On Spectrum Sensing And Channel Occupancy Estimation In Cognitive Radio

In recent years,with the development of communication industry,the shortage of spectrum resource is becoming more and more serious.By probing the surrounding electromagnetic environment cognitive radio(CR)access the licensed frequency band opportunistically.This greatly improves spectrum utilization.To avoid interfering licensed users,cognitive radio needs to accurately perceive the spectrum state and estimate the channel occupancy.This thesis focuses on the robust spectrum sensing method and the channel occupancy estimation algorithm.The main contributions are as follows:(1)Propose a Friedman-test based blind orthogonal frequency division multiplexing(OFDM)spectrum sensing method.It is observed that white Gaussian noise has a flat power spectral density,while the received OFDM signal through multipath fading channels often does not.The proposed detector exploits the difference of power spectral densities between white Gaussian white noise and OFDM signal.Closed-form expression for the decision threshold of the proposed method is derived.It is shown that the decision threshold does not depends on the number of samples,which will greatly reduce the implementation complexity in the scenario of varying number of samples.Numerical results validate the expression for the decision threshold and demonstrates the effectiveness of the proposed method.(2)Propose an asymptotically unbiased estimator of channel occupancy rate of licensed users.Existing estimators in the literature suffer from overestimation or underestimation at low signal-to-noise ratio(SNR)and with small number of samples.To overcome this issue,we propose an iterative estimator,which achieves an asymptotically unbiased estimate at low signal-to-noise ratio(SNR).The convergence of the iterative method is proved.Given a convergence probability,the requirement of minimum number of samples is derived and the computational complexity is analyzed.Extensive simulation results are provided to demonstrate the superior performance of the proposed estimator compared to state-often-art estimators.