Research On Spectrum Sensing Technology Of Cognitive Satellite Networks Based On Machine Learning

With the advent of the 5G era,we are always looking forward to the intelligent life based on the Internet of Everything,which is about to become possible.The Internet of Everything means the skyrocketing number of terminals in the network,and at the same time it brings end and end.Frequent communication services,which to some extent cause a shortage of spectrum resources.If the satellite network and the ground can be The combination of networks not only improves network stability,but also improves the performance of the original network.So we consider how to combine the satellite network with the terrestrial network while alleviating the shortage of spectrum resources,so we think of Knowing that radio technology is applied to the fusion of satellite networks and terrestrial networks,this is the satellite ground cognitive network we are going to study.The purpose of this paper is to study the application of spectrum sensing technology in cognitive radio in satellite terrestrial cognitive networks.Firstly,th e research background and significance of the subject and the research status at home and abroad are briefly introduced.Then several cognitive radios are introduced in principle.Then the basic model of satellite cognitive network is described,and the application of different bands in the satellite cognitive network model is analyzed.Finally,through the different needs of the satellite cognitive network,the application scenarios of satellite cognitive networks are analyzed by combining the actual situation of different bands of satellite cognitive communication.The three typical satellite cognitive network architectures studied at home and abroad are summarized,and their characteristics are analyzed and compared.The model of centralized satellite cognitive network architecture is established,which improves the spectrum sensing performance and makes the network easy to optimize and manage.And through the theoretical derivation of the sensing process with energy detection as an example,the spectrum sensing performance of a centralized satellite cognitive network is obtained when the cognitive user in terrestrial network perceives a satellite downlink signal.Based on the above research,the K-means clustering and Gaussian mixture model in unsupervised clustering in machine learning are applied to the energy detection algorithm.Through research,it is found that the application of K-means clustering and Gaussian mixture model in energy detection not only improves the detection performance,but also eliminates the threshold derivation process caused by noise uncertainty in traditional energy detection.Then a maximum minimum eigenvalue detection algorithm based on clustering is proposed.The algorithm adopts K-means clustering and Gaussian mixture model,and compared with the original algorithm,the complicated threshold derivation process is eliminated,and the improved maximum minimum eigenvalue algorithm is superior to energy detection in the same conditions.In this paper,CURE hierarchy clustering is applied to energy detection and maximum minimum eigenvalue detection,and is compared with traditional K-means clustering and Gaussian mixture model.Through simulation verification,it is found that in the energy detection and maximum and minimum eigenvalue detection,the CURE hierarchical clustering detection performance is slightly lower than the Gaussian mixture model but better than the K-means clustering.Moreover,compared with Gaussian mixture model,CURE hierarchy clustering has the advantages of low algorithm complexity and rapid process.