Fast sequential algorithms for least squares FIR filters with linear phase

Two algorithms are presented that adaptively compute the coefficients of a finite-input response (FIR) transversal filter with linear (affine) phase. The first algorithm is of the least-squares covariance type with a computational complexity of 18.5 m MADS (multiplications and additions), m being the order of the filter. The second one corresponds to the sliding-memory case and requires 24 m MADS. The prewindowed algorithm, a special case of the covariance type algorithm is also discussed and requires 10.5 m MADS. The algorithms are derived for the general multichannel case. The task of initialization is also considered.<
>