Inversion of H-ARMA models

We present in this contribution the problem of nongaussian H-ARMA models inversion. We show that very-classical methods of parameters identification based on the likelihood are unefficients in our case and we have chosen a fractionnal distance minimisation approach to estimate the nonlinearity. The ARMA coefficients are identified with maximum likelihood estimators and a comparison study with the cumulant based method has been conducted on synthetic data.