The noise power gain (NPG) is addressed and the error bound (EB) is specified via NPG for a general p-shift linear timevariant finite impulse response (FIR) transversal estimator intended for filtering (p = 0), prediction (p > 0), and smoothing (p < 0) of discrete-time K-state space signal models with M states measured. A fast iterative algorithm for NPG of the unbiased FIR estimator is provided along. An example is given for the two-state space model.
Yuriy S. Shmaliy, Oscar Gerardo Ibarra-Manzano