Research On Localization Algorithms Based On Compressed Sensing For Mixed Sources

Sources localization is one of the important research issues of array signal processing.The research of traditional sources localization technique is mainly based on subspace theory.However,the subspace-based algorithms inevitably have some drawbacks and limits.In recent years,with the raising and improvement of compressed sensing theory,scholars notice the deep link between array signal processing and sparse reconstruction,which has been gradually applied to sources localization research and shows some unique advantages.In sources localization research,according to the range between sources and sensor array,sources can be divided into two types: far-field and near-field source.Traditional sources localization technique mainly based on pure far-field or pure near-field model.However,in many practical applications,far-field sources and near-field sources often exist simultaneously and form the mixed sources model which the traditional localization techniques may failing in.Recently,the algorithm research of mixed sources localization is slow and has limited achievement which can be mainly divided into two types: the eigen-subspace-based localization algorithms and the sparsereconstruction-based algorithms.Based on the comparisions and analyses of the present algorithms,this thesis make two improvements,then extend to the cyclostationary signal case,research third-order cylic moment and sparse-reconstruction based localization algorithm.The main contents of the thesis can be summarized as follows:1.Consider the mixed sources localization model of symmetric uniform linear array,analyze the differences and links between mixed sources localization and traditional localization,then simulation,compare and analyze several representative algorithms.2.On the issue of fourth-order cumulant has higher cumulative variance,we propose an improved method of the Two-stage MUSIC localization algorithm,not only improve the range estimation accuracy of near-field source but also decrease the computational complexity.3.On the drawback of only utilizing partial data in the sparse-reconstruction and MUSIC based localization algorithm,we propose an improved method by combining the covariance matrix differencing with the sparse reconstruction and get higherestimation accuracy of near-field source.4.Consider the cyclostationary signal case,we research a algorithm which based on third-order cylic moment and sparse reconstruction.The simulation results show that the proposed algorithm not rely on the prior knowledge of source number and have better eatimation performance in the low SNR situation.5.Based on the horizontal project that the author mainly participate in,makes some researches of DOA estimation that based on the compressed and sparse reconstructed data.