Fading‐Memory (Discounted Least‐Squares) Filter
Eli Brookner DSc · 1998
In Chapter 7 the least-squares polynomial fit to the data is given for the case where the error of the fit is allowed to grow the older the data. In effect, or pay less and less attention to the data the older it is. This type of filter is called a fading-memory filter or discounted least-squares filter. This filter is shown to lead to the useful recursive fading-memory g–h filter of Section 1.2.6 when the polynomial being fitted to is degree m=1. Recursive versions of this filter that apply to the case when the polynomial being fitted has degree m=2, 3, 4 are also given. The issues of stability, rms error, track initiation, and equivalence to the growing-memory filters are also covered.