Study On Robust DOA Estimation Method Under Non-gaussian Noise

Direction-of-arrial(DOA)estimation,which is widely used in radar detection,underwater sonar,aeronautical navigation and other fields,is a very important research direction in array signal processing.There are a large number of coherent signals in the space that can affect the DOA estimation performance,due to the multipath propagation of signals or the presence of electronic interference.In addition,the traditional DOA estimation algorithm is usually proposed in the context of Gaussian noise.However,many noises in the actual environment exhibit non-Gaussian characteristics and have obvious pulse spikes.As a result,these noises can cause serious degradation or even failure of the performance of the DOA estimation algorithms based on Gaussian distribution.This thesis mainly focuses on the above two aspects of DOA estimation.Firstly,through comparing and further analyzing some DOA estimation algorithms based on fractional lower-order moments under non-Gaussian noise.Secondly,the research finds some limitations of the these methods:(1)It is usually necessary to model the non-Gaussian noise in advance and DOA estimation can be carried out according to relevant parameters in the distribution model;(2)The number of sources must be known in advance to ensure the accuracy of the estimation.Based on hypothesis testing theory,a DOA estimation algorithm for coherent signals with unknown source number based on SW(Shapiro-Wilk)test is proposed in this thesis.The algorithm considers the non-Gaussian noise in the signal as outlier based on the robust estimation theory,the signal with non-Gaussian impulse noise is pre-processed using the SW test,the outlier in the signal data is removed in an adaptive manner;then the preprocessed data is used as a new input signal to obtain its covariance matrix;next,the Toeplitz matrix is constructed for each row of the covariance matrix,it makes the rank only related to the direction of arrival of the signal,without the effects of the correlation of signals to achieve the purpose of decoherence;finally,a cost function without the number of sources is generated according to the joint diagonal structure,and the DOA estimation is performed by a one-dimensional spatial spectrum search derived from the cost function.Through simulation experiments under different signal-to-noise ratios,snapshots and other conditions,the results verify that the proposed algorithm still has a robust estimation ability for the target signal under the conditions of low signal-to-noise ratio,and it can work in low snapshots cases,compared with some existing methods,the proposed algorithm has higher efficiency and accuracy.