The paper shows that the problem of noncausal identification of a time-varying FIR (finite impulse response) system can be reformulated, and solved, as a problem of smoothing of the preestimated parameter trajectories. Characteristics of the smoothing filter should be chosen so as to provide the best tradeoff between the bias and variance of the resulting estimates. It is shown that optimization of the smoothing operation can be performed adaptively using the parallel estimation technique.
展开▼