RLS polynomial lattice algorithms for modelling time-varying signals

A novel autoregressive RLS lattice algorithm for modelling time-varying signals where the model parameters are allowed to vary as polynomials over each effective data window is presented. Since the model parameters are allowed to vary over each effective data window, the tracking properties of the algorithm are much better than those of the classical RLS algorithms, and the lattice implementation ensures good numerical robustness. This is achieved at the cost of a greater computational load than that of the classical RLS algorithms. The exact computational load depends on the order of the polynomials used. Having a model of the time-varying characteristics of the model parameters (instead of just parameter estimates) greatly expands the possibilities in such applications as prediction and fault detection. Numerical examples and comparisons to other RLS algorithms are presented