Research On Recovery Algorithms Of Compressed Sensing For Multiband Signals

Nowadays, multiband signals play a significant role in communications. Thebandwidth resources can be harnessed efficiently since different transmitters occupydifferent band ranges. It is important that we acquire and radiate radio signals. Theclassical methods need to approximate frequency spectrum of signals in advance and itshould sample at Nyquist rate if signals are expected to be recovered accurately.Compressed Sensing (CS) appearing recently can sample efficiently without support setas a prior knowledge. And it can compress data when sampling it as it is calledcompressed sensing. This can reduce space of memory and relieve the burden of datatransmitting.Compressed sensing technology existed is utilized to sample multiband signals,and we focus on reconstruction algorithms. Modulated Wideband Convertor (MWC) isused in our experiments, which is a state-of-the-art system that samples multibandsignals in CS frame. The main work of this thesis is to improve reconstructionalgorithms so that it will lead to higher recovery rate. First of all, convex optimizationreconstruction algorithms are discussed to state that Smoothed0Norm (SL0) methodis a superior one. And then we research on the performance of SL0as well as do someanalyses. We expand it into Multiple Measurement Vector model to suit itself in MWCto reconstruct multiband signals which is the real purpose in our work. Analyses andsimulation experiments are conducted to illustrate that the modified SL0method issuperior when channel number of MWC is low, compared with the original algorithm.The second item in our work is to improve the performance of Rank Aware OrderRecursive Matching Pursuit (RA-ORMP). Firstly, we start at researching on greedyalgorithm. Matching Pursuit is a typical greedy algorithm which contains basic thoughts.Then Orthogonal Matching Pursuit is studied. Since the rank of signals is ratherimportant, we studied RA-ORMP. In order to validate RA-ORMP in MWC,3innovation points have been proposed. The high recovery rate caused by modified RA-ORMP and robustness in noise has been illustrated via multitude of simulationexperiments.