CN104469811A - Clustering cooperative spectrum sensing hard fusion method for cognitive wireless sensor network - Google Patents

Abstract

The invention discloses a clustering cooperative spectrum sensing hard fusion method for a cognitive wireless sensor network. All cognitive sensing nodes are clustered firstly, a cluster head of each cluster is selected, the cognitive sensing nodes achieve spectrum sensing according to the signal energy of a master user, local spectrum sensing information is acquired, then first-time fusion is carried out on all the local spectrum sensing information in the clusters through the cluster heads, the spectrum sensing information of the corresponding cluster is acquired, and finally fusion is carried out on the spectrum sensing information of all the clusters through a fusion center to obtain the final spectrum sensing information. Through cluster head screening and two times of fusion, the computation burden of the fusion center is reduced, and the effect of fast judging whether frequency spectrum is free or not and improving the correct probability of detection is achieved.

Description

The hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing

Technical field

The present invention relates to radio communication and input field, particularly relate to the hard fusion method of a kind of cognition wireless sensing network sub-clustering cooperative spectrum sensing.

Background technology

Cognition wireless sensing network is the novel sensing network structure of one that cognitive radio technology and wireless sensor network combine with technique produce, network node in this network has the function of frequency spectrum perception and Dynamic Selection frequency spectrum resource, this idle frequency range can not utilized to carry out radio communication by during use at the frequency band of primary user, substantially increase the availability of frequency spectrum, alleviate the problem of wireless sensor network frequency spectrum resource anxiety, therefore, how efficiently and accurately perception idle frequency spectrum improves the availability of frequency spectrum, the key of guarantee information transmission reliability.

Cooperative sensing can improve perceptual performance greatly, but intercourse perception information meeting occupying system resources between node and produce propagation delay time, when nodes is larger, the problem bringing communication overhead excessive while performance improves, the not good node of some perceptual performance adds can reduce overall detection perform on the contrary.

Existing cooperation frequency spectrum sensing method does not take into full account the complexity of node layout, the geographical location information of network node layout and simple topology information, and its fusion method complexity adopted is high, inapplicable large-scale cognition wireless sensing network.

Summary of the invention

In view of this, in order to solve the communication overhead problem of cooperative sensing, the invention provides the hard fusion method of a kind of cognition wireless sensing network sub-clustering cooperative spectrum sensing, to reach the effect of perception idle frequency spectrum efficiently and accurately.

In order to solve communication overhead problem, the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing of the present invention comprises step:

1) sub-clustering is carried out to all cognitive sensing node in network, form K cognitive sensing bunch;

2) described fusion center selects a cognitive sensing node as a bunch head from each described cognitive sensing bunch;

3) whether a certain frequency spectrum of perception is idle separately forms local frequency spectrum perception information for all sensing nodes in each described cognitive sensing bunch, and described local frequency spectrum perception information being sent to bunch head at place bunch, described local frequency spectrum perception information is divided into idle and busy two states;

4) described local frequency spectrum perception information all in this bunch received is carried out first time fusion and is formed this bunch of frequency spectrum perception information by each described bunch head, and described bunch of frequency spectrum perception information is sent to described fusion center, described bunch of frequency spectrum perception information is divided into idle and busy two states;

5) this bunch of frequency spectrum perception information of all described bunch head received is carried out second time and is merged the final frequency spectrum perception information forming described a certain frequency spectrum by described fusion center, and described final frequency spectrum perception information is divided into idle and busy two states.

As the further improvement of the inventive method, step 1) in fusion center to carry out to all cognitive sensing node in network the method that sub-clustering adopts be Fuzzy C-Means Cluster Algorithm.

As the further improvement of the inventive method, step 2) in select the concrete steps of bunch head to be: described fusion center and arbitrary described cognitive sensing node carry out radio communication by frequency spectrum perception channel, described fusion center calculates the signal to noise ratio of all described frequency spectrum perception channels, selects cognitive sensing node that in each described cognitive sensing bunch, signal to noise ratio is maximum as bunch head of this bunch.

As the further improvement of the inventive method, described step 3) in form the concrete steps of local frequency spectrum perception information as follows:

All described cognitive sensing node in described cognitive sensing bunch carries out local energy to described a certain frequency spectrum and detects and obtain energy value;

Described energy value and the thresholding preset are compared, described energy value is greater than described threshold value, then described local frequency spectrum perception information is busy state, otherwise is idle condition.

As the further improvement of the inventive method, described step 4) in first time the concrete steps that merge as follows:

Described bunch of head adds up the number of the local frequency spectrum perception information of busy state in this bunch;

By described number and the threshold value L preset ₁compare, if described number be more than or equal to described in the threshold value L that presets ₁, then described bunch of frequency spectrum perception information is busy state, otherwise is idle condition.

As step 4) further improvement, described threshold value L ₁be set to 1 or be not less than the arbitrary integer of cognitive sensing node sum half in this bunch.

As the further improvement of the inventive method, step 5) in form the concrete steps of final frequency spectrum perception information as follows:

The number of this bunch of frequency spectrum perception information of all busy states of fusion center statistics;

By described number and the threshold value L preset ₂compare, if described number be more than or equal to described in the threshold value L that presets ₂, then final frequency spectrum perception information is busy state, otherwise is idle condition.

As step 5) further improvement, described threshold value L ₂be set to 1 or be not less than the arbitrary integer of all described bunch of heads sum half.

Compared with prior art, advantage of the present invention is: sub-clustering collaboration frequency spectrum detection algorithm is applied to cognition wireless sensing network, the cognitive sensing node that Fuzzy C-Means Cluster Algorithm can be roughly the same to performance is rapidly utilized to classify, improve the detection perform of each cognitive sensing bunch, the maximum sensing node of each sub-clustering intermediate frequency spectrum channel perception signal to noise ratio is selected to improve the performance of sensing node as bunch head, the traffic of sensing node and fusion center is reduced by twice fusion, reduce the operand of fusion center simultaneously, fusion center can be judged efficiently, and whether frequency spectrum is idle, by the setting of comparison threshold, reduce the probability of erroneous judgement, effectively improve the availability of frequency spectrum.

Accompanying drawing explanation

Fig. 1 is cognition wireless sensing network application scenarios schematic diagram;

The cognition wireless Sensor Network sub-clustering cooperative spectrum sensing topology diagram that Fig. 2 provides for the embodiment of the present invention;

The hard fusion method flow chart of cognition wireless Sensor Network sub-clustering cooperative spectrum sensing that Fig. 3 provides for the embodiment of the present invention;

The cognition wireless Sensor Network sub-clustering cooperative spectrum sensing hard calculating fusion machine simulated effect figure that Fig. 4 provides for the embodiment of the present invention.

Embodiment

Below in conjunction with embodiment, the present invention is further described.

As shown in Figure 1, primary user's network and be partly overlapping geographic area from user network, shares identical frequency spectrum resource.Primary user's topology of networks is all central controlled, and primary user's terminal 1 is controlled by the base station, different districts 2 in network.Be the cognition wireless sensing network that the present invention relates to from user network, wherein from user terminal be cognitive sensing node 3.Be clustering from the topological structure of user network, whether a certain frequency spectrum in cognitive sensing node 3 sensor coverage region idle, and bunches 4 is choose in the cognitive sensing node 3 from each cognitive sensing bunch.

As shown in Figure 2, cognitive sensing node CR, as the fundamental node of cognition wireless Sensor Network, has certain wireless communication ability and operational capability, is divided into multiple cognitive sensing bunches 2, exists and detect channel C between cognitive sensing node CR and primary user ₁, the frequency spectrum service condition of Real-Time Monitoring primary user 1, forms local frequency spectrum perception information; Between each cognitive sensing node CR, there is transmission channel C ₂, for transmitting local frequency spectrum perception information each other; Between each cognitive sensing node CR and fusion center 3, there is frequency spectrum perception channel C ₃, for transmitting this bunch of frequency spectrum perception information.

As shown in Figure 3, the hard fusion method concrete steps of cognition wireless sensing network sub-clustering cooperative spectrum sensing are:

1) clustering algorithm is adopted to realize sub-clustering, concrete employing Fuzzy C-Means Cluster Algorithm implementation procedure;

2) in bunch member, frequency spectrum perception channel SNRs ρ is selected _i,jmaximum cognitive sensing node as a bunch head, wherein ρ _i,jin representing the i-th bunch, a jth user is sent to the frequency spectrum perception channel SNRs of fusion center, ρ _{max, i}represent that bunch hair delivers to the frequency spectrum perception channel SNRs of fusion center;

3) each cognitive sensing node in the i-th bunch carries out local energy and detects and obtain energy value u _i,j, each cognitive sensing node is according to thresholding λ _i,jmake decisions as A _i,j=Ω (u _i,j).

4) result of decision is sent to a bunch hair by cognitive sensing node, each bunch of head setting threshold parameter L ₁, according to the first time hard fusion rule Φ of setting ₁() makes first time data fusion, i=1,2 ..., K, j=1,2 ... M _i, the wherein K number that is bunch, M _iit is the cognitive sensing node number in the i-th bunch;

5) each bunch of head is by its decision-making B _isent by frequency spectrum perception channel, fusion center setting threshold parameter L ₂, adopt the hard fusion rule Φ of second time ₂() makes last judgement Δ, Δ=Φ ₂(B ₁, B ₂..., B _k).

Wherein step 1) in adopt Fuzzy C-Means Cluster Algorithm to realize sub-clustering concrete steps be:

S1, setting sample set wherein s is the dimension of sample space, for cognition wireless sensing network s=2, and sample set element x _ibe 2 × 1 vectors, represent the position of cognitive sensing node in two-dimensional space; M is the sum of cognitive sensing node;

The number of S2, setting sub-clustering is K (K > 1), selects the position coordinates of K cognitive sensing node as cluster centre vector V=[v ₁, v ₂..., v _k], sample set element v _ibe 2 × 1 vectors, the sensing node of vector element correspondence position is as initial cluster head.

S3, distribution situation according to cognitive sensing node, setting realizes the distance of cluster, adopts Euclidean distance here.D _ij=|| x _j-v _i|| represent from sample point x _jto center v _idistance, namely cognitive sensing node is to the distance of bunch head;

The fuzzy membership matrix U=[u of S4, setting Fuzzy C-Means Cluster Algorithm _ij] be the fuzzy partition matrix of a K × n, u _ija jth sample x _jwhat belong to the i-th class is subordinate to angle value;

S5, the majorized function designing fuzzy Fuzzy C-Means Cluster Algorithm are as follows:

$\mathrm{Min}{J}_{\mathrm{fcm}}(U,V)=\underset{i=1}{\overset{K}{\mathrm{\Σ}}}\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{u}_{\mathrm{ij}}^m{d}_{\mathrm{ij}}^2---\left(1\right)$

Constraints is: 1≤j≤M, 1≤i≤K, u _ij>=0,1≤i≤K, 1≤j≤M.

S6, parameter initialization: given cluster number K (1 < K < n) and Fuzzy Exponential m (1≤m <+∞); Initialization cluster centre vector V is set ⁽⁰⁾; The precision ε > 0 of convergence is set; Make iteration count t=0, setting maximum iteration time t _max;

S7, calculating Distance matrix D ^(t)=[d _ij], wherein d _ij=|| x _j-v _i||;

S8, renewal fuzzy membership matrix U ^(t)as follows

$u_{\mathrm{ij}}^{\left(t\right)}={\left\{\underset{r=1}{\overset{K}{\mathrm{\&Sigma;}}}\right[{(d_{\mathrm{ij}}^{\left(t\right)}/d_{\mathrm{rj}}^{\left(t\right)})}^{\frac{2}{m-1}}\left]\right\}}^{-1}---\left(2\right)$

S9, according to formula (1) calculation optimization function J ^(t) _fcm(U, V);

S10, renewal cluster centre vector matrix V ^(t):

$v_i^{\left(t\right)}=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ij}}^{\left(t\right)}\right)}^m\&CenterDot;x_i/\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ij}}^{\left(t\right)}\right)}^m,i=\mathrm{1,2},...,K---\left(3\right)$

If S11 || V ^(k)-V ^(k-1)|| < ε or reach maximum iteration time t=t _max, then algorithm stops and exporting best fuzzy membership matrix U _optwith Optimal cluster center vector V _opt, Optimal cluster center vector V _optthe cognitive sensing node of correspondence position is initial cluster head, then according to fuzzy membership matrix U _optother cognitive sensing nodes that to draw with this bunch of head be cluster centre, form sub-clustering; Otherwise making t=t+1, turning to step S7 until finding bunch head and till forming K sub-clustering.

Wherein step 3) in each sensing node concrete steps of obtaining local frequency spectrum perception information be:

Whether perception primary user is used the local frequency spectrum perception information of certain frequency range, is summarized as a binary system Hypothesis Testing Problem:

$\left\{\begin{array}{c}{H}_0:y_{i,j}\left(n\right)=v\left(n\right)\\ H_1:y_{i,j}\left(n\right)=h_{i}x\left(n\right)+v\left(n\right)\end{array}\right.---\left(4\right)$

Wherein, v (n) is additive white Gaussian noise, and x (n) represents primary user's signal, h _ibe the channel perception fading coefficients of i-th sub-clustering, H ₀represent that this frequency range is not for be used by primary user, namely local frequency spectrum perception information is spatiality, H ₁represent that this frequency range is for be used by primary user, namely local frequency spectrum perception information is busy state.

The statistic of energy measuring is wherein N represents number of samples.Be expressed as the detection threshold of energy detector with λ, as Y > λ, then local frequency spectrum perception information is spatiality.Detection probability and the false alarm probability of single energy detector are expressed as: P _d=Pr (Y > λ | H ₁), P _f=Pr (Y > λ | H ₀).Suppose that interchannel noise Gaussian distributed is v (n) ~ N (0, δ ²), so detection limit Y then obeys χ ²distribution.So the frequency spectrum detector performance under AWGN channel perception can be expressed as $P_d=Q_u(\sqrt{2\mathrm{\&gamma;}},\sqrt{\mathrm{\&lambda;}})$ With $P_f=\mathrm{\&Gamma;}(u,\frac{\mathrm{\&lambda;}}{2})/\mathrm{\&Gamma;}\left(u\right).$

Step 4) with step 5) in fusion method identical, concrete steps are: to all state informations received, statistic behavior is the information number of busy state, and compare with the thresholding preset, the value preset described in if described number is more than or equal to, then adjudicating perception information is busy state, otherwise is idle condition.Difference is only step 4) in all state informations of receiving be all local frequency spectrum perception information that bunch head receives in this bunch, judgement perception information is this bunch of frequency spectrum perception information, and step 5) in all state informations of receiving be this bunch of frequency spectrum perception information of all bunches that fusion center receives, judgement perception information is final frequency spectrum perception information.

The present embodiment is for step 4) in fusion method be described, concrete steps are as follows: for the i-th bunch, add up the local frequency spectrum perception information A that in this bunch, all cognitive sensing nodes send _i,j, add up the number of the local frequency spectrum perception information of busy state,

By the numerical value of gained and the threshold value L preset _r(r=1,2) compare, if exceed this threshold value L _r, then this bunch of frequency spectrum perception information is busy state, otherwise is idle condition, and expression formula is as follows:

${\mathrm{\&Phi;}}_r(\&CenterDot;)\&DoubleRightArrow;\left\{\begin{array}{c}{H}_1:\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{A}_{i,j}\&GreaterEqual;L_r\\ H_0:\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{A}_{i,j}<L_r\end{array}\right.,r=\mathrm{1,2}---\left(5\right)$

Wherein, Φ _r() represents the r time hard fusion rule, r=1, and 2 represent that sensing node merges to bunch head first time hard fusion and bunch head to fusion center second time is hard respectively.L ₁represent first time hard fusion rule Φ ₁the threshold parameter that () is selected, L ₂represent first time hard fusion rule Φ ₂the threshold parameter that () is selected.

First time hard fusion rule Φ ₁() is described as: M _ifor the sum of sensing node in this bunch, as long as there is L ₁individual user detects and thinks that this frequency range is just used primary user, then judge that this bunch of frequency spectrum perception information is busy state; L ₁value can be: 1 or be greater than this bunch of sensing node sum M _ithe arbitrary integer of half.Work as L ₁when=1, as long as represent that having a sensing node to adjudicate a certain frequency spectrum in this bunch is busy state, then this bunch of frequency spectrum perception information is busy state; When time, as long as represent that having sensing node over half to adjudicate a certain frequency spectrum in this bunch is busy state, then this bunch of frequency spectrum perception information is busy state; Work as L ₁=M _itime, represent that only having sensing nodes whole in this bunch to adjudicate a certain frequency spectrum is busy state, this bunch of frequency spectrum perception information is just busy state.

When time, the detection probability P of this bunch of frequency spectrum perception information of i-th bunch _d,iwith false alarm probability P _f,ibe respectively:

$P_{d,i}=\underset{m=L}{\overset{M_i}{\mathrm{\&Sigma;}}}{C}_M^m\underset{j=1}{\overset{m}{\mathrm{\&Pi;}}}{P}_{d,j}\underset{k=1}{\overset{M-m}{\mathrm{\&Pi;}}}(1-P_{d,k})$ (6)

$P_{f,i}=\underset{m=L}{\overset{M_i}{\mathrm{\&Sigma;}}}{C}_M^m\underset{j=1}{\overset{m}{\mathrm{\&Pi;}}}{P}_{f,j}\underset{k=1}{\overset{M-m}{\mathrm{\&Pi;}}}(1-P_{f,k})$

Wherein, M _ibe expressed as the sum of cognitive sensing node in the i-th bunch, P _d,iand P _f,irepresent detection probability and the false alarm probability of a jth sensing node respectively, P _d,k, P _f,krepresent detection probability and the false alarm probability of a kth sensing node respectively.

Work as L ₁=M _itime, the detection probability P of this bunch of frequency spectrum perception information of i-th bunch _d,iwith false alarm probability P _f,ibe respectively:

$P_{d,i}=\underset{m=1}{\overset{M_i}{\mathrm{\&Pi;}}}{P}_{d,m}$ (7)

$P_{f,i}=\underset{m=1}{\overset{M_i}{\mathrm{\&Pi;}}}{P}_{f,m}$

Wherein, M _ibe expressed as the sum of cognitive sensing node in the i-th bunch, P _d,i, P _f,irepresent detection probability and the false alarm probability of i-th sensing node respectively.

Work as L ₁when=1, the detection probability P of this bunch of frequency spectrum perception information of i-th bunch _d,iwith false alarm probability P _f,ibe respectively:

$P_{d,i}=1-\underset{m=1}{\overset{M_i}{\mathrm{\&Pi;}}}(1-P_{d,m})$ (8)

$P_{f,i}=1-\underset{m=1}{\overset{M_i}{\mathrm{\&Pi;}}}(1-P_{f,m})$

Equally, the hard fusion rule Φ of second time ₂() is described as: K is cognitive Sensor Network sub-clustering number, as long as there is L ₂individual cognitive sensing bunch thinks that this frequency range is just used primary user, then judge that final frequency spectrum perception information is busy state; Work as L ₂when=1, as long as represent that having cluster to adjudicate frequency spectrum is busy state, then final frequency spectrum perception information is busy state; When time, represent that a cognitive sensing more than half bunch judgement frequency spectrum is busy state, then final frequency spectrum perception information is busy state; Work as L ₂during=K, represent that whole cognitive sensing bunch a certain frequency spectrum of judgement is busy state, then final frequency spectrum perception information is busy state.

In order to the effect that the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing that the present embodiment provides produces is described better, the present embodiment method and non-cluster-dividing method have been carried out Computer Simulation, and concrete steps are as follows:

1) produce primary user's signal, adopt QPSK modulation, channel perception is set to AWGN;

2) initialization false alarm probability, at P _fwhen certain, the initialization of detection probability counting, starts Monte Carlo simulation;

3) suppose that cognitive sensing node number total in cognitive sensing network is wherein, M _ifor the cognitive sensing node number in each bunch, K is the sub-clustering number utilizing clustering algorithm gained.

4) the energy measuring data of each cognitive sensing node are by the noise effect of channel perception, and namely cognitive sensing node receives the data of primary user's signal plus noise; Primary user to be detected by Energy-aware for the i-th bunch of Sino-German cognitive sensing node j, namely cognitive sensing node obtains energy statistics value is A _i,j;

5) cognitive sensing node is by energy statistics amount Y _i,jcompare with theoretical threshold value λ, obtain the binary result A of local frequency spectrum perception information _i,jand send to bunch head at place bunch;

6) bunch head is by the binary result A of all local frequency spectrum perception information in this bunch of receiving _i,jcarry out first time fusion firmly, select suitable threshold parameter L ₁, obtain this bunch of frequency spectrum perception information B by this threshold judgement _iand send to fusion center;

7) the fusion center end Received signal strength C of K sub-clustering _isecond time will be carried out at fusion center to merge, to obtain the final result Δ of K sub-clustering; Merging in the second time of fusion center is still hard amalgamation mode, selects suitable threshold parameter L ₂, whether existed by this threshold judgement primary user signal;

8) Monte Carlo simulation is adopted to repeat to realize above-mentioned steps;

9) result at every turn emulated is stored, in the performance parameter of terminal statistics receiver, to be transferred to the dynamic spectrum access module of cognition wireless Sensor Network.

In simulating scheme, the first time of leader cluster node merges firmly firmly merge with the second time of fusion center emulation supposition primary user is known, and the number of cognitive sensing node is M=50, and cognitive sensing node is divided into K bunch by us, supposes K=6 here.The reception SNR of channel perception is assumed that it is equivalent, and it is γ that emulation is selected to receive SNR _i=γ=15dB.Frequency spectrum perception channel SNR is set as η ₁=η ₃=η ₅=10dB and η ₂=η ₄=η ₆=5dB.Simulated program adopts Fuzzy C-Means Cluster Algorithm, and we suppose that in every cluster, cognitive sensing node is difference subsequently: 5,9,11,10,7,8.Monte Carlo simulation number of times is set as 10 ⁵.Complementary curve (CROC) performance curve of receiver identity that simulation result mainly contains detector is described.

As shown in Figure 4, when false alarm probability is P _dwhen=0.2025, the hard receiver false dismissal probability merging sub-clustering is P _m=0.871, the receiver false dismissal probability of non-sub-clustering is P _m=0.963; When false alarm probability is P _dwhen=0.3025, the hard receiver false dismissal probability merging sub-clustering is P _m=0.401, the receiver false dismissal probability of non-sub-clustering is P _m=0.56.Simulation result shows that, when cognitive sensing node is identical, the frequency spectrum perception receiver performance based on sub-clustering is better than the receiver performance of non-sub-clustering, and when false alarm probability one timing, the false dismissal probability of sub-clustering is lower than the false dismissal probability for sub-clustering.

More than describe the preferred embodiment of the present invention in detail; but the present invention is not limited to the detail in above-mentioned execution mode, within the scope of technical conceive of the present invention; can carry out multiple equivalents to technical scheme of the present invention, these equivalents all belong to protection scope of the present invention.

Claims (8)

1. the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing, is characterized in that, comprise step:

1) sub-clustering is carried out to all cognitive sensing node in network, form K cognitive sensing bunch;

2) described fusion center selects a cognitive sensing node as a bunch head from each described cognitive sensing bunch;

3) whether a certain frequency spectrum of perception is idle separately forms local frequency spectrum perception information for all sensing nodes in each described cognitive sensing bunch, and described local frequency spectrum perception information being sent to bunch head at place bunch, described local frequency spectrum perception information is divided into idle and busy two states;

4) described local frequency spectrum perception information all in this bunch received is carried out first time fusion and is formed this bunch of frequency spectrum perception information by each described bunch head, and described bunch of frequency spectrum perception information is sent to described fusion center, described bunch of frequency spectrum perception information is divided into idle and busy two states;

5) this bunch of frequency spectrum perception information of all described bunch head received is carried out second time and is merged the final frequency spectrum perception information forming described a certain frequency spectrum by described fusion center, and described final frequency spectrum perception information is divided into idle and busy two states.

2. the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing as claimed in claim 1, is characterized in that, described step 1) in fusion center sub-clustering is carried out to all cognitive sensing node in network method be Fuzzy C-Means Cluster Algorithm.

3. the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing as claimed in claim 1, is characterized in that, described step 2) in select the concrete steps of bunch head to be:

Described fusion center and arbitrary described cognitive sensing node carry out radio communication by frequency spectrum perception channel, described fusion center calculates the received signal to noise ratio of all described frequency spectrum perception channels, selects cognitive sensing node that in each described cognitive sensing bunch, received signal to noise ratio is maximum as bunch head of this bunch.

4. the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing as claimed in claim 1, is characterized in that, described step 3) in form the concrete steps of local frequency spectrum perception information as follows:

All described cognitive sensing node in described cognitive sensing bunch carries out local energy to described a certain frequency spectrum and detects and obtain energy value;

Described energy value and the thresholding preset are compared, described energy value is greater than described threshold value, then described local frequency spectrum perception information is busy state, otherwise is idle condition.

5. the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing as claimed in claim 1, is characterized in that, described step 4) in first time the concrete steps that merge as follows:

Described bunch of head adds up the number of the local frequency spectrum perception information of busy state in this bunch;

By the number of described local frequency spectrum perception information and the threshold value L preset ₁compare, if the number of described local frequency spectrum perception information be more than or equal to described in the threshold value L that presets ₁, then described bunch of frequency spectrum perception information is busy state, otherwise is idle condition.

6. the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing as claimed in claim 5, is characterized in that, described threshold value L ₁be set to 1 or be not less than the arbitrary integer of cognitive sensing node sum half in this bunch.

7. the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing as claimed in claim 1, is characterized in that, described step 5) in the concrete steps that merge of second time as follows:

The number of this bunch of frequency spectrum perception information of all busy states of fusion center statistics;

By the number of described bunch of frequency spectrum perception information and the threshold value L preset ₂compare, if the number of described bunch of frequency spectrum perception information be more than or equal to described in the threshold value L that presets ₂, then final frequency spectrum perception information is busy state, otherwise is idle condition.

8. the hard fusion method of cognition wireless sensing network sub-clustering cooperative spectrum sensing as claimed in claim 7, is characterized in that, described threshold value L ₂be set to 1 or be not less than the arbitrary integer of all described bunch of heads sum half.