Parameter estimation of single-input multiple-output systems using the finite impulse response approximation and gradient search

This paper considers the parameter estimation problem of a single-input multiple-output (SIMO) system by using some finite impulse response (FIR) models to approximate the fictitious subsystem´s transfer functions. We derive the least squares parameter estimates of the equivalent FIR models with the FIR model orders increasing from available input/output data. Moreover, we use the multi-innovation identification theory to derive a multi-innovation stochastic gradient algorithm for estimating the parameters of the original systems from the estimated FIR model parameters. The proposed algorithm can be extended to other multiple-input multiple-output systems with colored noises.