Title :
A performance-weighted mixture of LMS filters
Author :
Kozat, Suleyman S. ; Singer, Andrew C.
Author_Institution :
Koc Univ., Istanbul
Abstract :
In this paper, we explore the use of a particular multistage adaptation algorithm for a variety of adaptive filtering applications where the structure of the underlying process to be estimated is unknown. The proposed algorithm uses a performance-weighted mixture of LMS filters of various orders to construct its final output. The algorithm is analyzed in a stochastic context with respect to its convergence and mean-square error (MSE) behaviors and is shown to achieve the best MSE performance of the constituent algorithms in the mixture. Through simulations, it has been observed that the mixture structure can offer considerable performance improvement for both stationary and time varying observation sequences.
Keywords :
adaptive filters; least mean squares methods; LMS filters; adaptive filtering; minimum mean-square error optimal filtering; multistage adaptation algorithm; performance-weighted mixture; Adaptive filters; Algorithm design and analysis; Convergence; Filtering algorithms; Information filtering; Lattices; Least squares approximation; Performance analysis; Prediction algorithms; Predictive models; LMS; Universal; model combination; model mixture; prediction;
Conference_Titel :
Acoustics, Speech and Signal Processing, 2009. ICASSP 2009. IEEE International Conference on
Conference_Location :
Taipei
Print_ISBN :
978-1-4244-2353-8
Electronic_ISBN :
1520-6149
DOI :
10.1109/ICASSP.2009.4960280
