Fast root-MUSIC for arbitrary arrays

By using the manifold separation techniques, root-MUSIC designed for uniform linear arrays has been extended to arbitrary geometries at the cost of increased computational complexity. A fast algorithm is proposed that exploits the Laurent structure of the polynomial to conduct fast spectral factorisation via the Schur algorithm. Then Arnoldi iteration is employed to compute only a few of the largest eigenvalues. This implies that a large number of the unwanted eigenvalues (or roots) are exempt from the calculation and therefore the computational complexity is reduced significantly.