Research On Idle Spectrum Detection Technologies In Cognitive Radio

With the rapid development of wireless communication technologies, wireless service is now focusing on broadband multimedia business. However, the lack of wireless spectrum is becoming the obstacle of implementing these service. Dynamic spectrum access is considered an effective approach to solve shortness of wireless spectrum resources and lowness of spectrum utilization. Cognitive radio can dynamically reuse those idle authorized spectrum in temporal and spatial domain, and provide usable spectrum resources for new wireless applications. Therefore, cognitive radio is considered as an ideal carrier of implementing dynamic spectrum access in the future. As an key technology of cognitive radio physical layer, idle spectrum detection is the premise of cognitive radio system to operate normally. Starting from the request of quickness, exactness, and sensitiveness by spectrum detection, This dissertation studys idle spectrum detection deeply.In chapter one, first, the particularity of spectrum detection in cognitive radio is discussed. Second, the spectrum detection technologies are classified in detail. Third, the merit and shortcoming of various spectrum detection methods has been summrized,and the outline of representative literature of each spectrum method is expatiated at the same time.In chapter two, first, it is introduced the based concepts and rules of signal detection. Second, to overcome the serious shortcoming of failing to noise uncertainty by energy detection, it is proposed a blind spectrum detection algorithm based on generalized likehood ratio rule by this dissertation. First, the unknown parameters of likehood function are estimated through maximum likehood ratio rule under two hypothesis. Then, it is produced a detection statistic which is not ralated to nosie power by likehood ratio test rule. This statistic is free of noise uncertainty through mathematical deduction and computer simulation. When the noise uncertainty exists, if the length of received data and SNR are the same, the simulation results show that the detection probability of GLRT algorithm exceeds the energy detection clearly.In chapter three, first, the basic theory and esitimation methods of cyclic spectral are expatiated. Second, it is introduced the cyclic spectral statistic detection algorithm brought forward by Dandawate and Giannakis. Starting from deducing cyclic spectral detection statistic constructing complexity,how to set the threshold and improving detection probability, an cyclic spectral statistic spectrum detection algorithm based on freqency domain smoothing(FSCSS) is proposed by this dissertation. First, the distribution of detection statistic is analyzed theoretically. Second, it is deduced analytic expression of the false alarm probability and detection probability under AWGN channel and flat Rayleigh fading channel, the threshold setting is solved at the same time. Then, the FSCSS algorithm which has construct detection statistic at one cyclic frequency is extended to construct detection statistic at two frequency. The simulaiton results show when the length of received data and SNR are the same, the detection probability of FSCSS algorithm exceeds the DG algorithm.In chapter four, because of the merit of resisting fading, the multi-antenna technology can be combined with other physical layer detection methods, so as to further improve detection probability. First, the cyclic spectral statistic spectrum detection algorithm based on freqency domain smoothing is combined with multi-antenna technology, then it is deduced analytic expression of false alarm probability, and the threshold setting is solved at the same time. Second, analytic expression of the false alarm probability and detection probability of multi-antenna equal gain combining energy detection method are present. Third, it is introduced the optimal linear weighted combining detection algorithm proposed by Quan etc. At last, the three detection algorithms are compared through simulaiton, the results show that, when the length of received data and SNR are the same, the detection probability of the proposed algorithm is the best of all.In addition, starting from deducing calculation of cyclic spectral detection algorithm, a compromising scheme is proposed. Next, starting from maximizing the average throughput of sensing user, the optimization of detection time of compromising multi-antenna cyclic spectral statistic spectrum detection algorithm is study. It is proved the optimal detection time is existent through mathematical deduction and computer simulation. The simulation experiments give the performance curve of throughput under different number of antennas and paremeters. The simulation results show that increasing the number of antennas can not only improve the maximum average throughput, but also shorten the optimal detection time.In chapter five, to resist deep fading and improve reliability of spectrum detection, a new cooperative spectrum detection algorithm based on centralized decision fusion is proposed. The proposed algorithm starts from maximizing the global cooperative detection, solves optimized the fusion rule of control center and thresholds of each cognitive user by sequential quadratic programming. First, the mathematical description of cooperative detection algorithm is present. Second, solution of uniting judging probability are given under different channel. Third, the optimizing problem of cooperative detection algorithm, solution method, implementation steps are expatiated. At last, the soft decision optimal cooperative spectrum detection algorithm and optimal cooperative spectrum detection algorithm under correlated decision are discussed. The simulation experiments get results as followed: first, whether from the curve of detection probability versus SNR, or from receiver operating characteristic curve, compared with other cooperative detection algorithm whose fusion is fixed, the detection probability of the proposed algorithm is always the best. Second, when the detection probability is the same, the detection time is shorter by users cooperation.Third, when SNR is the same, compared with hard decision algorithm, the detection probability of soft decision algorithm has only a little improvement. And with the raise of sensing users, the improvement is worse and worse. Fouth, compared with two users independent, the bigger the correlated coefficient of two users is, the bigger the loss of improvement of cooperative detection probability is.