Higher order cumulant based parameter estimation in nonlinear time series models

Parameter estimation in nonlinear time-series models based on higher order cumulant matching is proposed in this paper. The cumulant estimates are computed from measured and simulated data and a cost function computed from second and fourth order cumulants is used. To simplify the calculations a reduced cost function is suggested using low-dimensional slices of the cumulant functions. The estimation method is illustrated on a numerical example