Wavelet-based denoising method for real phonocardiography signal recorded by mobile devices in noisy environment

• Published in: Computers in biology and medicine Vol. 52; pp. 119 - 129
• Main Authors: Gradolewski, Dawid, Redlarski, Grzegorz
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.09.2014; Elsevier Limited
• Subjects: Algorithms; Applied Signal Processing; Colleges & universities; Colored noise; Decomposition; Heart; Heart - physiology; Humans; Internal Medicine; Kalman filters; Methods; Mobile healthcare; Monitoring systems; Mortality; Neural networks; Noise; Noise cancellation; Other; Phonocardiography (PCG); Sensors; Signal Processing, Computer-Assisted; Tillämpad signalbehandling; Wavelet transforms; Wavelet transforms Phonocardiography (PCG) Colored noise Noise cancellation Mobile healthcare; Wavelet transforms; Mobile healthcare; Phonocardiography (PCG); Noise cancellation; Colored noise
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2014.06.011

Abstract The main obstacle in development of intelligent autodiagnosis medical systems based on the analysis of phonocardiography (PCG) signals is noise. The noise can be caused by digestive and respiration sounds, movements or even signals from the surrounding environment and it is characterized by wide frequency and intensity spectrum. This spectrum overlaps the heart tones spectrum, which makes the problem of PCG signal filtrating complex. The most common method for filtering such signals are wavelet denoising algorithms. In previous studies, in order to determine the optimum wavelet denoising parameters the disturbances were simulated by Gaussian white noise. However, this paper shows that this noise has a variable character. Therefore, the purpose of this paper is adaptation of a wavelet denoising algorithm for the filtration of real PCG signal disturbances from signals recorded by a mobile devices in a noisy environment. The best results were obtained for Coif 5 wavelet at the 10th decomposition level with the use of a minimaxi threshold selection algorithm and mln rescaling function. The performance of the algorithm was tested on four pathological heart sounds: early systolic murmur, ejection click, late systolic murmur and pansystolic murmur.