An Innovative Adaptive Noise Canceler Family for Cardiac Signal Filtering: Application to Wireless Body Sensor Network

Abstract

In clinical scenario, during acquisition through a sensing system the cardiac signal (CS) encounters both physiological and non-physiological contaminations. These components mask the tiny features of the cardiac activity and affects diagnosis. To avoid gradient noise amplification problem in gaussian environment, we used normalization with higher order algorithm. This results in variants of least mean fourth (LMF) algorithms. The excess mean-square error of the LMS algorithm is depends only on the second order moment of the noise but excess mean-square error of the LMF algorithm depends on fourth moments of the noise that results in lower steady-state error as compared to the LMS algorithm. Based on normalization quantity, data and error normalized LMF algorithms facilitate adaptive noise cancellers (ANC) for CS de-noising. Finally, we tested the proposed implementations on original cardiac signals acquired from the MIT-BIH database and analyzed their performance with the basic LMF based ANC. The results show that the performances of the proposed normalized higher order algorithms are superior to the LMF counterparts in gaussian environment.