A New Tvar Modeling in Cascaded Form for Nonstationary Signals

In the nonstationary process identification, a time varying autoregressive (TVAR) model may possess temporal model instability in the conventional direct form. In this study, we propose a new TVAR cascaded form to overcome model instability. The model stability in TVAR the cascaded form is accomplished through the parameterization of time varying poles using pole tracking and monitoring methods. Our simulation on synthetic data demonstrates that the TVAR cascaded form stability can easily be achieved, monitored, and controlled. The performance evaluation of TVAR cascaded model has shown that the Cartesian coordinate with orthogonal representation performs better than other pole representation