Fast sequential least-squares processing

A fast algorithm with reduced memory requirements for updating in time the covariance autoregressive model of a time series with constant memory is proposed. The "algorithm is derived using elementary linear algebra concepts and manipulations and is based on the special structure of the covariance matrix characterizing the underlying linear equations. The algorithm is extended to updating the all-zero (or all-pole) least-squares model of a system, computed from its input and output signals.