Research On Spectrum Sensing Algorithms Based On Energy Detection In Cognitive Radio Networks

The core idea of cognitive radio is to access and utilize the licensed frequency bands opportunistically without causing interferences on the licensed users to improve the utilization of spectrum. Therefore, spectrum sensing is considered as a key technology in cognitive radio networks(CRNs). Among all existing sensing algorithms, energy detection is widely used in CRNs since its simple implementation and low computational complexity. However, when there exist malicous users, when the channel conditions become complicated and changeful caused by user mobility and when there exists the problem of noise uncertainty, the detection performance of spectrum sensing algorithms based on energy detection will become quite terrible. Aiming at those three practical problems in actual systems, this thesis proposes three solutions, i.e., multi-bit decision cooperative spectrum sensing algorithm based on trust value, reputation-based cooperative spectrum sensing algorithm and weighted correlation-based blind detector.First, single-node energy detection and energy-detection-based multi-node spectrum sensing algorithms, such as hard decision, soft decision and multi-bit decision methods which allow the compromises between system overheads and detection performance, are studied and relevant conclusions are verified through simulation experiments. Then a few existing problems of energy detection are illustrated from two aspect, i.e., theoretical analysis and simulation experiments. To be specific, issues like hidden terminals lead to terrible performance of single-node sensing. As to multi-node sensing, though it is able to weaken the restrictions of single-node sensing by utilizing diversity gains, it offers malicious users(MUs)opportunities to attack, which may result in dramatically degradation of detection performance. When it comes to the noise uncertainty problem, this thesis explains its detailed impacts on sensing system performance by presenting simulation results and then provides feasible solutions, i.e., eigenvalue-based spectrum sensing algorithms. Finally, after comparison tests, it is verified that eigenvalue-based algorithms can actually overcome the negative effects caused by noise uncertainty.A cooperative algorithm based on the idea of trust value under the frame of multi-bit decision is proposed which is able to remove MUs, aiming at the problem that malicious attacks in cooperative systems would lead to terrible performance.And simulation experiments are conducted to illustrate that the proposed algorithm indeed is able to precisely remove MUs and significantly increase the performance of CRNs attacked by MUs. However, the mobility of users is an inherent property of wireless networks and conventional trust-value-based algorithms always tend toidentify reliable users which move into areas in deep fading as MUs mistakenly.Thus they cannot be applied in a mobile scenario directly. Aiming at this problem,this thesis proposes another reputation-based cooperative algorithm which adequately considers the channel condition disparity to remove mobile MUs. The principle of identifying MUs and the effectiveness of date fusion are explained from the aspect of theoretical derivation. Finally, the MU identification capability and the accuracy of detection primary users(PUs) are compared between our proposed algorithm and other trust-value-based cooperative algorithms by simulation,verifying the outstanding performance of the proposed algorithm.To cope with the two problems of noise uncertainty and complicated channel conditions caused by cognitive user mobility, this thesis first studies existing covariance-based spectrum sensing algorithms, derives the expressions of corresponding detection performance metrics and explains the reason with they are robust against noise uncertainty. And again, simulation results are presented to prove the correctness of derivation results. Considering a significant decline in performance of existing algorithms in a more practical scenario with weak signal correlations and low signal-to-noise ratio(SNR), a weighted spectrum sensing algorithm based on signal correlations is proposed to increase the performance further. The proposed algorithm can achieve better performance for the following two reasons. On the one hand, it adequately utilizes both the auto correlations and the cross correlations of received signals by secondary users(SUs). And on the other hand, it assigns each test statistic term with a proper weighting coefficient. The effectiveness of the proposed weighting operation is demonstrated. Moreover, the expressions of performance metrics of the proposed algorithm are derived and verified through the fact that theoretical results match simulation results well.Finally, numerous simulations are carried out to prove the detection performance of the proposed algorithm is better than that of other existing covariance-based ones.