Research On Approaches Of Robust Cooperative Temporal Spectrum Sensing For Cognitive Radio

The rapid development of the mobile Internet and the Internet of Things has directly led to conflicts between limited spectrum resources and rising demand.Corresponding to this conflict,under the traditional spectrum allocation method,there is a large amount of waste of spectrum resources in space,time and frequency of authorized spectrum,which results in an increasingly severe spectrum deficit and low spectrum utilization.In response to this basic contradiction,cognitive radio has been introduced as a core technology to ef-fectively improve spectrum utilization and alleviate the shortage of spectrum resources.The primary task of cognitive radio is spectrum sensing.Under the existing spectrum environment,there are also several technical challenges such as unreliable data sources,inaccurate detection results,insufficient opportunity discovery and incomplete historical data.In response to the these challenges,based on the systematic analysis of the charac-teristics of cognitive radio spectrum sensing data,this paper conducts a robust cooperative time-domain spectrum sensing algorithm from the perspective of the cross fusion of radio communication and machine learning.The main research contents are as follows:(1)Systematically analyzing the theory and method of single-node spectrum sens-ing and cooperative spectrum sensing.Firstly,the classical single-node spectrum sensing technology represented by energy detection,matcher filter detection and cyclostationary detection is introduced,and the advantages and disadvantages of various methods are ana-lyzed.Then,it introduces representative centralized and distributed cooperative spectrum sensing methods and their corresponding fusion rules,and summarizes the limitations of various methods.Finally,the spectrum sensing method based on machine learning is analyzed,and existing challenges are summarized.(2)Proposing a spectrum sensing algorithm based on extreme learning machine.This algorithm aims at the resolving the issue of spectrum perception that the spectrum state satisfies periodicity and historical spectrum data has prior information.Firstly,based on historical prior spectrum sensing data,a series of extreme learning machine's spectrum sensing regression models are constructed.Then,the results of different spectrum sensing models mentioned above are fused through an integrated idea.Finally,the fusion result is used to determine the spectrum state.Simulation cases show that this perception model not only effectively improves spectrum perception performance,but also gives a class of improved integrated extreme learning machine models.(3)Proposing a soft and hard fusion cooperative spectrum sensing algorithm based on clustering algorithm.This algorithm aims to resolve the limitation of poor performance of hard fusion sensing algorithm and high complexity of soft fusion algorithm.Firstly,clustering the perception information based on the clustering algorithm,then building a soft and hard fusion mechanism for the clustering information,and finally making a per-ception decision through its output.Compared with the traditional hard fusion and soft fusion models,this algorithm takes the advantages of both models,and the simulation results verify the feasibility and effectiveness of the algorithm.