New split algorithms for linear least squares prediction filters with linear phase

Abstract

This paper is concerned with the development of new split algorithms for the design of linear least squares prediction filters with linear phase. The proposed fast algorithm, which fully expresses the inherent symmetry of the problem, requires lower computational complexity than other existing ones. Moreover, unlike other existing ones, the new recurrences involve only the order updates, which lend themselves to more efficient hardware implementations. For parallelization consideration, a new split Schur-like algorithm is also proposed to overcome the nonparallelizable inner product. Some numerical simulation results are provided to verify the proposed fast algorithms and highlight possible applications.