Research On Cooperative Spectrum Sensing Of Cognitive Radio Based On Machine Learning

The main purpose of spectrum sensing is employed to detect spectrum holes accurately and quickly in complex wireless environment,which provides a reasonable way for unauthorized users to occupy the idle time of authorized channels.Thus,spectrum sensing is generally regarded as the cornerstone of cognitive radio technology.Spectrum sensing is essentially to judge the state of the frequency band.Considering the problem of improving system accuracy in spectrum s ensing environment with low signal-to-noise ratio(SNR)and the stability of system performance in spectrum sensing environment with malicious attacks,based on the proposed methods,this paper combines clustering algorithm in machine learning with spectrum sensing technology,several methods of collaborative spectrum sensing based on clustering algorithms are proposed.First,a cooperative spectrum sensing method based on coefficient and clustering algorithm is introduced.In order to improve the performance of spectrum sensing with low Signal-to-Noise Ratio(SNR)and less cooperative users,the method extracts signal features by combining Decomposition and Recombination(DAR)with the Correlation Coefficient(CC).Furthermore,a k-means clustering algorithm-based classifier is trained to improve the accuracy of classification results.The classifier can accurately divide the signal features from unknown environments into two classes,avoid the derivation of complex decision threshold in traditional sensing algorithms.In the experiment part,comparing with different popular algorithms under the same conditions,this method improved the spectrum sensing performance effectively under the condition of few cooperative users and high noise uncertainty.Second,a robust cooperative spectrum sensing algorithm based on clustering algorithm is proposed.In order to deal with the more complex actual communication environment and against malicious user attacks on the sensing system.Based on the principle of soft fusion mechanism,two clustering-based data fusion schemes are proposed.One of them is data fusion algorithm based on k-medoids clustering algorithm,which is called as DF-medoids algorithm in this paper.The other is called as DFMS-medoids algorithm,which is based on mean-shift clustering algorithm.The effect of malicious usage on data fusion is eliminated by iterative method,and real feature statistics are obtained.No malicious users need to be identified during this process.Further,two clustering algorithms-based Cooperative Spectrum Sensing(CSS)frameworks are constructed,which are the C SS framework based on fast k-medoids clustering algorithm and the CSS framework based on mean-shift clustering algorithm.The CSS framework is used in the sensing part,the final statistical samples are studied unsupervised by clustering algorithm,which i s divided into two categories to realize spectrum sensing.In the experimental section,the performance of these sensing data fusion methods is given,and the classification effect of clustering algorithm is analyzed.Finally,the performance of the propos ed robust cooperative spectrum sensing method is verified.