Spectrum Sensing Research Based On Signal Separation And Principal Component Analysis

Currently,the number of wireless devices is increasing,and spectrum resources are becoming more and more tense.However,the spectrum utilization of many licensed channels is very low,and the channel is idle for many times and is not used by authorized users,which causes the spectrum resources to become more and more tense.In order to solve the problem of utilization of spectrum resources,cognitive radios have thus been proposed.Cognitive radio is different from conventional fixed thinking.It proposes a new way to use spectrum resources to meet the spectrum requirements by reusing the authorized idle channels,thereby alleviating the shortage of spectrum resources.Spec trum sensing is the primary task of cognitive radio,which improves spectrum utilization by finding and determining the idle spectrum to allocate it to unauthorized users.In order to improve the performance of spectrum sensing in different environments,t his paper combines signal separation,principal component analysis and information geometry to conduct in-depth research on spectrum sensing technology.The main work and research contents are as follows:Firstly,the background and significance of cognitive radio are introduced,and the research status and results of the technology are summarized.Some classic single-user and multi-user spectrum sensing methods are introduced.Through the shortcomings of single-user technology,the multi-user-aware soft combining and hard combining methods are introduced,and the advantages and disadvantages between them are analyzed and summarized.Then introduce some other spectrum sensing technologies to improve the spectrum sensing model and improve detection performance.In this paper,by denoising and decomposing the signal separation,it is introduced into the receiving front end,and the received signal is processed,and then the receiving back end is used for spectrum sensing.By decomposing the signal matrix and co mbining the Choletsky decomposition correlation matrix,a kind of decomposition correlation matrix extraction based on null space pursuit(NSP)and Cholesky decomposition is proposed.The feature scheme then trains the signal features through the k-medoids algorithm to obtain the required cluster centers.Finally,the spectrum perception is realized by calculating the similarity between the received signals and the cluster center feature vectors.Theoretical analysis and experimental simulations show that the proposed algorithm solves the problem of poor performance of low SNR detection and is superior to traditional energy detection and eigenvalue detection.In order to improve the detection performance under low SNR,two improved methods based on Principal Component Analysis(PCA)are proposed.Firstly,a feature extraction method combining principal component analysis and random matrix is proposed.The principal component analysis is applied to the combination scheme of multi-antenna system to extract the signal principal component,and then the corresponding signal characteristics are extracted by the signal principal component matrix.Finally,the K is extracted.K-medoids clustering for spectrum sensing.Then a feature extraction scheme of principal component analysis and information geometry is proposed.According to the difference of statistical features of the perceptual data obtained by occupying the channel,the corresponding statistical manifolds are different points,and then according to the geometry processing on the manifold.To extract the features on the manifold,and finally through the k-medoids cluster training and test to achieve spectrum sensing,thereby further improving the detection performance under low SNR.