A unified derivation for fast estimation algorithms by the conjugate direction method

Fast estimation algorithms inspired by the classical method of Levinson have been developed in the areas of time series analysis, system identification, and signal processing. This paper provides a unified derivation for the Levinson-Durbin-Whittle-Wiggins-Robinson, fast recursive least squares (RLS), ladder (or lattice), and fast Cholesky algorithms as special cases of the conjugate direction method (CDM). This gives a novel derivation and interpretation for all these methods.