Identification Method for Coupled MIMO Wiener Model with Output Noise

This study introduces a parameter estimation technique for multiple input multiple output (MIMO) Wiener model incorporating measurement noises. In the Wiener model established, the linear block is modeled by auto-regressive with extra inputs (ARX) model, and the nonlinear block is represented through a neuro-fuzzy approach. For the purpose of identification, a combination of independent Gaussian signals and random signals was employed to analyze the MIMO Wiener model, decoupling the identification of linear block from that of nonlinear block. Initially, the correlation analysis technology is applied to estimate the ARX model parameters applying available Gaussian input-output. Furthermore, taking into account the effect of the unknown noise term on the Wiener model identification precision, the estimation of the residuals is used instead of the noise term, then a recursive extended least squares method is developed for updating neuro-fuzzy model weights. Numerical simulation results validate the effectiveness of the proposed identification method.