Computer Experiments in System Identification Using a Prefltering Bank

This paper proposes a scheme for decreasing the number of parameters identified by adaptive filters with finite impulse response. The real target is to decrease the computational complexity by using the proposed identification scheme. A filter bank is used to decompose the input signal into several components, which are then used as input sequence for an adaptive algorithm. RLS is used as adaptive algorithm; current implementations require a computational complexity that exponentially grows with the number of parameters. Three sets of functions are used as impulse responses for the filter bank: rectangular, orthogonalized exponential and triangular functions. The cases are analyzed from identification accuracy point of view; the third case is also analyzed from obtained quality versus needed computational complexity point of view, and experimental data on RLS convergence is provided.