Identification Of Time-Varying Nonlinear System Using Wavelet Decomposition

A new mean of describing time-varying nonlinear system and corresponding identification approach using wavelet decomposition is proposed. The general nonlinear auto regressive with exogenous inputs (NARX) models are provided with time varying parameters to construct time varying NARX models. The time-varying parameters and the time invariant polynomials of NARX model are represented by translations and dilations of wavelet function and thus, a complex identification process is simplified as a linear parameter estimation procedure. The recursive least square (RLS) algorithm can be used to estimate the parameters, which make it feasible to realize on-line identification.