Research On Spectrum Sensing Technology In Alpha Stable Distributed Noise Environment

As wireless communication technologies like 5G and 6G networks continue to advance rapidly,the increasing number of wireless devices has resulted in limited spectrum resources and low spectrum utilization.Spectrum sensing,a radio technology that detects and analyzes signals in the wireless spectrum,is crucial for improving spectrum utilization and management efficiency.It finds applications in wireless communication,radar,satellite communication,and wireless positioning.However,the presence of non-Gaussian noise in real-world environments adds complexity to wireless communication.This study focuses on investigating spectrum sensing algorithms under non-Gaussian noise,specifically using Alpha stable distribution noise modeling.By employing signal estimation and detection theory,the research delves into the comparison of spectrum sensing algorithms in the presence of Gaussian noise and Alpha stable distribution noise.The main research objectives are as follows:This thesis first studies the covariance matrix spectrum sensing algorithm in the classical spectrum sensing algorithm in the Gaussian noise background with a cognitive wireless network equipped with multiple spatially correlated antennas as an example.In a multi-antenna system,the detection performance of such algorithms is greatly affected if the number of antennas is small.The thesis introduces the method of "sampling signal splitting and recombination" into this type of algorithm and analyzes the performance of the perception algorithm maximum eigenvalue to the mean value of the remaining eigenvalues,using the "sampling signal splitting and recombination" method.The thesis then proposes the method of "cyclic shifting of the sampled signal to receive and recombine" to further improve the detection performance when the number of antennas is small.Secondly,in the context of Alpha-stable distribution noise,the thesis analyzes all the eigenvalues of the matrix,introduces the geometric mean of the matrix eigenvalues,and improves the maximum to minimum eigenvalue algorithms and difference between the maximum-minimum eigenvalue algorithm.in the context of Alpha-stable distribution noise.The thesis proposes the spectrum-aware algorithm difference between maximum eigenvalue and geometric mean of eigenvalue of fractional low-order covariance matrix.The perceptual threshold of this perception algorithm is re-derived using fractional low-order statistics theory.The theoretical analysis and simulation experimental results show that this algorithm is more suitable for impulsive noise environment and has better perception performance under low signal-to-noise ratio compared with maximum to minimum eigenvalue algorithm and difference between the maximum-minimum eigenvalue algorithm.Finally,the thesis addresses the problem that the spectrum sensing algorithm based on fractional low-order moments does not have the adaptive capability in the face of dynamically changing non-Gaussian noise background,and proposes a spectrum sensing algorithm that can adaptively sample signals and gives the optimal sample point closure solution for spectrum sensing by a single user.The simulation results show that the proposed spectrum sensing algorithm has good adaptive capability in the noise power dynamically changing Alpha-stable distribution noise background compared with the traditional fractional low-order moment spectrum sensing algorithm.