Some structured matrix approximation problems

Abstract

An improved structured matrix approximation approach for simultaneous estimation of frequencies and wavenumbers from 2-D array data is proposed. A quasi-linear relationship of the error with the polynomial coefficients of both the spatial and temporal domains is derived. This leads to an iterative optimization of the error criterion simultaneously in both the domains. By performing simulations it is shown that the method is capable of resolving signals closely spaced in frequency and wavenumber at low SNR. Next, the extendibility of the method for least-squares fitting of Toeplitz/Hankel/data matrix to a given non-Toeplitz/Hankel/data matrix is also discussed.<>