Title :
The block-processing FTF adaptive algorithm
Author_Institution :
IBM Research Laboratory, San Jose, CA
Abstract :
A new order-recursive least-squares algorithm for adaptive processing of discontinuous blocks of data is presented. For any one of these data blocks, the new algorithm is significantly more efficient in both total and "per-order" computation than any others that solve this same problem, including those recently introduced by Morf et al (1977), by Marple (1981), and by Kalouptsidis, Carayannis, and Manolakis (1985). In addition, the new numerically stable algorithm permits the recursive incorporation of new blocks of data, unlike those previous algorithms, which cannot be modified for use with more than one data block. The new algorithm does not vary block length with order during the intermediate computations for a particular block and is derived using the geometric Fast Transversal Filters (FTF) concepts that have also been used in time-recursive least-squares algorithms.
Keywords :
Adaptive algorithm; Costs; Equations; Laboratories; Length measurement; Linear systems; Nonlinear filters; Time measurement; Transversal filters; Vectors;
Conference_Titel :
Acoustics, Speech, and Signal Processing, IEEE International Conference on ICASSP '85.
DOI :
10.1109/ICASSP.1985.1168388