Research On Cooperative Spectrum Sensing Technology Based On Belief Propagation In Cognitive Radio

Along with the rapid development of wireless communication technology, thecontradiction between the growing demands of business bandwidth and the scarcity ofspectrum resources becomes more and more sharp. Wireless spectrum is so preciousthat its allocation and utilization must be authorized by legal departments. However,the fixed allocation of spectrum in this way has reduced the utilization efficiency ofthe spectrum resource. Cognitive Radio is an emerging technology that has thepotential of dealing with the scarcity of the radio spectrum. The basic starting point ofthis technology is to detect the idle spectrum resources of the licensed users, and thenfully utilize them to achieve that multi-user share the same frequency band. The mostimportant problem is how to identify the spectrum holes accurately and to detect thepresence of the licensed users. Spectrum sensing is an important technology to solvethis problem. So, around this hot topic, we begin our study in this paper.Generally, the study on spectrum sensing is classified as single user detection andmulti-user cooperative sensing. Firstly the dissertation systematically investigatesthree kinds of single user spectrum sensing techniques based on transmitter detection.Meanwhile, it analyzes the performance of each method according to the simulationresults, and summarizes the advantages and disadvantages so as to learn theirapplications. As cooperative spectrum sensing successfully overcomes the ‘hiddenterminal’ problem by combining multiple users’ sensing results, it significantlyimproves the accuracy of spectrum sensing. We then make further study oncooperative sensing technology on the basis of single user spectrum sensing. Detailedanalysis of information fusion algorithm has been carried out from two points of viewthat one is decision fusion, the other is data fusion. In addition, comprehensivecomparisons have been made between different kinds of information fusionalgorithms in centralized sensing networks.So far, most of the cooperative spectrum sensing methods need centralized controller or coordinator like fusion centers, which decreases the robustness,flexibility and scalability of the whole system. In order to overcome this question, wefinally focus on spectrum sensing in decentralized networks, introducing a fullydistributed approach based on URW-BP algorithm to solve the problem of cooperativesensing. The core of the proposed approach is the representation of the networkthrough a probabilistic graphical model. Then we can realize cooperative spectrumsensing through probabilistic inference on the probabilistic graphical model.Considering the spatial correlations existing between pairs of neighboring nodes andthe temporal evolution of the state of each node may have influence on the sensingresults, we emphasize these two factors in our graphical model. Probabilistic inferenceon the resulting graph can be performed by URW-BP, which is an improved BeliefPropagation algorithm. Finally we get the estimations of the probabilities of channeloccupation for secondary network nodes which can provide convenience for furtherspectrum allocation and management. Our new method significantly improves therobustness, flexibility, effectiveness and scalability of the sensing network comparedwith conventional cooperative spectrum sensing approaches.