| Abstract |
The LMS algorithm, which is widely used in the adaptive filtering community, has been proved to be H(infinity) optimal in [1]. In [2] we have analyzed the other performance measures in the H(infinity) setting which are of direct relevance to adaptive filtering and system identification. In that paper we considered the system identification and estimation employing exponential window problems. This problems are basically of rank I updating class, where we have to update the estimation as the new information comes into picture, while reducing the effect of the past data with a predefined factor. Due to this the effect of past data is not removed completely. The H(infinity) performance measure in the situation of removing the past data effect completely and optimal H(infinity) filter in this situation was still an open problem. In this paper we examine the performance measure in the H(infinity) setting employing sliding window. We present explicit algorithms and the achievable bound in this case. |
ICAR Research Data Repository for Knowledge Management