Sector-focused stability for high resolution array processing

Abstract

When the ambient noise field is spatially correlated the eigenvectors of the cross spectral matrix associated with the lowest eigenvalues are nearly orthogonal to all planewave arrivals within the region of noise concentration, making their nulls unstable for high resolution bearing estimation. By modifying the quadratic minimization problem that yields these eigenvectors as solutions, to include a subspace constraint dependent on a sector of interest in wavevector space, we obtain high resolution bearing estimators that are stable within the chosen sector, while reducing the size of the matrix on which we must operate.