2013 Volume 6 Issue 6 Pages 381-386
The Hyper H∞ Filter is an adaptive filtering algorithm, where a forgetting factor ρ is built into the H∞ optimization framework as a function of the robustness parameter γf and determined through the process of γ-iteration. This paper examines how the choice of γf affects the resultant steady-state mean-square error of this algorithm. For moderately large γf, a theoretical expression of the excess mean-square error is derived, which turns out to coincide with that of RLS. Based on this expression and numerical simulations, it is shown that there is a trade-off between the robustness and the mean-square performance at steady-state. For balancing this trade-off, preferable values of γf are also considered.