A Frequency Domain LMS Algorithm based on Semidefinite Programming

Y. Wakasa and K. Tanaka (Japan)


adaptive filters, LMS algorithm, semidefiniteprogramming, active noise control.


The frequency domain least mean square (LMS) algorithms are attractive due to the faster convergence compared to the time domain ones. However, there is still room for im provement of the convergence speed by exploiting the free dom of the step size parameter in the recursive rule. This paper proposes a design method for the frequency domain LMS algorithm which is robust in some sense and con verges faster than the conventional one. The design prob lem is reduced to a semidefinite program which is an ef ficiently solvable optimization problem. A simulation ex ample is provided to illustrate the effectiveness of the pro posed method.

Important Links:

Go Back