Research On Joint Spectrum Sensing Technology Based On Compressed Sampling

The problem of spectrum congestion is getting worse due to the explosive growth of wireless users and high bandwidth applications.Aiming at this problem,the cognitive radio constructs a framework for dynamic distribution of spectrum resources based on spectrum sensing.The basic function of cognitive radios is spectrum sensing,whose goal is to locate the potential spectrum holes by spectrum estimation,and thereby meet the spectrum access needs of more users.Two recent research trends have grown up with respect to spectrum sensing technology.One is the joint frequency and direction of arrival(DOA)spectrum sensing based on array technology,and the other is the spectrum sensing based on compressed sampling.The former can reuse the spectrum resource in spatial dimension to improve the efficiency of spectrum utilization,and the latter can cope with the high sampling rate of wideband signals.Combining these two technologies concerning spectrum sensing can simultaneously acquire both advantages.However,problems will arise in terms of system structure,data model,estimation algorithm,and the relative research is deficiency so far.Therefore,in this thesis,an in-depth study of joint frequency and DOA spectrum sensing technology based on compressed sampling is carred out.For three different scenes,namely a subband containing single signal,multisignals,and signals exceeding the number of array elements,two kinds of array receiver structures based on compressed sampling are designed,and corresponding receiving models are established.Furthermore,the Cram?er Rao bounds(CRBs)of the models are analyzed,and three joint frequency and DOA spectrum sensing algorithms,which are based on trilinear decomposition,subspace theory,and iterative maximum likelihood(ML),respectively,are proposed.The main contributions of this thesis are given as follows:1.For the scene with single signal in a subband,array receiving model based on compressed sampling is established via joint space and frequency information.Towards this model,the joint frequency and DOA spectral sensing algorithms based on trilinear decomposition and subspace decomposition are proposed,and the CRBs for spatial phase,frequency,and DOA are derived.It is theoretically proven that with the same samplings,the CRB of the spatial phase estimation of the proposed array model is lower than that of the traditional array model based on Nyqusit sampling,and the estimation performances of the spatial phase between the subbands have no mutual affect.We demonstrate that both the frequency and DOA estimation performances of the proposed algorithms are proximate to the corresponding CRB model,and the trilinear decomposition algorithm based on the singular value decomposition reduces the computation burden caused by big snapshots.2.For the scene where multiple signals exist in a subband,joint space and frequency array receiving model based on compressed sampling is established,and the corresponding joint spectrum sensing algorithms are proposed.Besides,the number of maximum estimable signals in a subband and the whole band are discussed.Similarly,theoretical analysis shows that: with the same samplings,the CRB of the spatial phase estimation of the proposed array model is still lower than that of the traditional array model based on Nyqusit sampling,and the orthogonality of the base vectors between the subbands makes the estimation performances of the spatial phase between the subbands have no mutual affect.It also indicates that the maximum numbers of estimable signals in the subband and the whole band are respectively -1 and(-1),where and are the number of elements and subbands,respectively.The proposed algorithm solves the problem of joint frequency and DOA spectrum sensing towards the scene with multiple signals in a sub-band,and realizes estimate of signals whose number is more than that of array elements.Moreover,the estimated performance reaches the CRB of the corresponding model.3.For the scene with signals whose number is more than that of array elements in a subband,block vectorization is applied to eliminate the effect on the covariance matrix structure caused by compressed sampling.On this basis,a sparse array covariance matrix expansion technique based on compressed sampling is proposed.This technique enables the maximum number of estimable signals in a subband of the compressed sampling array system to improve from -1 which is the number without using this technique up to -1,where ( > )is the number of the virtual uniform linear array corresponding to the physical array.Comparing with the compressed sampling system with uniform linear array of sensors,the compressed sampling system based on the proposed technique balance the estimated performance and the consumption of sensor,thus can be employed to reduce the elements number with acceptable performance loss.4.To reduce the hardware complexity of the receiver system,a simplified array receiver structure is proposed,and the corresponding data receiving model and joint spectrum sensing algorithm are deduced,following the deriving of the CRBs of the parameter estimates.Comparing with the structure that each sensor utilizes all the compressed sampling channels,the proposed simplified receiver structure can realize the joint spectrum sensing at lower data rate,thus providing a solution for compromising the estimated performance and consumption of channel resources.5.For the traditional array model based on Nyquist sampling,the convexity and convex range near the true values of the determined ML estimation cost function in terms of DOA are analyzed.On this basis,an iterative ML DOA estimation algorithm based on compressed sensing is proposed to realize the fast estimation.Furthermore,the convex analysis and estimation algorithm of the traditional array model are extended to the proposed array model based on compressed sampling so as to efficiently realize the joint spectrum sensing under the framework of ML estimation.