A simplicial canonical piecewise linear adaptive filter (Q1762613)

The paper studies a nonlinear adaptive filtering algorithm based on the so-called canonical piecewise linear (CPWL) approximation of nonlinear functions. In particular, a representation using the concept of simplicial partition of the domain of interest is considered on the real line (SPWL). The resulting adaptive filter consists of a linear finite impulse response system followed by a static nonlinearity (represented by an SPWL map) and is characterized by low computational and implementation costs due to its reduced number of parameters. The convergence properties of the filter are also discussed and its performance is illustrated by simulation examples.