Research And Improvement On The Cooperative Spectrum Sensing In Cognitive Radio

Cognitive radio arises to be an important solution to the spectrum scarcity problem by finding spectrum space holes to improve the spectral efficiency of licensed radio frequency bands. Spectrum sensing is one of the most important constituents for the establishment of cognitive radio networks which could identify the available spectrum and prevent the harmful interference with licensed users. Ensuring cooperation among secondary users could gain improved detection performance. In this thesis, we will focus on the cooperative spectrum sensing technique and make some improvements.We start by explaining various aspects of cognitive radio and spectrum sensing and introducing the existing local spectrum sensing techniques including energy detection, matched filtering detection, etc. To overcome the hidden terminal problem, in which a single user is unable to detect primary users, several cooperative spectrum sensing methods are investigated. However, cooperative gain can be affected by malfunctioning or malicious users in sensing networks.In the thesis, we will propose a centralized cooperative sensing algorithm which utilizes multi-threshold method on local nodes to cope with different signal-to-noise ratios (SNR), and makes synthetic judgment in fusion centre by introducing functions of distance between bodies of evidence. Safety coefficient can be used to identify selfish nodes. Simulation results reveal that recommended algorithm, while encounters several complex environments including noises of great uncertainty and selfish nodes that attempt to occupy channels exclusively, can get better perfonnance in the probability of detection, the probability of false alarm and security.We also study a mixed distributed cooperative sensing approach using traces of matrix and estimation of eigenvalues based on random matrix theory. The Poisson Pareto burst process which caters to multi-mode sensing will describe our primary users’activities. As far as the low SNR and the small number of users are concerned, the performance results are provided to confirm the efficiency of the proposed approach which avoids evaluation of eigenvalues.