Algorithm Design And Implementation For Multiband Signal Reconstruction Based On Compressed Sensing

Multiband signal is one of the most common signal types in wireless communication and RF signal transmission.To utilize the idle spectrum or demodulate the received signal,it is important to determine the location of carrier frequency.At first sight,it seems that sampling the whole signal at twice of the highest frequency based on the Shannon Nyquist theorem is necessary.However,in order to resist the interference and increase the capacity,the carrier frequency of the RF signal is usually very high,which causes enormous strains on the front-end ADC,and the excessive amount of data will cause difficulties for subsequent storage,transmission and processing.Applying Compressed Sensing to multiband signal sampling will reduce the sampling rate without affecting the signal recovery,which no doubt will greatly reduce the signal sampling and storage cost,decrease the processing time,bring a new dawn for signal processing.The research focus of this thesis are reconstruction algorithm design and implementation for multiband signal based on compressed sensing.The main contents of this paper are as follows:Firstly,the purpose,significance and research status of multiband signal sampling and reconstruction based on compressed sensing are analyzed,and the advantages and disadvantages of existing compressed sensing reconstruction algorithms are analyzed.Compressed sensing theory and analog signal compression sensing techniques are studied.The AIC random demodulation scheme and Xampling scheme are simulated,and the Xampling scheme is transformed into C language.Secondly,the matching pursuit algorithm and its derivation algorithms are studied.Four classical algorithms are extended to solve MMV problem,and the performances of different algorithms are compared in the case of multiband signal reconstruction.An improved algorithm based on correlation coefficients is proposed to improve the reconstruction accuracy of the existing algorithms.Sparsity adaptive Co Sa MP and SP for MMV problem are proposed to reconstruct the signal when the sparsity of the signal is unknown.Clustered ROMP algorithm is proposed to reduce the sampling rate.Simulation result shows that the improved algorithms achieved the desired goals.All the algorithms above are transformed into C language.Finally,the sparse Bayesian learning algorithm and its derivation algorithms are studied.Four classical MMV sparse Bayesian learning algorithms is extended,so that they can be applied to complex situation and reconstructing multiband signal,The performances of different algorithms are compared.TMSBL,which is the highest-performing of them all,is improved so that its adaptability and robustness to noise is promoted.The multiband signal reconstruction problem is also converted into block Sparse Bayesian Learning(BSBL)problem,which provides a new solving approach.Three BSBL algorithms are extended to complex situation and their performances are compared.