Direction Finding Method Research Based On Compressed Sensing And Sparse Component Analysis

The electronic reconnaissance will face more complicated electromagneticenvironment because of the development of military electronic and communicationtechnology. The use of all kinds of high frequency wideband signals and the existenceof various complex application environments put forward severe challenge to the arraydirection-finding technology. The traditional direction-finding method all base on theNyquist sampling, in addition to some inherent problems such as low resolution andunable for coherent sources and spatial aliasing, for wideband signals, the data arrayreceived and the burden of hardware acquisition equipment becomes very huge.Compressed sensing theory which is based on the sparse signal processing theory canachieve the information acquisition and the parameter estimation for a signal inunder-sampling mode. It is in urgent need to study the application to the direction ofarrival estimation field.In this thesis, we are engaged in the work in above. Then, the novel and improvedalgorithms are proposed:1. The high-accuracy narrowband direction-finding algorithms are studied in-depth.Based on the eigenvalue decomposition, a novel algorithm which can achieve reduceddimension of the received data and high resolution is proposed. The reduction operationtransforms the direction of arrival estimation problem into solving a multiplemeasurement vectors problem which is appropriate for multiple orthogonal matchingpursuit algorithm.2. The high resolution wideband direction-finding algorithms are studied in-depth.Based on the traditional broadband model, we present a broadband source slice-sparserepresentation model. Combined with the greedy algorithm, a novel multiplemeasurement slices orthogonal matching pursuit algorithm is proposed to exploit thejoint frequency processing for wideband scenarios. It has higher accuracy and betternoise robustness. Besides, based on the harmonic source model, a new algorithm calledwideband harmonic source model sparse representation is proposed. This model usesthe superposition of a limited number of harmonic waves to approximate an authentic broadband signals. The estimation can be transformed into a sparse recovery problemwhich can be solved by convex optimization techniques. The flat spectra constraint isresolved compared with the spatial-only sparse representation method.3. We make a preliminary study of the direction-finding of distributed sources.Based on the block-sparse model, we propose a joint norm convex optimization method.The performance improvement compared with the traditional method is revealed in thesimulations, which provides a new train of thought.