Research On Cognitive Radio Spectrum Sensing Method Based On Information Geometry And K-means Clustering

With the rapid development of wireless communication technology and the lack of spectrum resources,the contradiction has become increasingly prominent.As a technology to solve spectrum resource shortages,Cognitive Radio(CR)provides a new solution for improving spectrum sensing resource utilization.Spectrum sensing technology is the core technology in cognitive radio technology,enabling spectrum access by detecting available spectrum in complex wireless environments.Information geometry is developed for the study of the geometric properties of statistical manifolds,providing a new direction for the solution of spectrum sensing problems.With the rapid development of machine learning,more and more scholars have used the algorithms in machine learning to use spectrum sensing.This paper studies the spectrum sensing technology by combining information geometry with clustering algorithms in machine learning.Spectrum sensing is realized by transforming statistical signal detection problems into geometrical problems on statistical manifolds by information geometry theory.The specific research contents are as follows:Firstly,the research status of cognitive radio is introduced.According to the type of spectrum sensing,the difference between single-user independent spectrum sensing and multi-user collaborative spectrum sensing is analyzed.At the same time,the basic principles of several classical single-user independent spectrum sensing methods and their advantages and disadvantages are discussed,and the data fusion methods in multi-user collaborative spectrum sensing methods are introduced.In this paper,the spectrum sensing method based on information geometry and the spectrum sensing method based on machine learning are studied,and the feasibility of spectrum sensing algorithm based on information geometry and clustering algorithm is verified.Secondly,a spectrum sensing method based on information geometry and k-means clustering algorithm is proposed.According to the information geometry theory,the statistical characteristics of the sensing signal can be mapped to the manifold,and the geometric characteristics of the signal can be obtained according to the measurement method on the manifold.Then,the geometric characteristics are put into the k-means clustering algorithm to train to obtain the classifier.Finally,the geometric characteristics of the signal to be perceived are placed in the classifier to achieve spectrum sensing.The effectiveness and superiority of the algorithm are verified by simulation.Thirdly,considering the complex noise environment problem,a spectrum sensing method based on Riemannian mean and Riemannian median is proposed.The difference between Riemann mean and Riemann median is discussed and analyzed.Two reference points are obtained by estimating the noise environment with Riemannian mean and Riemannian median respectively,and then the metric is used to calculate the distance between the spectrum signal and the two reference points to obtain the distance characteristics.At the same time,the k-median clustering algorithm is used to replace the k-means clustering algorithm.The algorithm can effectively reduce the impact of outliers on the detection performance.Finally,for the spectrum sensing problem with low SNR,the spectrum sensing based on signal processing and information geometry is proposed.The sensing signal is filtered by a filter,then the filtered signal is extracted,and then the characteristics is put into the clustering algorithm.The classifier is obtained,and the classifier is used to judge whether the primary user exists.The simulation results show that the proposed method can effectively improve the performance of spectrum sensing in low SNR environment.