Efficient Implementation of LMS Adaptive Filter Using Distributed Arithmetic

Abstract

A LMS adaptive noise canceller has been proposed to remove the maternal heartbeat interference noise from the foetal heartbeat signal by using Distributed arithmetic. DA is capable of computing the inner product of vectors by calculating the partial products with shifting and accumulation. These partial products are stored in look-up table. .Adaptive filter has been proposed with conventional LMS adaptive filter and Sign LMS adaptive filters. It is being found that the noise canceller proposed using sign LMS is more efficient as compared to the conventional LMS adaptive filter. The convergence rate of sign LMS is very fast in comparison to the other adaptive algorithms. The throughput of the sign LMS adaptive filter using distributed arithmetic is high. The maternal heartbeat signal is stronger as compared to the foetal heartbeat signal. The maternal heartbeat signal is measured from the chest and the foetal heartbeat signal is measured from the abdomen. Since, the maternal heartbeat signal is strong it causes interference in the foetal measured heartbeat signal. This interrupted signal is needed to be recalculated so as to remove noise from the foetal heartbeat signal. The analysis has been made on the basis convergence rate, varying step size and mean square error of different adaptive algorithms. The convergence rate is measured on behalf of mean square error vs. Iterations consumed to converge the noise canceller. The conventional LMS adaptive filter converges for thousands of iterations while the LMS adaptive filter using distributed arithmetic converges for hundreds of iterations.