Sequential Monte Carlo simulation of dynamical models with slowly varying parameters: An extension

In this paper, we improve on the slow time-varying partial correlation (STV-PARCOR) model recently suggested by us [2] to include any deterministic interpolator. We then suggest a modification to the on-line filtering algorithm to accomodate the changes. It is believed that the modification will improve on the simulation results as it takes into account the underlying trend of the parameter evolution. The suggested algorithm is tested with real speech data and preliminary results are shown and compared with those generated using existing approaches.