A nonlinear architecture involving a combination of proportionate functional link adaptive filters

In this paper, we consider a functional link-based architecture that separates the linear and nonlinear filterings and exploits any sparse representation of functional links. We focus our attention on the nonlinear path in order to improve the modeling performance of the overall architecture. To this end, we propose a new scheme that involves the adaptive combination of filters downstream of the nonlinear expansion. This combination enhances the sparse representation of functional links according to how much distorted the input signal is, thus improving the nonlinear modeling performance in case of time-varying nonlinear systems. Experimental results show the performance improvement produced by the proposed model.