Research On Algorithm Of Information Fusion In Cognitive Radio Networks

Cooperative spectrum sensing is an important method of spectrum detection incognitive radio (CR), furthermore, information fusion is the key technology in theprocess of cooperative sensing. Different optimization methods were employedaccording to the different features of spectrum information. Information fusionalgorithms for cognitive radio were described in this dissertation.A novel decision fusion algorithm of distributed multi-objects by reweighted isproposed to resolve the poor efficiency of information fusion by single sub-band incognitive radio, and the problems that optimization could not achieve with fixedweight during the fusion process. The algorithm converted the spectrum informationinto decision form by threshold-processing, adaptively selected the optimalcollaborative users and its numbers through combining the users and channelinformation, and solved the optimization problem by gradient method duringiteration. Multiple sub-bands were simultaneously detected, and the utilization rateof entire system was improved. The experimental results showed that the averagedetective probability raised about13%in the same condition, and it was alsoimproved in the environment of low SNR.Due to the high computational complexity caused by matrix calculating duringthe process of information fusion, the spectrum estimation fusion based on quickalternative direction multipliers method reported in the dissertation couldsimultaneously satisfy the requirements of both speed and computational complexity.Each CR communicated the estimated spectrum information with each other, itcontained more information than decision form. The algorithm improved theaccuracy by constraints of the squared error. During the process of iterations,alternative direction multipliers method was applied, and the parameters wereprovided by augmented Lagrange function. The corresponding parameters generatedduring each iteration, and the iteration completed until the error accuracy met therequirement of residuals convergence. The simulation results showed that themethod could satisfy the usage of cognitive radio, and improve the convergence time.Higher detection probability and lower false detection could be obtained by the newalgorithm comparing with other algorithms in the same test conditions.