Research On Cooperative Spectrum Sensing Algorithm Based On Support Vector Machine

Spectrum sensing is the premise of cognitive radio and has important theoretical research value.In the actual sensing process,cognitive users are distributed in a two-dimensional space.The distance and relative position of each cognitive user and the primary user are quite different,resulting in a large difference in the signals received by each cognitive user.Moreover,the noise in the actual communication channel is uncertainty noise.And the perceived noise in each cognitive user has a large difference,which leads to the complexity of the cognitive user?s sensing signal.The traditional energy detection does not solve the influence of spatial distribution and noise uncertainty on spectrum sensing.In this thesis,the support vector machine(SVM)algorithm is applied to the spectrum sensing process for spatially distributed cognitive radio networks and uncertainty noise.Cognitive users submit their sensing data to the fusion center,which statistics the obtained feature vectors.The spectrum sensing algorithm classification parameter is obtained throught support vector machine.Due to noise uncertainty,the Fisher criterion is used to find the kernel function parameters with the highest degree of separation of the feature vector in the feature space.For the problem that all data points have the same "degree of contribution" to the classification surface,the K-nearest neighbor algorithm is used to obtain the membership parameter for the fuzzy support vector machine training,and the weight of the noise and the wild point is reduced.The simulation results show that the cooperative spectrum sensing algorithm based on kernel space optimization and fuzzy support vector machine effectively improves the spectrum sensing performance.When the number of cognitive users is large,the energy vector dimension also increases,resulting in a longer training time.Aiming at this problem,this thesis studies the cooperative spectrum sensing algorithm based on support vector machine with the probability vector as the feature vector.The simulation results show that the cooperative spectrum sensing algorithm based on support vector machine using probability vector significantly reduces the training duration and classification decision time.