Direction Of Arrival Estimation In Impulsive Noise Environments

Wireless passive location is one of the most important research branches of array signal processing.It is widely involved in many military and national economy applications,such as radar,communicaitons,sonar,earthquake,radio astronomy and biomedical engineering.Direction of arrival(DOA)estimation is one of the most classical research problems in wireless passive location.Most of the existing DOA methods rely heavily on the Gaussian assumption of the underlying noise.However,in some of the practical wireless passive location environments,such as atmospheric environments,sea clutter,ground clutter,radar backscatter echoes,and urban mobile radio channels,sudden bursts or sharp spikes are exhibited at the array outputs which can be characterized as impulsive.In this thesis,the DOA estimation in extremely high impulsive noise environments along with low generalized signal to noise ratio(GSNR)levels and few available snapshots,the DOA estimation for noncircular sources in impulsive noise environments,and the dynamic DOA estimation in impulsive noise environments are intensively studied.The main contributions are listed as follows.(1)To deal with the DOA estimation in extremely high impulsive and low GNSR noise environments,inspired by the "mix norm" property of the correntropy induced metric(CIM),a new cost function based on CIM is formulated.Combining the cost function with the M-estimation theory,the correntropy based correlation(CRCO)is proposed.Incorporating CRCO with the MUSIC(multiple signal classification)technique,a novel DOA estimation algorithm is proposed which presents better performance than the existing DOA algorithms do in highly impulsive noise environments or in low GSNR situations.Despite the robustness of CRCO-MUSIC,the formulation for the robust CRCO statistics needs quite a number of snapshots.For the DOA estimation in the scenario that only few snapshots are available,based on the investigation of Cauchy distribution and M-estimation theory,a novel cost function of the residual error matrix is proposed for the search of the signal subspace.The resulting nonconvex problem is solved by alternating convex optimization.With the obtained signal subspace,the DOA estimates are retrieved by the MUSIC technique.The simulation results demonstrate that,with few snapshots,the new proposed algorithm can still achieve robust performance when the GSNR is fairly low,or the underlying noise is extremely impulsive.(2)To address the DOA estimation for noncircular signals in impulsive noises which can be modeled as complex symmetrical alpha-stable(SaS)processes,in this thesis,the non-circularity is extended to SaS signals by the utilization of covariation.And then,the covariation based matrix for the extended sensor array outputs is fonnulated and proved to span the same characteristic subspaces as the covariance matrix for the extended sensor array outputs which is formulated under Gaussian noise assumptions does.Applying the covariation matrix with the classical subspace techniques,the robust DOA estimation algorithms for noncircular sources are proposed.Besides,the fractional lower-order parameter p is discussed through the study of the modified fractional lower order moment estimator for covariation.Finally,to evaluate the performance,the Cramer-Rao bound for noncircular sources' DOA estimation in SaS noise scenarios is derived.The simulations demonstrate that the proposed direction finding algorithms outperform the traditional DOA estimation algorithms for noncircular signals in the presence of a wide range of impulsive noise environments.(3)In this thesis,the dynamic DOA estimation is studied by being transformed into the problem of subspace tracking.Motivated by the "mix norm" property of CIM,the maximum correntropy criterion(MCC)is applied as a substitute for the MSE(mean square error)criterion in the PAST(projection approximation subspace tracking)algorithm.Based on the recursive least square(RLS)technique,the MCC based PAST algorithm is developed.To extend the tracking capability of the MCC-PAST for substantially varying subspaces,guided by the improved measure of subspace variations,a variable forgetting factors(VFF)technique is developed and employed in the recursion.Moreover,to further enhance the tracking performance,the technique of nonlinear principle component analysis(PCA)is combined in the algorithm.Before the iteration,the Gaussian function based nonlinear transformation is performed on the projection approximation so as to restrain the impulsiveness more effectively.Compared with the existing algorithm,the new proposed algorithm presents more robust tracking performance especially when the GSNR is fairly low or the underlying noise is extremely impulsive.