The Wideband Spectrum Detection Based On Compressed Sensing

Cognitive radio is a new technology which is able to automatically sense theenvironment and detect the spectrum holes,which significantly improve the spectrumutilization. Cognitive radio is one of the hot research field of wirelesscommunications.A novel collaborative spectrum detection based on backtracking blind sparsitymatching pursuit algorithm is proposed for sparse signals with unkown sparsitywhich the cognitive radio users received.The algorithm could control the rapidityand accuracy of spectrum detection by choosing the candidate setautomatically,adopting staged changing process which estimates sparsity andbacktracking mechanism which obtains the global optimal support sets; andchoosing the optimal collaborative users although SNR estimate.The experimentalresults show that the algorithm is superior to other algorithms in the same testconditions,and detection probability about the collaboration detection of selectiveobject is increased by about25%than nonselective object.For the spectrum detection, the key job of the cognitive radio users is not toreconstruct the signal entirely,but is to estimate the presence or absence of theprimary users.So in the paper we adopt the fast sparse bayesian learning to completespectrum detection tasks.The unknown variables are endowed with the hyperparameters about following the certain prior conditional distribution. we update thehyper parameters and select the basis function through the fast algorithm,the numberof basis function is increasing from one to another until it has obtioned the relevantvector,so observation matrix contains only basis functions which exist in the currentmodel. The algorithm eliminated the complex matrix inversion process andimproved the speed.Meanwhile,the posterior distribution of the unknown signal obeythe Student-distribution,which is more sparse than the Gaussian distribution. Theprimary users’ information described with three parameters about the spectrumlocation,the mean and covariance in the posterior distribution.These three parameterswere fused to complete the spectrum detection. The simulation results show that themethod saves the resources and reduces the computational complexity withoutreconstructing the signal.