Research Of Wideband Spectrum Sensing Algorithms Based On Compressed Sensing

In recent years, the number of wireless communications business users has beengrown rapidly and they have more strict performance requirements and more businesstypes. As a result, the scarce spectrum resources are increasingly crowded. Traditionalfixed spectrum allocation causes the imbalance of spectrum utilization and a lot ofspectrum resources are idle for a long time, causing further congestion of spectrumresources. CR is an intelligent wireless communication system that it can search and accessto spectrum holes adaptively without interference to authorized users by continuallysensing the surrounding RF environment. In a CR system, spectrum sensing is the coretechnology. As the development of future wireless communication technologies needshigh-speed data communications, broadband spectrum sensing technology has become animportant direction of the current study. However, the hardware of broadband spectrumsensing algorithm is facing tremendous pressure. The excessively high sampling rate andbig data overload have become a bottleneck restricting its development. CS combines thesignal sampling and compressive coding theory, so the sampling rate is determined by thesignal structure and the content of information in the signal. Luckily, Low spectrumresource utilization makes broadband signal sparse in the frequency domain, therefore,compressed sensing theory can be applied to the broadband spectrum sensing problem. It isproved that CS theory provides a perfect solution to the dilemma which broadbandspectrum sensing is facing.Timeliness and accuracy are two important measure indexes of spectrum sensing.Some of the existing broadband spectrum sensing algorithms need to be improved in termsof the two performances. In this paper, we have an in-depth research on the broadbandcompressed spectrum sensing algorithms and make improvements for their shortcomings.With prior knowledge of signal sparsity, iterative greedy OMP algorithm canaccurately reconstruct original signal. However, the signal sparsity in practice is oftenunknown due to dynamic changes of the spectrum. In consideration of problem of blindsparsity, an adaptive OMP algorithm is proposed. It estimates reconstruction error of theoriginal signal through additional observations in the iteration process and adaptivelydetermines the iteration corresponding to the best spectrum reconstruction. Simulationresults show that the proposed adaptive OMP algorithm can effectively improve signalreconstruction performance under low SNR.The gradient pursuit based on the steepest descent method has a fast speed initially,but it descents slowly when approaching the optimal solution convergence, affecting the overall running time of the algorithm. The gradient pursuit based on the Newton methodhas a fast global convergence speed, but it requires an initial point near the optimalsolution. In addition, calculating the inverse matrix of Hessian matrix increases thecomplexity of the algorithm and reduces the efficiency of spectrum sensing. Inconsideration of these problems, a gradient pursuit based on hybrid optimizationalgorithm(GNP) is proposed. It combines the steepest descent method and Newton methodwhich are applied to the greedy iterative algorithm. Simulation results show that comparedto the OMP algorithm, GNP algorithm can still guarantee the effect of signal spectrumreconstruction when reducing the computational complexity.