A fast covariance type algorithm for sequential least-squares filtering and prediction

Fast implementation of sequential least-squares (LS) algorithms is of great importance in various applications of signal processing, estimation, system identification, and control. The purpose of this note is the introduction of an efficient sequential LS algorithm for multichannel unwindowed signals (covariance case). The new scheme, in the single channel case, requires 10 MADPR (multiplications and divisions per recursion), being the number of estimated parameters. This offers a saving of 5 MADPR compared to other existing algorithms.