Low-complexity equivalent realizations of nonlinear multirate filters

In this paper low-complexity equivalent realizations for multirate Volterra systems are presented. All analysed filter configurations contain in addition to the nonlinear filter both an upsampler and a downsampler. Using Volterra filters a wide class of nonlinear systems can be approximated with arbitrary precision. Also special cases, like cascades of linear filters and memoryless nonlinearities (LNL filters, Hammerstein model, Wiener model) are considered. To derive the new realizations linear and nonlinear polyphase decompositions are employed. All operations are performed at the lowest possible sampling rate. Furthermore, some coefficients disappear in the polyphase representation and therefore, the computational complexity can be reduced significantly.