A RLS variable-length sliding-window nonlinear filtering algorithm for system identification and adaptive noise cancellation

A new recursive least-squares (RLS)-nonlinear filtering algorithm of second-order discrete Volterra type with variable length sliding-window is derived. In order to provide better tracking capability the length of the finite memory window can be adapted recursively according to the changing time-varying speed of a nonlinear (NL) system. Viewing noise cancellation as an input-output identification problem, this algorithm can be used directly as an NL adaptive noise canceller. Simulations are done to demonstrate the performance of this algorithm, and to compare it with the corresponding exponentially weighted prewindowed algorithm