Sparse representation based direction-of-arrival estimation in nonuniform noise via tail minimisation

DOA estimation based on sparse representation in a non-uniform noise environment is proposed using a tail minimisation technique. The noise-free covariance matrix is modelled as diagonal and off-diagonal components relating to incoming signals. In order to combat coherent signals, the auto- and cross-correlation entries of the incident signal are considered separately by using the least square (LS) criterion. Subsequently, from the perspective of difference co-array a vectorised covariance matrix is obtained to emulate the received signal similar to the output of a single snapshot. The effect of non-uniform noise is significantly mitigated via a removal operation. A tail minimisation scheme that is capable of deriving sharper spectra peaks is then applied to detect the DOAs. The proposed approach has the advantages of a larger virtual array aperture beneficial to DOA estimations of uncorrelated and/or coherent signals. The technique can suppress false spectral peaks to a certain extent and is also computationally effective. Numerical simulation verifies the effectiveness of the proposed method.