Discrete-wavelet-transform-based noise removal and feature extraction for ECG signals |
| |
| Institution: | 1. Department of Electrical Engineering, National Tsing Hua University, Hsinchu, Taiwan;2. Division of Cardiology, Department of Internal Medicine, National Taiwan University Hospital and National Taiwan University, Taipei, Taiwan;2. Deusto Institute of Technology – DeustoTech (University of Deusto), Av. Universidades 24, 48007, Bilbao, Spain;1. Ohio University, Stocker Center EE P. Box #71, Athens 45701, USA;2. Ohio University, Stocker Center 322A, Athens 45701, USA;1. Department of Biomedical Engineering, KUET, Khulna, Bangladesh;2. Faculty of Design and Engineering Technology, University Sultan Zainal Abidin (UniSZA), Kuala Terengaanu, Terengganu, Malaysia;1. LAMIH, UMR CNRS 8201 UVHC Laboratory of industrial and Human Automation, Mechanics anc Computer Sciences, Université de Valenciennes et du Hainaut Cambrésis, Bat Malvache, 1er étage, bureau 204, Le mont Houy, 59313 Valenciennes Cedex 9, France;2. Laboratoire Signaux et Images (LSI), Département Electronique, Faculté Génie Electrique Université USTO-MB, B.P 1505, El M’Naouar, Bir el Djir- Oran, Algeria;3. Laboratoire Signaux et Systèmes (LSS), Université Abdelhamid Ibn Badis de Mostaganem, Route Belahcel, 27000 Mostaganem, Algeria;1. MTech Student, AECE of ECE Department, Rajiv Gandhi Institute of Technology, Kottayam, Kerala-686501, India;2. Principal of Govt. Engineering College, Thrissur, Kerala-680009, India |
| |
| Abstract: | Nowadays, doctors use electrocardiogram (ECG) to diagnose heart diseases commonly. However, some nonideal effects are often distributed in ECG. Discrete wavelet transform (DWT) is efficient for nonstationary signal analysis. In this paper, the Symlets sym5 is chosen as the wavelet function to decompose recorded ECG signals for noise removal. Soft-thresholding method is then applied for feature detection. To detect ECG features, R peak of each heart beat is first detected, and the onset and offset of the QRS complex are then detected. Finally, the signal is reconstructed to remove high frequency interferences and applied with adaptive searching window and threshold to detect P and T waves. We use the MIT-BIH arrhythmia database for algorithm verification. For noise reduction, the SNR improvement is achieved at least 10 dB at SNR 5 dB, and most of the improvement SNR are better than other methods at least 1 dB at different SNR. When applying to the real portable ECG device, all R peaks can be detected when patients walk, run, or move at the speed below 9 km/h. The performance of delineation on database shows in our algorithm can achieve high sensitivity in detecting ECG features. The QRS detector attains a sensitivity over 99.94%, while detectors of P and T waves achieve 99.75% and 99.7%, respectively. |
| |
| Keywords: | |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
