Authors:
Shi Zhao
;
Yiding Wang
and
Hong Yang
Affiliation:
Graduate University of Chinese Academy of Sciences, China
Keyword(s):
ECG, noise reduction, wavelet, Besov, nonlinear shrinkage function.
Related
Ontology
Subjects/Areas/Topics:
Applications and Services
;
Biomedical Engineering
;
Biomedical Signal Processing
;
Computer Vision, Visualization and Computer Graphics
;
Medical Image Detection, Acquisition, Analysis and Processing
;
Wavelet Transform
Abstract:
This paper proposes a novel technique to eliminate the noise in practical electrocardiogram (ECG) signals. Two state-of-the-art denoising techniques, which both based on wavelet bases, are combined together. The first one is discussing wavelet bases in Besov spaces. Compared to traditional algorithms, which discuss wavelets in L2 (R) spaces, the proposed technique projects ECG signals onto Besov spaces for the first time. Besov space is a more sophisticated smoothness space. Determining the threshold of shrinkage function in Besov space could eliminate Gibbs phenomenon. In addition, instead of using linear shrinkage function, the proposed algorithm uses nonlinear hyper shrinkage function, which is proposed by Poornachandra. The function tends to keep a few larger coefficients representing the function while the noise coefficients tend to be reduced to zero. Combining the two techniques, we obtain a significant improvement over conventional ECG denoising algorithm.