Optimal Parallel Wavelet ECG Signal Processing

Domazet, Ervin and Gusev, Marjan (2017) Optimal Parallel Wavelet ECG Signal Processing. In: PROCEEDINGS of the 14th Conference on Informatics and Information Technology. Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University in Skopje, Macedonia, Skopje, Macedonia, pp. 97-102. ISBN 978-608-4699-07-1

[img]
Preview
Text
978-608-4699-07-1_pp97-102.pdf

Download (636kB) | Preview
Official URL: http://ciit.finki.ukim.mk

Abstract

Real time detection of heart abnormalities can prevent serious health problems. This requires real time processing of ECG data by a corresponding web service. Considering the case of wearable devices to collect ECG data, the signal is actually contaminated by noise. Noise can seriously change the ECG signal and occur in the form of a baseline drift representing various physical movements and breathing. Unless it is removed, correct analysis on ECG data is impossible. Being characterized by very low frequencies, its elimination can not be efficiently realized by simple DSP filters, such as Finite Response Filters (FIR) or Infinite Response Filters (IIR). Wavelet Transformation is a promising technique to eliminate the noise with very low frequencies, and its digital version (DWT) is capable of efficient removing the ECG baseline drift. In this paper, we set a research question to investigate the dependence between the nodes in the DWT implementation (and therefore to their corresponding threads) and the available number of cores that can execute the code. This analysis leads to valuable conclusions that will allow construction of even better optimizations. Results indicate that proper allocation of cores can yield faster code.

Item Type: Book Section
Subjects: International Conference on Informatics and Information Technologies > Parallel Processing
?? CIIT_ CLOC ??
International Conference on Informatics and Information Technologies > Computer networks
Depositing User: Vangel Ajanovski
Date Deposited: 29 Nov 2017 18:32
Last Modified: 29 Nov 2017 18:32
URI: http://eprints.finki.ukim.mk/id/eprint/11383

Actions (login required)

View Item View Item