DOA estimation for coherent and incoherent wideband sources via sparse representation of the focused array covariance vectors

摘要

The frequency smoothing technique makes the coherent signal subspace method (CSSM) capable of resolving the wideband coherent sources. However, when time delay between coherent signals becomes short, its performance deteriorates. To fix this, we propose a novel method based on the sparse representation. After averaging the focused array covariance matrices at different frequencies, we construct the new focused array covariance vectors (FACVs). Then the wideband direction-of-arrival (DOA) estimation is to find the sparsest representation of the FACVs under the newly-formed dictionary by sparse Bayesian learning (SBL). When constructing the focusing matrices, we adopt the sector focusing instead of the point focusing. This method has better performance compared with several CSSMs under both coherent and incoherent scenarios, which is confirmed by the Monte Carlo simulations.