Research On Technologies For Wide-band Signal Compressive Sensing

Recently,Compressive Sensing(CS)has become an important research issue in signal processing field.It mainly deals with the problem of sparse signal measurement and reconstruction.A lot of applications using CS theory were proposed these years,especially in the realm of image processing and Cognitive Radio(CR).The signal CS process should have sparse property,which means the signal is either sparse by itself or sparse in certain transform domain.In fact in many CR scenarios,signal may not be sparse if we inspect the spectrum within a narrow bandwidth.However,if we measure it from a wide range of bandwidth,we may still find the frequency occupancy is sparse,thus CS can be well adopted in CR scenario.Compared to traditional signal acquisition method,CS acquisition method can save a lot of sampling cost.Instead of collecting plenty of samples then dumping those redundant parts by coding technique,CS mechanism directly measure the compressed results so as to avoid the waste of sampled information.Using CS as a tool of signal acquisition,we can break Shannon's constrains and collect the signal in sub-Nyquist sampling rate.This paper aim at solving several important issues on CS,namely,the architecture of compressed measurement for wide-band signals and the reconstruction algorithms on considering the intra-relation and inter-relation of sparse signals.The main contributions of this paper are listed as follows:1)A dual branch CS structure for analog signalis proposed.By taking advantage of the inner correlation ofcomplex-valued frequency,this structure greatly improves theperformance of sparse signal reconstruction.2)Adistributed compressive sensing algorithm using Bayesian hierarchy and Bayesian inference is proposed.By applying probability to model the inter-correlation between common sparse signals and using Gibbs sampler to solve the model,the proposed algorithm greatly improves reconstruction performance.