Research On Fast Wideband Spectral Sensing Based On Compressed Sampling In Cognitive Radio

The technology of cognitive radio can improve the utilization of wireless spectrum. Sensing thewideband spectrum quickly and accurately is one of the key technologies in CR, of which highsampling rate and massive data challenge the hardware for spectrum sensing. Compressed samplingmakes it possible to reconstruct the original signal via fewer measurements than that is requiredtraditionally. However, the computing complexity of reconstruction algorithms is very high and onmany signal processing occasions, it is not necessary to fully reconstruct the original signal. Focusingon that, in this paper we use compressed sampling method to solve the high speed sampling problemof wideband spectrum sensing. And we make research on signal detection without fullyreconstruction.The main contribution is shown as follow:1. In this paper, we make further discussion of the compressive detection of random signal andunknown amplitude signal. Firstly, we set up a model of signal detection based on compressivesampling without signal recovery, and then derive the maximum likelihood detection of unknownamplitude signal in detail, and analysis the performance. Theoretical analysis and simulation showthat it can achieve higher detection rates.2. Focusing on the detection of sparse random signal, we propose a sequential compressivesensing scheme. Then we discuss the performance of detection and use it in distribute collaborationspectrum sensing. Theoretical analysis and simulation results show that sequential compressivedetection can significantly save the number of measurements under a given detection performance.This algorithm reduces the detection time, and also avoids the reconstruction of original signal, ofwhich computer complexity is very high.3. We improve one of BPDN, the classic interior-point method and propose partly reconstructalgorithm. In this paper, we simulate in wideband spectrum scenario respectively in the sparse degreeof constant and of change. The result show that this partly reconstruct algorithm can decrease thenumber of iterations without reducing detection performance.