Research On Gridless Compressed Sensing Method For Direction Of Arrival Estimation

As an important part of array signal processing,Direction of arrival(DOA)estimation has been widely used in radar,sonar,medical,communication and other fields.Sparse and parametric approach(SPA),as one of the representative gridless sparse methods in DOA estimation,has a high degree of attention at home and abroad.But,in the experiment,we find that SPA tends to underestimate the power of coherent source in the case of large snapshots.In addition,in a low SNR environment,SPA may estimate one real source as two sources with similar direction of arrival.We call this phenomenon "frequency splitting."How to improve the estimation result of the source power and reduce the occurrence of frequency splitting is the main content of this paper.Inspired by the relationship between SPA and atomic norm,we found that in the case of large snapshots,SPA is equivalent to a weighted atomic norm method.SPA underestimates the power of the source,precisely because the weighted function used in this method is very sensitive to the correlation between the sources.In view of the fact that SPA and atomic norm method can accurately estimate the direction frequency of coherent sources,we take the estimation result of direction frequency as a known condition,and combine the estimated noiseless array signal to calculate the power of each source.The simulation results show that,the new method successfully realized the improvement of the source power estimation results.In addition,we have observed that SPA tends to underestimate the power of noise,thus inferring that the frequency splitting is actually a phenomenon of signal overfitting due to unknown noise power.Therefore,we use SORTE(Second ORder s Tatistic of the Eigenvalues)algorithm to estimate the number of sources,and pre-estimate the noise power based on the estimation results.This approach reduces the frequency splitting,and simulation examples verifies our conclusions.Finally,we combine the above two improved algorithms.The new algorithm is not only suitable for power estimation of coherent sources,but also reduces the occurrence of frequency splitting in a low SNR environment.Compared with SPA,the accuracy and reliability of estimation results are improved.