Statistical averaging and PARTAN-some alternatives to LMS and RLS

Statistical averaging and PARTAN (parallel tangent) are shown as potential alternatives to the conventional least mean square (LMS) and recursive least squares (RLS) for adaptive filtering and signal processing. These new algorithms exhibit the RLS-like fast adaptation, and yet relate well to the popular LMS in conceptual simplicity and in implementation. Compared with the RLS, these new algorithms seem to be more robust in guarding against the signal bursting phenomenon. The computational complexity of these algorithms, in their direct forms, is less than that of the standard RLS but higher than the fast RLS and LMS