A new transform domain LMS algorithm

In this paper a Krylov subspace transform domain lease mean square (LMS) algorithm is proposed. The unknown system can be sparse after Krylov subspace transform, thus a much smaller tap length can be used for the update of adaptive filer coefficients in transform domain, which results in a significant improvement of convergence rate. The small tap length in transform domain can be found by using variable tap-length LMS algorithm. Simulation is performed to show the advantage of the proposed algorithm. As can be seen from simulation results, the proposed algorithm has an improved convergence rate as compared with the LMS algorithm.