| In recent years,wireless communication technologies have continued to evolve,and more and more communication devices require the use of wireless spectrum.At present,spectrum is a valuable and scarce resource that is uniformly distributed and authorized by the state.However,a significant portion of the frequency bands allocated to authorized users are rarely used.Therefore,the existing fixed spectrum allocation str ategy can not meet the increasing spectrum demand.Cognitive Radio(CR)technology is considered as a new intelligent wireless communication method to effectively solve the problem of low spectrum utilization.Spectrum Sensing is a key step in cognitive radio.Its main purpose is to accurately and quickly detect the free spectrum in a complex wireless environment.Thus,the basis for making full use of the idle time of the authorized user not using the spectrum.This paper focuses on applying clustering algorithms in machine learning to spectrum sensing in CR.The spectrum sensing problem can be considered as a two-category problem,that is,whether the authorized user is using the licensed spectrum.Based on the existing spectrum sensing technology,this paper proposes a variety of cooperative spectrum sensing methods based on clustering algorithm.The main innovations are as follows:A cooperative spectrum sensing method based on K-means is proposed.In order to improve the spectrum sensing performance under the condition that the number of collaborative cognitive users is small,a feature extraction method combining IQ decomposition and Decomposition and Recombination(DAR)is proposed.The method first needs to extract the features of spectrum sensing signals,then use the K-means algorithm for offline training to obtain the required cluster centers,and finally calculate the similarity between the received signals and the cluster center feature vectors.To achieve spectrum sensing.The whole process is divided into two parts: offline training and online perception.The offline training phase can effectively overcome the dependence of the classical feature detection algorithm on a priori information without increasing the complexity of the algorithm.In the simulation part,the performance of different spectrum sensing algorithms is compared and analyzed.Experimental results show that the method effectively improves the spectrum sensing performance.Furthermore,an improved method based on Fuzzy c-means for cooperative spectrum sensing is proposed.In order to further improve the spectrum sensing performance when the number of cooperative users is small,a feature extraction method combining splitting and recombining information geometry is proposed.This method transforms the signal detection problem into a geometric problem on the manifold,so the signal detection problem can be visually analyzed geometrically.Finally,the fuzzy c-means clustering algorithm is used for offline training and online perception.The experimental results show that the proposed method can improve the spectrum sensing performance to some extent.Further,an improved method based on K-medoids for cooperative spectrum sensing is proposed.In order to improve the spectrum sensing performance under low signal-to-noise ratio and the number of cooperative sub-users,a feature extraction method based on empirical mode decomposition and split recombination is proposed.The method firstly performs EMD noise reduction processing on the signals perceived by the cooperative secondary users,and then splits and recombines the noise-reduced signals,thereby logically increasing the number of coordinated secondary users and improving signal feature accuracy.Finally,K-medoids clustering algorithm is used for offline training and online perception.Experimental results show that the method further improves the spectrum sensing performance. |
