Research On Cyclostationary Feature Based Spectrum Sensing

As an intelligent wireless communication technology,cognitive radio allows secondary users to get opportunistic access to an idle licensed spectrum band under the premise of not causing any influence on the communication of primary users.Spectrum sensing is the key technology and fundamental procedure in the field of cognitive radio.The main function of spectrum sensing is to detect spectrum holes in the licensed spectrum band as the criterion of dynamic spectrum allocation by sensing the surrounding radio environment.Cyclostationary feature based spectrum sensing is a key research direction of spectrum sensing,the core concept of which is detecting the presence of primary user signals by the cyclostationary feature of received signals.Focusing on the cyclostationary feature based spectrum sensing,this thesis makes further researches in aspects of independent spectrum sensing and cooperative spectrum sensing.The main work and contribution of this thesis can be concluded as follows.1.In aspect of independent spectrum sensing,this thesis proposes a novel scheme that applies the low-rank and sparse decomposition algorithm to cyclostationary feature based spectrum sensing.On the basis of traditional cyclostationary feature based spectrum sensing,this scheme decomposes the cyclic spectrum matrix into two matrices,of which the low-rank one represents the noise floor and the sparse one retains the cyclostationary features in the cyclic spectrum matrix.Taking advantage of the cyclostationary features in the sparse matrix to make spectrum decision can effectively reduce the influence of noise and thus improve detection probability.Simulation results show that under low SNR conditions,the proposed scheme achieves better spectrum sensing performance and stronger robustness than traditional cyclostationary feature based spectrum sensing as well as energy detection.2.In aspect of cooperative spectrum sensing,this thesis proposes a novel cooperative scheme based on low-rank and sparse decomposition and machine learning.According to the scheme,secondary users in the cognitive radio network perform the improved cyclostationary feature based spectrum sensing,and report the feature vector composed of features in the sparse matrix to the fusion center.After collecting sensing results from all secondary users,the fusion center makes final decision of the presence of primary user signal by employing machine learning algorithms such as KNN,support vector machine and artificial neural network to make cluster analysis.Simulation results show that compared to independent spectrum sensing and traditional cooperative spectrum sensing,the proposed scheme achieves further optimization in terms of spectrum sensing performance under low SNR conditions.