Recursive identification of time-varying systems via incremental estimation

A new approach, recursive incremental estimation, is presented for identification of time-varying systems. The basic idea is to assume a locally changing system model, instead of a constant one as in most variants of the standard least squares algorithm, and then to identify the characteristics of this model. The derived algorithm can enlarge the memory length of the estimator while maintaining comparable tracking ability, so that some of the major problems experienced with existing algorithms can be reduced.