259A - Multi-bit judgment cooperation self-adaptation spectrum sensing method based on confidence degrees in cognition OFDM system - Google Patents

Abstract

The invention relates to the technical field of information and communication, in particular to a multi-bit judgment cooperation self-adaptation spectrum sensing method based on confidence degrees in a cognition OFDM system to solve the problem that all cognition user local detection results are not fully utilized. The confidence degree concept is introduced into a multi-bit judgment model for the probably existing situation that malicious user attack and the inconsistent signal to noise ratio exist and several users are in the deep fading scene in the actual application scene. According to the multi-bit judgment cooperation self-adaptation spectrum sensing method, the cognition users carry out local judgment firstly, a fusion center carries out weight fusion and comparison judgment, the confidence degree increment is calculated according to the weighted average difference between the cognition user local judgment results and all cognition node judgment results, and the confidence degrees of all cognition users are updated and are utilized to solve respective weights for next time detection. The multi-bit judgment cooperation self-adaptation spectrum sensing method is suitable for information and communication occasions.

Description

A kind of many bit decision cooperation adaptive spectrum cognitive methods based on degree of belief in cognitive ofdm system

Technical field

The present invention relates to Information & Communication Technology field, be specifically related to a kind of cognitive radio cooperative frequency spectrum sensing method.

Background technology

Cognitive radio (Cognitive Radio, CR) is a kind of technology that effectively can be used for improving frequency spectrum resource utilization rate.CR can solve the problems such as current radio communication intermediate frequency spectrum resource is nervous, the availability of frequency spectrum is low by dynamic spectrum access technology.Cognitive radio is proposed at first by Mitola.Basic thought is: thus cognitive user improves the availability of frequency spectrum by select a good opportunity mode insertion authority user's frequency range or unauthorized user frequency range.But prerequisite is to having authorized user or the unauthorized user of frequency spectrum, not produce harmful interference.

Conventional frequency spectrum perception method has following 4 kinds at present: energy measuring method, cyclostationary characteristic detection method, matched filtering method and cooperative detection method.Because various alone families detection method is difficult to overcome the problems such as decline, multipath, stealthy terminal, collaborative spectrum sensing method is suggested, and obtains showing great attention to of numerous researchers.

Collaborative spectrum sensing, according to whether there is independently fusion center in cognitive radio networks, can be divided into centralized frequency spectrum perception and distributed frequency spectrum perception.In centralized frequency spectrum perception, need an independently fusion center, each cognitive user is sent to fusion center by the analog quantity of local court verdict or signal by special-purpose control channel, and fusion center is made final judgement by data fusion.Current most research all concentrates in the cooperation perception of this type.In distributed frequency spectrum perception, do not set up independently fusion center, each cognitive user and other users exchange local court verdict or the signal imitation amount of sharing each other, finally according to the fuse information of each cognitive user, obtain final judging result.

OFDM (Orthogonal Frequency Division Multiplexing, OFDM) modulation technique has numerous superior performances, and for example its availability of frequency spectrum is high, can resist the interference that multipath effect is brought.And OFDM technology can support frequency-selecting scheme very flexibly, can combine with cognitive radio networks, realize the distribution of adaptive spectrum resource in cognitive radio.In ofdm system, channel distribution is become to a plurality of subchannels, this also provides good realization basis for cognitive radio frequency spectrum detects.Because OFDM technology has good flexibility aspect spectral shaping, OFDM becomes the modulation technique of cognitive radio system first-selection.OFDM Technology Need is used fast fourier transform (Fast Fourier Transformation, FFT) module, and the complexity that therefore can come reduction system to realize by the FFT module with receiver, has simplified again the design of hardware simultaneously.Therefore the frequency spectrum detection algorithm of research based on cognitive ofdm system has high practical significance.

Summary of the invention

The present invention is the following problem existing in order to solve existing frequency spectrum sensing method:

1), abundant not to the local testing result utilization of each cognitive user;

2), opposing malice cognitive user attacking ability is low;

3), do not take into full account because each cognitive user geographical position of living in is different with operational environment, the court verdict of not identical each the caused cognitive user of signal to noise ratio is not identical on the degree of impact of global decision, and the testing result of different cognitive users exists the fact of reliability difference;

4) the real-time adjustment aim function of variation that, can not adaptive environment.

Thereby provide a kind of many bit decision cooperation adaptive spectrum cognitive methods based on degree of belief in a kind of cognitive ofdm system.

A kind of many bit decision cooperation adaptive spectrum cognitive methods based on degree of belief in cognitive ofdm system, is characterized in that: it is realized by following steps:

Step 1, set the initial value of each cognitive user degree of belief, that is: r _i=0; Set the initial value of each cognitive user degree of belief, that is: ω _i=1;

Step 2, each cognitive user independent detection, and utilize respectively formula:

$\left\{\begin{array}{cc}{H}_1& \underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{u}_i\&GreaterEqual;\mathrm{\&lambda;}\\ H_0& \underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{u}_i<\mathrm{\&lambda;}\end{array}\right.$

Whether judgement primary user exist, and obtains each user's local court verdict u _i; In formula: U representative system fusion center final judging result; H ₁represent that primary user exists; H ₀represent that primary user does not exist; N represents cognitive user number; λ is the judging threshold of fusion center;

Step 3, according to formula:

${\mathrm{\&delta;}}_i=|\stackrel{\&OverBar;}{u}-u_i|$

Calculate the local court verdict u of each cognitive user _iweighted average with each cognitive user court verdict poor, that is: the departure function δ of each cognitive user _i;

In formula: the weighted average of each cognitive user court verdict value be:

$\stackrel{\&OverBar;}{u}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_i\&times;{\mathrm{\&omega;}}_i$

Step 4, each cognitive user departure function δ obtaining according to step 3 _i, utilize formula:

${\mathrm{\&Delta;}}_i=\left\{\begin{array}{cc}1.5\tan (\frac{7}{2}\&times;\frac{-{\mathrm{\&delta;}}_i}{2^b-1}),& {\mathrm{\&delta;}}_i>\frac{1}{2}\\ 1,& {\mathrm{\&delta;}}_i\&le;\frac{1}{2}\end{array}\right.$

Obtain the corresponding degree of belief increment of each cognitive user △ _i;

Step 5, the corresponding degree of belief increment of each cognitive user △ obtaining according to step 4 _i, utilize formula:

r _i＝r _i+△ _i

Upgrade each cognitive user degree of belief r _i;

Step 6, utilize formula:

$r_i=\left\{\begin{array}{cc}{r}_{\max },& r_i>r_{\max }\\ r_i,& r_{\min }\&le;r_i\&le;r_{\max }\\ r_{\min },& r_i<r_{\min }\end{array}\right.$

Upgrade the degree of belief r of each cognitive user _i; In formula: r _maxfor the default degree of belief upper limit; r _minfor default degree of belief lower limit, r _min<0; | r _min| >>|r _max|;

Degree of belief r after step 7, each cognitive user obtaining according to step 6 are upgraded _i, utilize formula:

${\mathrm{\&omega;}}_i^{\&prime;}=\frac{h_i^2}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{h}_i^2}\&times;N$

Obtain the weights ω after each cognitive user is upgraded _i';

In formula:

$h_i=\left\{\begin{array}{cc}0,& r_i\&le;\stackrel{\&OverBar;}{r}+g\\ \frac{r_i-g}{r_{\max }-g},& r_i>\stackrel{\&OverBar;}{r}+g\end{array}\right.$

for the weighted average of current all cognitive user degree of beliefs, and:

$\stackrel{\&OverBar;}{r}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i\&times;{\mathrm{\&omega;}}_i;$

Parameter g is predetermined threshold value, g<0; Complete a kind of many bit decision cooperation adaptive spectrums perception based on degree of belief in cognitive ofdm system;

The degree of belief r of step 8, each cognitive user after upgrading according to step 6 _iweights ω after upgrading with each cognitive user _i', return to execution step two, carry out a kind of many bit decisions cooperation adaptive spectrums perception based on degree of belief in cognitive ofdm system next time.

In step 2, adjudicating the concrete grammar whether primary user exist is:

If the decision threshold of each cognitive user is respectively △, 2 △, 3 △ ..., M △, wherein: M=2 ^b-1, b represents the each transmission information bit number of each cognitive user;

Therefore the quantized interval of each cognitive user be [0, △), [△, 2 △) ..., [M △, ∞);

If cognitive user detects the energy value of signal between above-mentioned arbitrary interval, upload corresponding court verdict u _ito fusion center, i=0,1,2 ..., M;

Target function after fusion center judgement weighting with the magnitude relationship of λ, if be greater than, primary user exists, if be less than, primary user does not exist;

Wherein: N represents cognitive user number.

The present invention is directed to malicious user attack and the signal to noise ratio that in practical application scene, may exist inconsistent, the situation of individual user in deep fade scene, introduces many bit decisions model by the concept of degree of belief.In the present invention, each cognitive user is first carried out this locality judgement, fusion center is weighted and merges and relatively judgement, then according to the difference of the weighted average of the local court verdict of each cognitive user and whole cognitive nodes court verdicts, calculate degree of belief increment, and then upgrade each cognitive user degree of belief and utilize each cognitive user degree of belief to ask weights separately to prepare against detection next time.Simulation result shows, no matter be that network exists malicious user or each cognitive user signal to noise ratio is different, individual user is in deep fade scene, and the system that the effectively improvement system of many bit decisions collaborative spectrum sensing algorithm based on degree of belief that the present invention proposes is weighed by detection probability and error probability detects performance.

Accompanying drawing explanation

P in following figure _fexpression system false alarm probability, P _dexpression system detection probability, Probability of Error represents system mistake probability, P _mexpression system false dismissal probability.

Fig. 1 is in the situation of each cognitive user SNR=-6dB, totally 10 cognitive user in network, the many bits of not weighting (2bit, 3bit and 4bit) judgement method and traditional amalgamation judging mode (" with " judgement and "or" judgement) detection performance (detection probability P _dwith false alarm probability P _fbetween relation) analogous diagram;

Fig. 2 is in the situation of each cognitive user SNR=-6dB, totally 10 cognitive user in network, the many bits of not weighting (2bit, 3bit and 4bit) judgement method and traditional amalgamation judging mode (" with " judgement and "or" judgement) detection performance (error probability P _ewith false alarm probability P _fbetween relation) analogous diagram;

Fig. 3 is in the situation of each cognitive user SNR=-8dB, totally 10 cognitive user in network, the many bits of not weighting (2bit, 3bit and 4bit) judgement method and traditional amalgamation judging mode (" with " judgement and "or" judgement) detection performance (detection probability P _dwith false alarm probability P _fbetween relation) analogous diagram;

Fig. 4 is in the situation of each cognitive user SNR=-8dB, totally 10 cognitive user in network, the many bits of not weighting (2bit, 3bit and 4bit) judgement method and traditional amalgamation judging mode (" with " judgement and "or" judgement) detection performance (error probability P _ewith false alarm probability P _fbetween relation) analogous diagram;

Fig. 5 is in the situation of each cognitive user SNR=-10dB, totally 10 cognitive user in network, the many bits of not weighting (2bit, 3bit and 4bit) judgement method and traditional amalgamation judging mode (" with " judgement and "or" judgement) detection performance (detection probability P _dwith false alarm probability P _fbetween relation) analogous diagram;

Fig. 6 is in the situation of each cognitive user SNR=-10dB, totally 10 cognitive user in network, the many bits of not weighting (2bit, 3bit and 4bit) judgement method and traditional amalgamation judging mode (" with " judgement and "or" judgement) detection performance (error probability P _ewith false alarm probability P _fbetween relation) analogous diagram;

Fig. 7 is in the situation of each cognitive user SNR=-8, totally 10 cognitive user in network, and whether, there is detection performance (the detection probability P of malicious user under attacking in the many bits of not weighting (2bit) method _dwith false alarm probability P _fbetween relation) contrast simulation figure, wherein in there is malicious user Attack Scenarios, totally 2 malicious users in network;

Fig. 8 is in the situation of each cognitive user SNR=-8, totally 10 cognitive user in network, and whether, there is detection performance (the error probability P of malicious user under attacking in the many bits of not weighting (2bit) method _ewith false alarm probability P _fbetween relation) contrast simulation figure, wherein in there is malicious user Attack Scenarios, totally 2 malicious users in network;

Fig. 9 is in the situation of each cognitive user SNR=-6, totally 10 cognitive user in network, and whether, there is detection performance (the detection probability P of malicious user under attacking in the many bits of not weighting (2bit) method _dwith false alarm probability P _fbetween relation) contrast simulation figure, wherein in there is malicious user Attack Scenarios, totally 2 malicious users in network;

Figure 10 is in the situation of each cognitive user SNR=-6, totally 10 cognitive user in network, and whether, there is detection performance (the error probability P of malicious user under attacking in the many bits of not weighting (2bit) method _ewith false alarm probability P _fbetween relation) contrast simulation figure, wherein in there is malicious user Attack Scenarios, totally 2 malicious users in network;

Figure 11 is that each cognitive user signal to noise ratio is respectively SNR, SNR, SNR-1, SNR-1, SNR-2, SNR-3, SNR-3, SNR-12, SNR-14 and SNR-16 and each cognitive user signal to noise ratio is in two kinds of situations of SNR, detection performance (the detection probability P of the many bits of not weighting (2bit) method _dwith false alarm probability P _fbetween relation) analogous diagram, totally 10 cognitive user in network wherein, SNR=-8dB;

Figure 12 is that each cognitive user signal to noise ratio is respectively SNR, SNR, SNR-1, SNR-1, SNR-2, SNR-3, SNR-3, SNR-12, SNR-14 and SNR-16 and each cognitive user signal to noise ratio is in two kinds of situations of SNR, detection performance (the error probability P of the many bits of not weighting (2bit) method _ewith false alarm probability P _fbetween relation) analogous diagram, totally 10 cognitive user in network wherein, SNR=-8dB;

Figure 13 is that each cognitive user signal to noise ratio is respectively SNR, SNR, SNR-1, SNR-1, SNR-2, SNR-3, SNR-3, SNR-12, SNR-14 and SNR-16 and each cognitive user signal to noise ratio is in two kinds of situations of SNR, detection performance (the detection probability P of the many bits of not weighting (2bit) method _dwith false alarm probability P _fbetween relation) analogous diagram, totally 10 cognitive user in network wherein, SNR=-6dB;

Figure 14 is that each cognitive user signal to noise ratio is respectively SNR, SNR, SNR-1, SNR-1, SNR-2, SNR-3, SNR-3, SNR-12, SNR-14 and SNR-16 and each cognitive user signal to noise ratio is in two kinds of situations of SNR, detection performance (the error probability P of the many bits of not weighting (2bit) method _ewith false alarm probability P _fbetween relation) analogous diagram, totally 10 cognitive user in network wherein, SNR=-6dB;

Figure 15 is in the situation of be-6dB of each cognitive user signal to noise ratio, totally 10 cognitive user in network, wherein 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the detection probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _dwith false alarm probability P _fbetween relation) contrast simulation figure, wherein r _min=-4000, r _max=800, g=-400;

Figure 16 is in the situation of be-6dB of each cognitive user signal to noise ratio, totally 10 cognitive user in network, wherein 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the error probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _ewith false alarm probability P _fbetween relation) contrast simulation figure, wherein r _min=-4000, r _max=800, g=-400;

Figure 17 is in the situation of be-6dB of each cognitive user signal to noise ratio, totally 10 cognitive user in network, wherein 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the false dismissal probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _mwith false alarm probability P _fbetween relation) contrast simulation figure, wherein r _min=-4000, r _max=800, g=-400;

Figure 18 is in the situation of be-8dB of each cognitive user signal to noise ratio, totally 10 cognitive user in network, wherein 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the detection probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _dwith false alarm probability P _fbetween relation) contrast simulation figure, wherein r _min=-4000, r _max=800, g=-400;

Figure 19 is in the situation of be-8dB of each cognitive user signal to noise ratio, totally 10 cognitive user in network, wherein 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the error probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _ewith false alarm probability P _fbetween relation) contrast simulation figure, wherein r _min=-4000, r _max=800, g=-400;

Figure 20 is in the situation of be-8dB of each cognitive user signal to noise ratio, totally 10 cognitive user in network, wherein 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the false dismissal probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _mwith false alarm probability P _fbetween relation) contrast simulation figure, wherein r _min=-4000, r _max=800, g=-400;

Figure 21 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-15, SNR-15 and SNR-15 situation, the many bits of not weighting (2bit) method and detection performance (the detection probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _dwith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=-2dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 22 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-15, SNR-15 and SNR-15 situation, the many bits of not weighting (2bit) method and detection performance (the error probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _ewith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=-2dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 23 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-15, SNR-15 and SNR-15 situation, the many bits of not weighting (2bit) method and detection performance (the false dismissal probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _mwith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=-2dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 24 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-15, SNR-15 and SNR-15 situation, the many bits of not weighting (2bit) method and detection performance (the detection probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _dwith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=0dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 25 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-15, SNR-15 and SNR-15 situation, the many bits of not weighting (2bit) method and detection performance (the error probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _ewith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=0dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 26 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-15, SNR-15 and SNR-15 situation, the many bits of not weighting (2bit) method and detection performance (the false dismissal probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _mwith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=0dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 27 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-7, SNR-8 and SNR-9 situation, the many bits of not weighting (2bit) method and detection performance (the detection probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _dwith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=0dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 28 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-7, SNR-8 and SNR-9 situation, the many bits of not weighting (2bit) method and detection performance (the error probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _ewith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=0dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 29 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-7, SNR-8 and SNR-9 situation, the many bits of not weighting (2bit) method and detection performance (the false dismissal probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _mwith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=0dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 30 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-7, SNR-8 and SNR-9 situation, the many bits of not weighting (2bit) method and detection performance (the detection probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _dwith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=-2dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 31 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-7, SNR-8 and SNR-9 situation, the many bits of not weighting (2bit) method and detection performance (the error probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _ewith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=-2dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 32 is that each cognitive user signal to noise ratio is respectively in SNR, SNR-1, SNR-2, SNR-3, SNR-4, SNR-5, SNR-6, SNR-7, SNR-8 and SNR-9 situation, the many bits of not weighting (2bit) method and detection performance (the false dismissal probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _mwith false alarm probability P _fbetween relation) contrast simulation figure, totally 10 cognitive user in network wherein, SNR=-2dB, wherein r _min=-2000, r _max=400, g=-200;

Figure 33 is that each cognitive user signal to noise ratio is respectively in the situation of SNR, SNR, SNR+1, SNR+1, SNR-1, SNR-1, SNR+2, SNR+2, SNR-2 and SNR-2, totally 10 cognitive user in network, wherein first 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the detection probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _dwith false alarm probability P _fbetween relation) contrast simulation figure, SNR=-6dB wherein, r _min=-4000, r _max=800, g=-400;

Figure 34 is that each cognitive user signal to noise ratio is respectively in the situation of SNR, SNR, SNR+1, SNR+1, SNR-1, SNR-1, SNR+2, SNR+2, SNR-2 and SNR-2, totally 10 cognitive user in network, wherein first 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the error probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _ewith false alarm probability P _fbetween relation) contrast simulation figure, SNR=-6dB wherein, r _min=-4000, r _max=800, g=-400;

Figure 35 is that each cognitive user signal to noise ratio is respectively in the situation of SNR, SNR, SNR+1, SNR+1, SNR-1, SNR-1, SNR+2, SNR+2, SNR-2 and SNR-2, totally 10 cognitive user in network, wherein first 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the false dismissal probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _mwith false alarm probability P _fbetween relation) contrast simulation figure, SNR=-6dB wherein, r _min=-4000, r _max=800, g=-400;

Figure 36 is that each cognitive user signal to noise ratio is respectively in the situation of SNR, SNR, SNR+1, SNR+1, SNR-1, SNR-1, SNR+2, SNR+2, SNR-2 and SNR-2, totally 10 cognitive user in network, wherein first 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the detection probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _dwith false alarm probability P _fbetween relation) contrast simulation figure, SNR=-8dB wherein, r _min=-4000, r _max=800, g=-400;

Figure 37 is that each cognitive user signal to noise ratio is respectively in the situation of SNR, SNR, SNR+1, SNR+1, SNR-1, SNR-1, SNR+2, SNR+2, SNR-2 and SNR-2, totally 10 cognitive user in network, wherein first 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the error probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _ewith false alarm probability P _fbetween relation) contrast simulation figure, SNR=-8dB wherein, r _min=-4000, r _max=800, g=-400;

Figure 38 is that each cognitive user signal to noise ratio is respectively in the situation of SNR, SNR, SNR+1, SNR+1, SNR-1, SNR-1, SNR+2, SNR+2, SNR-2 and SNR-2, totally 10 cognitive user in network, wherein first 2 is malicious user, the many bits of not weighting (2bit) method and detection performance (the false dismissal probability P that the present invention is based on many bits (2bit) judgement collaborative spectrum sensing method of degree of belief _mwith false alarm probability P _fbetween relation) contrast simulation figure, wherein SNR=-8dB.

Embodiment

A kind of many bit decision cooperation adaptive spectrum cognitive methods based on degree of belief in embodiment one, cognitive ofdm system, in cognitive ofdm system, frequency spectrum perception can be used similar energy measuring method to judge each subcarrier.Supposing has l OFDM symbol, each subcarrier l plural detected value, i.e. 2l real number detected value can be detected in detection time.When primary user does not exist, what energy detector detected is the energy value of noise, supposes that now system is obeyed H ₀suppose; When primary user exists, what energy detector detected is the energy value of the superposed signal of noise and primary user's signal, and now supposing the system is obeyed H ₁suppose:

$R_{2l}=\left\{\begin{array}{cc}{N}_{2l}& H_0\\ N_{2l}+S_{2l}& H_1\end{array}\right.---\left(8\right)$

R _2l＝(r ₁,r ₂,...,r _2l) ^T (9)

Wherein, R _2lbe the 2l dimensional signal vector that energy detector detects, establishing each component is r _i, i ∈ (1,2 ..., 2l); N _2lthe 2l dimensional vector that represents noise signal, it obeys variance is σ ², average is zero Gaussian Profile; S _2lrepresent the 2l dimensional vector of primary user's signal, suppose its each component independence and have same distribution function, each component variance is average is m _s.The energy value Y detecting is the quadratic sum of 2l component detecting of energy detector.

According to energy measuring method, by the energy value Y detecting and threshold value γ comparison:

$Y=\underset{i=1}{\overset{2l}{\mathrm{\&Sigma;}}}{r}_i^2=\left\{\begin{array}{cc}<\mathrm{\&gamma;}& H_0\\ >\mathrm{\&gamma;}& H_1\end{array}\right.---\left(10\right)$

When Y surpasses thresholding, think and have primary user's signal, otherwise think and do not have primary user's signal.Because a 2l mentioned above measured value is and whether independent irrelevant gaussian variable, so its quadratic sum Y obeys distribute.When noise signal is white Gaussian noise, because its average is zero, the signal energy value Y detecting obeys central distribution; When primary user exists, the signal detecting is the stack of noise and primary user's signal, and average is non-vanishing.Therefore, obey non-central distribute:

$Y~\left\{\begin{array}{cc}{\&aleph;}_{2l}^2& H_0\\ {\&aleph;}_{2l}^2\left(\mathrm{\&lambda;}\right)& H_1\end{array}\right.---\left(11\right)$

Therefore the probability density function of the target function Y under two kinds of hypothesis can be expressed as:

$P\left(Y\right|H_0)=\left\{\begin{array}{cc}\frac{Y^{l-1}{e}^{-\frac{\mathrm{\&gamma;}}{{2\mathrm{\&sigma;}}^2}}}{{\left({2\mathrm{\&sigma;}}^2\right)}^l\mathrm{\&Gamma;}\left(l\right)},& R\&GreaterEqual;0\\ 0,& R<0\end{array}\right.---\left(12\right)$

$P\left(Y\right|H_1)=\left\{\begin{array}{c}\frac{{\left(\frac{\mathrm{\&gamma;}}{u}\right)}^{\frac{l-1}{2}}{e}^{-\frac{\mathrm{\&gamma;}+u}{2({\mathrm{\&sigma;}}^2+{\mathrm{\&sigma;}}_s^2)}}}{2({\mathrm{\&sigma;}}^2+{\mathrm{\&sigma;}}_s^2)}{I}_{l-1}\left(\frac{\sqrt{\mathrm{uy}}}{{\mathrm{\&sigma;}}^2+{\mathrm{\&sigma;}}_s^2}\right),R\&GreaterEqual;0\\ \begin{array}{cc}0,& R<0\end{array}\end{array}\right.---\left(13\right)$

Wherein, Γ (x) represents Gamma function, Ι _l-1(x) represent l-1 rank Bessel function. represent non-central parameter distributes.

If false alarm probability P _fexpression is the probability of unavailable channel by the misjudgement of available (not having primary user's signal) channel, and it is that target function Y is at H ₀suppose the lower probability that surpasses the threshold value of setting.

$\begin{array}{c}{P}_f=P(Y>\mathrm{\&gamma;}|H_0)={\&Integral;}_{\mathrm{\&gamma;}}^{\&infin;}\frac{y^{l-1}{e}^{-\frac{\mathrm{\&gamma;}}{2{\mathrm{\&sigma;}}^2}}}{{\left({2\mathrm{\&sigma;}}^2\right)}^l\mathrm{\&Gamma;}\left(l\right)}\mathrm{dy}\\ =1-\frac{\mathrm{\&Gamma;}(\frac{\mathrm{\&gamma;}}{{2\mathrm{\&sigma;}}^2},l)}{\mathrm{\&Gamma;}\left(l\right)}\end{array}---\left(14\right)$

If detection probability P _dfor unavailable (having primary user's signal) channel is correctly judged to the probability of unavailable channel, it is that target function Y is at H ₁suppose the lower probability that surpasses the threshold value of setting.

$P_d=P(Y>\mathrm{\&gamma;}|H_1)={\&Integral;}_{\mathrm{\&gamma;}}^{\&infin;}\left(\frac{{\left(\frac{y}{u}\right)}^{\frac{l-1}{2}}{e}^{-\frac{y+u}{2({\mathrm{\&sigma;}}^2+{\mathrm{\&sigma;}}_s^2)}}}{2({\mathrm{\&sigma;}}^2+{\mathrm{\&sigma;}}_s^2)}{I}_{l-1}\left(\frac{\sqrt{\mathrm{uy}}}{{\mathrm{\&sigma;}}^2+{\mathrm{\&sigma;}}_s^2}\right)\right)\mathrm{dy}---\left(15\right)$

$=Q_d(\sqrt{\frac{u}{{\mathrm{\&sigma;}}^2(1+\mathrm{SNR})}},\sqrt{\frac{\mathrm{\&gamma;}}{{\mathrm{\&sigma;}}^2(1+\mathrm{SNR})}})$

In formula: SNR represents the signal to noise ratio of cognitive ofdm system, and corresponding false dismissal probability P _m=1-P _dexpression is mistaken for unavailable channel the probability of available channel:

$P_m=1-Q_d(\sqrt{\frac{\mathrm{\&lambda;}}{{\mathrm{\&sigma;}}^2(1+\mathrm{SNR})}},\sqrt{\frac{\mathrm{\&gamma;}}{{\mathrm{\&sigma;}}^2(1+\mathrm{SNR})}})---\left(16\right)$

$Q_m(a,b)={\&Integral;}_b^{\&infin;}\frac{x_m}{a_m}{e}^{-\frac{x^2+a^2}{2}}{I}_{m-1}\left(a_x\right)\mathrm{dx}---\left(17\right)$

The global error detection probability of system is:

P _e＝P _fP ₀+P _mP ₁ (18)

Wherein: P ₀represent the non-existent probability of primary user, and P ₁represent the probability that primary user exists.

The signal energy value that in many bit decisions model, each cognitive user first detects oneself quantizes, the more bit informations that quantize to obtain are sent to fusion center.If the decision threshold of cognitive user is respectively △, 2 △, 3 △ ..., M △, wherein M=2 ^b-1, b represents the each transmission information bit number of each cognitive user.Therefore the quantized interval of each cognitive user be [0, △), [△, 2 △) ..., [M △, ∞).If cognitive user detects the energy value of signal between above-mentioned arbitrary interval, upload corresponding court verdict u _i(0,1,2 ..., M).Fusion center judgement target function with the magnitude relationship of λ, wherein N represents cognitive user number.Be greater than and judge that primary user exists, be less than judgement and do not exist.Wherein λ is the judging threshold of fusion center.

When number of users is enough large, can be target function D _sLMCthe approximate Gaussian distributed of regarding as.Utilize central-limit theorem to try to achieve u _iaverage and the variance of distribution function.When primary user does not exist, its average and variance are respectively:

$\begin{array}{cc}{\mathrm{\&mu;}}_0=\underset{k=0}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{kP}}_k^0& {\mathrm{\&sigma;}}_0^2=\underset{k=0}{\overset{M}{\mathrm{\&Sigma;}}}{k}^2{P}_k^0-{\mathrm{\&mu;}}_0^2\end{array}---\left(19\right)$

When primary user exists, its average and variance are respectively:

$\begin{array}{cc}{\mathrm{\&mu;}}_1=\underset{k=0}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{kP}}_k^1& {\mathrm{\&sigma;}}_1^2=\underset{k=0}{\overset{M}{\mathrm{\&Sigma;}}}{k}^2{P}_k^1-{\mathrm{\&mu;}}_1^2\end{array}---\left(20\right)$

Wherein: the energy value that expression primary user detects while not existing is fallen thresholding k △ and is arrived the probability in (k+1) △; And the energy value detecting while representing to have primary user is fallen thresholding k △ and is arrived the probability in (k+1) △.Wherein, F ₁the probability density function of detected energy value when (.) expression primary user exists, F ₁the probable value of Y<k △ when (k △) represents that primary user exists, F ₀the probability density function of detected energy value when (.) expression primary user does not exist, F ₀(k △) represents the now probable value of Y<k △.Utilize P in formula (14) (15) _fand P _dderivation result can obtain with expression formula:

$\begin{array}{c}{P}_k^0=F_0\left((k+1)\mathrm{\&Delta;}\right)-F_0\left(\mathrm{k\&Delta;}\right)\\ =P(Y>\mathrm{k\&Delta;}|H_0)-P(Y>(k+1)\mathrm{\&Delta;}|H_0)\\ =(1-\frac{\mathrm{\&Gamma;}(\frac{\mathrm{k\&Delta;}}{{2\mathrm{\&sigma;}}^2},l)}{\mathrm{\&Gamma;}\left(l\right)})-(1-\frac{\mathrm{\&Gamma;}(\frac{(k+1)\mathrm{\&Delta;}}{{2\mathrm{\&sigma;}}^2},l)}{\mathrm{\&Gamma;}\left(l\right)})\\ =\frac{\mathrm{\&Gamma;}(\frac{(k+1)\mathrm{\&Delta;}}{{2\mathrm{\&sigma;}}^2},l)}{\mathrm{\&Gamma;}\left(l\right)}-\frac{\mathrm{\&Gamma;}(\frac{\mathrm{k\&Delta;}}{{2\mathrm{\&sigma;}}^2},l)}{\mathrm{\&Gamma;}\left(l\right)}\end{array}---\left(21\right)$

$\begin{array}{c}{P}_k^1=F_1\left((k+1)\mathrm{\&Delta;}\right)-F_1\left(\mathrm{k\&Delta;}\right)\\ =P(Y>\mathrm{k\&Delta;}|H_1)-P(Y>(k+1)\mathrm{\&Delta;}|H_1)\\ =Q_d(\sqrt{\frac{u}{{\mathrm{\&sigma;}}^2(1+\mathrm{SNR})}},\sqrt{\frac{\mathrm{k\&Delta;}}{{\mathrm{\&sigma;}}^2(1+\mathrm{SNR})}})\\ -Q_d(\sqrt{\frac{u}{{\mathrm{\&sigma;}}^2(1+\mathrm{SNR})}},\sqrt{\frac{(k+1)\mathrm{\&Delta;}}{{\mathrm{\&sigma;}}^2(1+\mathrm{SNR})}})\end{array}---\left(22\right)$

Formula (21) and (22) are calculated with substitution formula (19) and (20), can obtain:

${\mathrm{\&mu;}}_0=M-\underset{k=1}{\overset{M}{\mathrm{\&Sigma;}}}{F}_0\left(\mathrm{k\&Delta;}\right)$

${\mathrm{\&sigma;}}_0^2=M^2-\underset{k=1}{\overset{M}{\mathrm{\&Sigma;}}}(2k-1)F_0\left(\mathrm{k\&Delta;}\right)$

${\mathrm{\&mu;}}_1=M-\underset{k=1}{\overset{M}{\mathrm{\&Sigma;}}}{F}_1\left(\mathrm{k\&Delta;}\right)$

${\mathrm{\&sigma;}}_1^2=M^2-\underset{k=1}{\overset{M}{\mathrm{\&Sigma;}}}(2k-1)F_1\left(\mathrm{k\&Delta;}\right)$

So target function average can ask, be N μ ₀(when primary user does not exist) or N μ ₁(when primary user exists), variance is (when primary user does not exist) or (when primary user exists).Thereby can try to achieve:

$\mathrm{Pr}(D_{\mathrm{SLMC}}\&GreaterEqual;\mathrm{\&lambda;}|H_0)=Q\left(\frac{\mathrm{\&lambda;}-N{\mathrm{\&mu;}}_0}{\sqrt{N}{\mathrm{\&sigma;}}_0}\right)---\left(23\right)$

$\mathrm{Pr}(D_{\mathrm{SLMC}}\&GreaterEqual;\mathrm{\&lambda;}|H_1)=Q\left(\frac{\mathrm{\&lambda;}-N{\mathrm{\&mu;}}_1}{\sqrt{N}{\mathrm{\&sigma;}}_1}\right)---\left(24\right)$

By formula (23) and formula (24) substitution:

P _e＝P(H ₀)Pr(D _SLMC≥λ|H ₀)+P(H ₁)Pr(D _SLMC<λ|H ₁)

Can obtain probability of false detection:

$P_e=P\left(H_0\right)Q\left(\frac{\mathrm{\&lambda;}-N{\mathrm{\&mu;}}_0}{\sqrt{N}{\mathrm{\&sigma;}}_0}\right)+P\left(H_1\right)(1-Q\left(\frac{\mathrm{\&lambda;}-N{\mathrm{\&mu;}}_1}{\sqrt{N}{\mathrm{\&sigma;}}_1}\right))---\left(25\right)$

In collaborative spectrum sensing system, malicious user deliberately sends wrong local court verdict, thereby affects the court verdict of fusion center.

Malicious user Attack Scenarios in tradition simple gate limit cognitive radio networks can be divided into 3 kinds below:

1) no matter whether primary user exists, and malicious user is exported " 0 " or output " 1 " forever forever;

2) no matter what actual court verdict is, assailant is output " 0 " or " 1 " at random;

3) assailant's output valve is forever contrary with its real court verdict.

In many bit decisions system, corresponding malicious user Attack Scenarios can be described as so:

1) no matter whether primary user exists, and malicious user is exported " 0 " or output " M " forever forever;

2) no matter what actual court verdict is, assailant exports the integer of " 0 " to " M " at random;

3) when true testing result is " 0 ", malicious user output " M ", otherwise, malicious user output " 0 ".

See intuitively, the 3rd kind of situation than front 2 kinds more severe.Therefore below the performance of weight analysis many bit decisions cooperation perception in the 3rd kind of situation.

Supposing to participate in assailant's proportion in the cognitive user of cooperation perception is α.The local court verdict probability different from its true court verdict that each cognitive user sends is so α.When existing ratio to be the assailant of α, the false alarm probability of the energy value that each cognitive user detects between each energy range and detection probability become

${P_0^0}^{\&prime;}=P_1^0\&times;\mathrm{\&alpha;}+P_2^0\&times;\mathrm{\&alpha;}+...+P_M^0\&times;\mathrm{\&alpha;}$

${P_k^0}^{\&prime;}=P_k^0\&times;(1-\mathrm{\&alpha;}),k=\mathrm{1,2},...,M-1---\left(26\right)$

${P_M^0}^{\&prime;}=P_M^0\&times;(1-\mathrm{\&alpha;})+P_0^0\&times;\mathrm{\&alpha;}$

${P_0^1}^{\&prime;}=P_1^1\&times;\mathrm{\&alpha;}+P_2^1\&times;\mathrm{\&alpha;}+...+P_M^1\&times;\mathrm{\&alpha;}$

${P_k^1}^{\&prime;}=P_k^1\&times;(1-\mathrm{\&alpha;}),k=\mathrm{1,2},...,M-1---\left(27\right)$

${P_M^1}^{\&prime;}=P_M^1\&times;(1-\mathrm{\&alpha;})+P_0^1\&times;\mathrm{\&alpha;}$

With in formula (26) and (27) with in difference alternate form (19) and formula (20) with that is: the performance of many bit decisions cooperation perception algorithms in cognitive ofdm system when can to obtain existing ratio be the assailant of α.

In order to weaken malice cognitive user to detecting the impact of performance, in improvement system, there is the detection performance of the cognition network of malicious user, a kind of many bit decision cooperation adaptive spectrum perception algorithms based on degree of belief are proposed.This algorithm comprises two parts, the calculating of cognitive user degree of belief value and the calculating of the blending weight based on degree of belief.Its main thought is: first each cognitive user is carried out this locality judgement, at fusion center, utilize Weighted Fusion algorithm to draw final judging result again, then according to local court verdict separately and whole differences of the weighted average of cognitive nodes court verdicts, calculate degree of belief increment, and then upgrade each cognitive user degree of belief and utilize each cognitive user degree of belief to ask separately weights in order to detect next time.

By mentioned earlier, by local frequency spectrum detection whole M thresholding △ for region, 2 △ ..., M △ be divided into M+1 interval.Because whole region is divided into a plurality of intervals, local testing result will be represented by b bit, M=2 ^b-1, and final detection judgement must be binary decision, between so local testing result and conclusive judgement, will no longer there is simply whether consistent relation, according to both, whether unanimously come the strategy that degree of belief is adjusted will be no longer applicable, and the new degree of belief method of adjustment that is applicable to many interval division need to be set.Method of adjustment can have a lot, but will meet following two basic principles:

1), if local testing result compares the weighted average close to each cognitive user court verdict, so corresponding degree of belief should increase; On the contrary, if local testing result deviates from the weighted average of each cognitive user court verdict, correspondingly should reduce;

2) local testing result more deviates from each cognitive user court verdict weighted average, and less amplitude should be larger.

Weighted average for whether the local testing result of quantitative description approaches (departing from) and which kind of degree to approach (departing from) each cognitive user court verdict with, has defined a departure function δ i who weighs local testing result and each node testing result weighted average gap of system.The weighted average of each cognitive user court verdict for each cognitive user court verdict weighted average in the present invention represent, the expression formula of δ i is so:

${\mathrm{\&delta;}}_i=|\stackrel{\&OverBar;}{u}-u_i|---\left(28\right)$

For cognitive user, suppose that the increment that degree of belief is upgraded after certain cooperative detection is △ _i, work as so departure function time, present node court verdict is to differ minimum with the weighted average of each cognitive nodes court verdict in various possible outcomes (court verdict one is decided to be integer), thinks that this court verdict of this node is comparatively reliable, △ _i=1.And work as time, think that this court verdict of this node is relative unreliable, △ _i<0, and departure function δ _ilarger, △ _iabsolute value larger.The function meeting this requirement has a lot, and the present invention utilizes function shown in formula (29).

${\mathrm{\&Delta;}}_i=\left\{\begin{array}{cc}1.5\tan (\frac{7}{2}\&times;\frac{-{\mathrm{\&delta;}}_i}{2^b-1}),& {\mathrm{\&delta;}}_i>\frac{1}{2}\\ 1,& {\mathrm{\&delta;}}_i\&le;\frac{1}{2}\end{array}\right.---\left(29\right)$

Method therefor of the present invention possesses following feature:

1) prerequisite that algorithm is used is that in system, malice cognitive user number is less.If malice cognitive user number is more, the algorithm based on degree of belief of the strength of depending on the collective has just become invalid algorithm from essence; 2), if local testing result compares the weighted average close to each cognitive user court verdict, so corresponding degree of belief should increase; On the contrary, if local testing result deviates from the weighted average of each cognitive user court verdict, correspondingly should reduce.Local testing result more deviates from each cognitive user court verdict weighted average, and less amplitude should be larger.Departure function δ _ilarger, degree of belief reduction is larger.Malicious act is carried out to severe punishment, otherwise malice cognitive user can, by the higher degree of belief of normal work a period of time accumulative total, then be sent the attack of long period as capital; 3) to degree of belief, upper limit r is set _max, after reaching maximum, degree of belief value no longer increases, and using and prevent that malice cognitive user accumulation degree of belief is as the capital of later attack.To degree of belief, lower limit r is set _min, r _min<0, in the agreement causing in order to prevent degree of belief from infinitely reducing, store the buffer overflow of degree of belief, after reaching minimum, degree of belief value no longer reduces, when degree of belief is worth when too small, by these user's weights are set, be 0, remove the testing result of this cognitive user, thereby thoroughly eliminate the impact of malice cognitive user on sensing results.For preventing that maliciously cognitive user is accumulated degree of belief value fast, establish | r _min| >>|r _max|; 4) weight coefficient is the increasing function of degree of belief.For preventing that indivedual cognitive user from occupying the generation of absolute leading role situation, weight coefficient can not infinitely increase.Algorithm of the present invention is made as cognitive user by the summation of weight coefficient and counts N.

Based on above-mentioned 4 features, introduce respectively second part: r of algorithm of the present invention below _ithe degree of belief value that represents i cognitive user, initial value is 0, ω _ibe the weighted value of i cognitive user, initial value is 1.In one-time detection, target function defines in original many bit decisions model become the Weighted Fusion result of algorithm of the present invention the magnitude relationship of judgement target function and fusion center threshold value λ, thus finally judge whether primary user exists.Before being weighted fusion, first will be according to the degree of belief value r of each cognitive user _icalculate weight coefficient ω _i.The weight coefficient function meeting the demands has a lot, and the present invention chooses with minor function and calculates the weights after upgrading.

$h_i=\left\{\begin{array}{cc}0,& r_i\&le;\stackrel{\&OverBar;}{r}+g\\ \frac{r_i-g}{r_{\max }-g},& r_i>\stackrel{\&OverBar;}{r}+g\end{array}\right.---\left(30\right)$

${\mathrm{\&omega;}}_i^{\&prime;}=\frac{h_i^2}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{h}_i^2}\&times;N---\left(31\right)$

In formula (30) and (31), each parameter is described below: for the weighted average of current all cognitive user degree of beliefs, rmax is the degree of belief upper limit arranging in step 6.N is the number of cognitive user in cognition network.Parameter g (<0) is predetermined threshold value, if the degree of belief of cognitive user is reduced to below, think that this user is for malicious user, when calculating final inspection statistic, its local testing result will not participate in conclusive judgement.It is in order to guarantee that those inferior users with slight negative degree of belief value still have certain weight that parameter g is made as negative value.Because in the starting stage, some reliable cognitive user may provide incorrect testing result due to factors such as external interference in short-term, thereby its degree of belief may be negative value.

After obtaining the weight coefficient of each cognitive user, the testing result that fusion center utilizes each cognitive user to send, introduces weight coefficient ω i and is weighted fusion, and decision rule can be expressed as

$\left\{\begin{array}{cc}1& \underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{u}_i\&GreaterEqual;\mathrm{\&lambda;}\\ 0& \underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{u}_i<\mathrm{\&lambda;}\end{array}\right.---\left(32\right)$

Wherein, λ still represents the fusion center decision threshold in above-mentioned multi-threshold model.

In practical application scene, because the residing geographical position of each cognitive user is different with operational environment, its signal to noise ratio is also different, therefore the local decisions result of each cognitive user is not identical on the degree of impact of global decision yet.Traditional cooperation perception fusion criterion " with " or "or" fusion criterion all adopt equal weight blending algorithm.But because each cognitive user environment of living in is different and constantly variation, each point degree affected by noise is also different, so equal weight blending algorithm is inappropriate.Inconsistent in signal to noise ratio, indivedual cognitive user are in deep fade scene, and the bad user of those channel conditions, because its testing result is unreliable, is just equal to malicious user mentioned above, and the less user's testing result of SNR is just more unreliable.Algorithm of the present invention is based on degree of belief and carry out weighted data fusion, variation that can adaptive environment, by weight coefficient is carried out to self adaptation adjustment, give the cognitive user that SNR is larger larger weights, effectively improve cooperation perceptual performance, not only can also be applied to each user's signal to noise ratio inconsistent for the attack of opposing malicious user, in the scene of indivedual cognitive user in deep fade state.

The present invention has following characteristics and marked improvement:

1, the present invention is without the prior information of any primary user's signal.Under cognitive ofdm system model, carry out emulation and analytic explanation herein, but in any system, for any type of primary user's signal, the frequency spectrum sensing method that the present invention proposes is all effective.Be that the present invention has the extremely wide scope of application.

2, frequency spectrum sensing method of the present invention adopts many bit decisions model, thereby the energy value that each cognitive user is detected utilizes more fully.Compare with the judgement model of traditional simple gate limit single-bit, the present invention who proposes under this many bit decisions model significantly promotes the detection probability of system, and error probability also obviously declines.Yet the detection performance of 2bit judgement model and 3bit judgement model is more or less the same, and has but saved the data of a large amount of transmission.Will choose thresholding number by demand in actual applications, all simulation curve all obtains under 2bit judgement model herein.

3, the present invention can effectively resist the attack of indivedual malice cognitive user in cognition network.

4, the present invention can be for the inconsistent scene of each cognitive user signal to noise ratio in cognition network, and effectively elevator system detects performance.

Predetermined threshold value g size when 5, the present invention calculates each cognitive user weights by arranging, can thoroughly reject the indivedual cognitive user under deep fade environment, thereby guarantee the reliability of entire system court verdict.

6, the effective prerequisite of the present invention is that in system, malice cognitive user number is less.If malice cognitive user number is more, the algorithm based on degree of belief of the strength of depending on the collective has just become invalid algorithm from essence.

7, the present invention is based on degree of belief and carry out weighted data fusion, variation that can adaptive environment.Be no matter that cognitive user signal to noise ratio changes or occurs malicious user and the local court verdict reliability decrease that causes, by weight coefficient being carried out to dynamic self-adapting, can effectively improve cooperation perceptual performance.

Claims (2)

1. a kind of many bit decisions cooperation adaptive spectrum cognitive methods based on degree of belief in cognitive ofdm system, is characterized in that: it is realized by following steps:

Step 1, set the initial value of each cognitive user degree of belief, that is: r _i=0; Set the initial value of each cognitive user degree of belief, that is: ω _i=1;

Step 2, each cognitive user independent detection, and utilize respectively formula:

$\left\{\begin{array}{cc}{H}_1& \underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{u}_i\&GreaterEqual;\mathrm{\&lambda;}\\ H_0& \underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{u}_i<\mathrm{\&lambda;}\end{array}\right.$

Whether judgement primary user exist, and obtains each user's local court verdict u _i; In formula: U representative system fusion center final judging result; H ₁represent that primary user exists; H ₀represent that primary user does not exist; N represents cognitive user number; λ is the judging threshold of fusion center;

Step 3, according to formula:

${\mathrm{\&delta;}}_i=|\stackrel{\&OverBar;}{u}-u_i|$

Calculate the local court verdict u of each cognitive user _iweighted average with each cognitive user court verdict poor, that is: the departure function δ of each cognitive user _i;

In formula: the weighted average of each cognitive user court verdict value be:

$\stackrel{\&OverBar;}{u}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_i\&times;{\mathrm{\&omega;}}_i$

Step 4, each cognitive user departure function δ obtaining according to step 3 _i, utilize formula:

${\mathrm{\&Delta;}}_i=\left\{\begin{array}{cc}1.5\tan (\frac{7}{2}\&times;\frac{-{\mathrm{\&delta;}}_i}{2^b-1}),& {\mathrm{\&delta;}}_i>\frac{1}{2}\\ 1,& {\mathrm{\&delta;}}_i\&le;\frac{1}{2}\end{array}\right.$

Obtain the corresponding degree of belief increment of each cognitive user △ _i;

Step 5, the corresponding degree of belief increment of each cognitive user △ obtaining according to step 4 _i, utilize formula:

r _i＝r _i+△ _i

Upgrade each cognitive user degree of belief r _i;

Step 6, utilize formula:

$r_i=\left\{\begin{array}{cc}{r}_{\max },& r_i>r_{\max }\\ r_i,& r_{\min }\&le;r_i\&le;r_{\max }\\ r_{\min },& r_i<r_{\min }\end{array}\right.$

Upgrade the degree of belief r of each cognitive user _i; In formula: r _maxfor the default degree of belief upper limit; r _minfor default degree of belief lower limit, r _min<0; r _min>>|r _max|;

Degree of belief r after step 7, each cognitive user obtaining according to step 6 are upgraded _i, utilize formula:

${\mathrm{\&omega;}}_i^{\&prime;}=\frac{h_i^2}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{h}_i^2}\&times;N$

Obtain the weights ω after each cognitive user is upgraded _i';

In formula:

$h_i=\left\{\begin{array}{cc}0,& r_i\&le;\stackrel{\&OverBar;}{r}+g\\ \frac{r_i-g}{r_{\max }-g},& r_i>\stackrel{\&OverBar;}{r}+g\end{array}\right.$

for the weighted average of current all cognitive user degree of beliefs, and:

$\stackrel{\&OverBar;}{r}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i\&times;{\mathrm{\&omega;}}_i;$

Parameter g is predetermined threshold value, g<0; Complete a kind of many bit decision cooperation adaptive spectrums perception based on degree of belief in cognitive ofdm system;

The degree of belief r of step 8, each cognitive user after upgrading according to step 6 _iweights ω after upgrading with each cognitive user _i', return to execution step two, carry out a kind of many bit decisions cooperation adaptive spectrums perception based on degree of belief in cognitive ofdm system next time.

2. a kind of many bit decisions cooperation adaptive spectrum cognitive methods based on degree of belief in cognitive ofdm system according to claim 1, is characterized in that adjudicating in step 2 the concrete grammar whether primary user exist and are:

If the decision threshold of each cognitive user is respectively △, 2 △, 3 △ ..., M △, wherein: M=2 ^b-1, b represents the each transmission information bit number of each cognitive user;

Therefore the quantized interval of each cognitive user be [0, △), [△, 2 △) ..., [M △, ∞);

If cognitive user detects the energy value of signal between above-mentioned arbitrary interval, upload corresponding court verdict u _ito fusion center, i=0,1,2 ..., M;

Target function after fusion center judgement weighting with the magnitude relationship of λ, if be greater than, primary user exists, if be less than, primary user does not exist.