Comparing the performances of least mean squares based multirate adaptive filters

Abstract

In this study, sign-error, sign-data, and sign-sign modifications of the multirate least mean squares (LMS) filter are proposed as an alternative to the existing multirate LMS and normalized LMS filters for reconstructing the high rate desired signal from its several low rate noisy observations. The performance of all the existing and the proposed methods are compared by using an adaptive noise cancellation simulation example applied to an audio signal. It is seen from the simulation results that some of the proposed methods lead to better results than the existing ones especially in the convergence speed sense.