Research Of Spectrum Sensing Algorithm Based On Random And Stochastic Network Model

The rapid development of information technology has led to the explosive growth of wireless communication services,and compared with traditional wired communication system which transmites information through cables,wireless communication system uses the electromagnetic wave as the carrier to realize the remote transmission of information.So the need of spectrum resource faces enormous challenge.However,spectrum resources are limited,especially high-quality spectrum resources are becoming increasingly scarce with distribution mechanism.In the other hand,with the progress of society,the desires of people are constanty increasing,and the demand for communication services is also increasing.This thesis tries to solve the problem.Based on the previous reseach of spectrum sensing algorithm in the field of cognitive radio(CR),this thesis proposes some spectrum sensing algorithms for a CR network to detect primary user(PU)network that exhibits some randomness in topology.These algorithms are effective and boost,so that it is usefull for us to improve utilitation of spectrum and solve the problem of spectrum shortage.Details are as follows,Firstly,in order to provide a real communication environment for spectrum sensing algorithms,this thesis makes a research on stochastic network model,and proposes a model based on Poisson point process and Voronoi graph theory for communication nodes in space.Compared with Boolean model and Gilbert model,this model can better describe randomness of wireless communication networks,and is more in line with the actual communication rules which user communicates with the closest base stations.The interference of this model is given by rigorous mathematical analysis.And the reliability and scientificity of the model are verified by the simulation of communication connectivity and communication area in the random distributed network.Secondly,in order to improving the spectrum efficiency of CR,classical cyclic spectrum sensing algorithms are studied deeply,and a novel cyclic spectrum sensing algorithm based on partial QR decomposition(PQR-CSS)is proposed in this thesis.At first step of the algorithm,spectral correlation functions(SCFs)for sampled signals are calculated to get a SCFs matrix,and then partial QR decomposition for the SCFs matrix is performed to separate the signal and noise.Finally,the idle spectrum can be detected effectively with help of signal component and noise component.The algorithm can effectively solve the problem of noise wall,and has strong robustness.This thesis gives a rigorous mathematical deduction for the algorithm,and the performance of the algorithm is proved by Pd curve,ROC curve and robustness curve.The simulation results have shown the proposed algorithm perform much better than the traditional ones at the same experimental condition.Thirdly,in order to further improve the efficiency and practicability of spectrum sensing algorithm,this thesis makes full use of the PQR-CSS algorithm and the proposed network model to propose a novel cooperative spectrum sensing algorithm based on Fisher liner descriminent.Different from existing algorithms either assume that there is only one PU or do not consider the topology of the network at all,the algorithm takes full account of the randomness of communication environment,and the classical supervised learning algorithm Fisher linear discriminant analysis is used to optimize the results for the fusion central.This thesis gives a rigorous mathematical deduction for the algorithm.And simulation results have proved that the algorithm can effectively solve the problem of hidden authorized users,and even in complex network environment,it still has high spectrum detection performance.