A robust sliding window adaptive filtering technique for phonocardiogram signal denoising

作     者：Shervegar, Vishwanath Madhava 

作者机构：Department of Electronics & Communication Engineering Moodlakatte Institute of Technology Kundapura India 

出 版 物：《Multimedia Tools and Applications》 (Multimedia Tools Appl)

年 卷 期：2025年

页      面：1-27页

核心收录：

学科分类：0810[工学-信息与通信工程] 0808[工学-电气工程] 0809[工学-电子科学与技术（可授工学、理学学位）] 08[工学] 070206[理学-声学] 0804[工学-仪器科学与技术] 0805[工学-材料科学与工程（可授工学、理学学位）] 0813[工学-建筑学] 0835[工学-软件工程] 0803[工学-光学工程] 0702[理学-物理学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

主　　题：Adaptive filters 

摘      要：Phonocardiogram (PCG) signal is the digital sound recording of various heart sounds. To diagnose the different types of heart disorders, it is often necessary to analyse these PCG signals. However, PCG signal recording is challenging due to disturbing surrounding noise signals. So denoising the PCG signal is done before using PCG for advanced processing. This paper proposes a novel Sliding window Adaptive Noise Cancellers-based filter model for effectively denoising and recovering the PCG signal. This work introduces a sliding Window optimum adaptive filter structure for estimating a noise-free signal with high accuracy using Least Mean Square (LMS) algorithm. A noisy signal is processed through the Sliding Window adaptive filter stage in the proposed work. Sliding Window of fixed duration slides over the signal and in each window, the signal is filtered using the Adaptive filter. The method approximates the clean version of the PCG signal using this Sliding window adaptive filter architecture with high accuracy. The proposed robust Sliding window adaptive filter is tested against experimental PCG signals corrupted by Gaussian and pink noise with different noise levels. The experimental data are taken from the Physionet database and real world recordings. The results show that the robust Sliding Window Adaptive filter model performs remarkably well. Compared with traditional LMS filter configuration, the proposed filter structure achieves a 2–10 times reduction in MSE values. Further, there is an improved SNR by 3 times, PSNR improvement by 4%–25% comparatively. The correlation between the clean signal and its estimate is more than 0.92. Sliding Window LMS adaptive filter model offers a cost-effective hardware implementation of Adaptive Noise Canceller with high accuracy. In the future, such models can be tested for real-time performance for obtaining desired convergence speed and accuracy. © The Author(s), under exclusive licence to Springer Science+Business Media,