In this paper, it is proposed to improve the robustness of the Fast Recursive Least Squares (FRLS) algorithms with the exponential weighting, which is an important class of algorithms for adaptive filtering. It is well known that the FRLS algorithm is...
In this paper, it is proposed to improve the robustness of the Fast Recursive Least Squares (FRLS) algorithms with the exponential weighting, which is an important class of algorithms for adaptive filtering. It is well known that the FRLS algorithm is numerically unstable with exponential weighting factor λ<1.
However, introducing some gains into this algorithms, numerical errors can be reduced. A proper choice of the gains then leads to a numerically stable FRLS algorithm with a complexity of 8m mulitiplications and we shown it by computer simulations.