Adaptive overlapping-group sparse denoising for heart sound signals. (February 2018)

• Record Type: Journal Article
• Authors: Deng, Shi-Wen; Han, Ji-Qing
• Is Part Of: Biomedical signal processing and control. Volume 40(2018)
• Page Start: 49
• Page End: 57
• Publication Date: 2018-02
• Subjects: Heart sound signal denoising -- Overlapping-group sparsity -- Regularization optimization -- Adaptive denoising -- Majorization–minimization (MM) algorithm; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2017.08.027
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: The algorithm is proposed based on the group sparsity of the first-order difference. The algorithm does not need the predefined base function. The algorithm outperforms the conventional wavelet methods in lower noise level. Abstract: The heart sound (HS) is an important physiological signal of the human body and can provide valuable diagnostic information in the clinical auscultation. The HS signal, however, is often contaminated by noise and the noisy HS signal will cause adverse influence of making the diagnosis. In this paper, we proposed an adaptive denoising algorithm, named adaOGS denoising, based on the overlapping-group sparsity (OGS) of the first-order difference of the HS signal. Under the Bayesian framework, the adaOGS algorithm is derived and solved as an optimization problem with OGS regularization based on the majorization–minimization (MM) algorithm. Compared with the conventional wavelet method, the proposed algorithm has the advantage that it does not need the predefined base functions and can also be performed in an adaptive way according to the noise level. Moreover, the experimental results show that the proposed algorithm outperforms the conventional wavelet methods such as 'db10', 'db5', and 'bior5.5', for denoising the noisy HS signals in lower noise level.