Adaptive lattice filters for band-inverse covariance matrix approximations

Theory of adaptive lattice filters, developed originally for stationary processes with Toeplitz covariance matrices, was well-established by the pioneering contributions of Burg, Kailath, Friendlander. The lattice filters are intimately related to the well-known Levinson algorithm, that requires 0(N2) operations for N¿N Toeplitz covariance matrix inversion, and exploit Burg technique for adaptive estimation of reflection coefficients that naturally preserves positive definiteness of the estimated Toeplitz covariance matrix.