[Paper Review] Denoising ECG by Adaptive Filter with Empirical Mode Decomposition

Bingze Dai, Wenlei Bai|arXiv (Cornell University)|Aug 18, 2021

ECG Monitoring and Analysis29 references5 citations

TL;DR

This paper proposes a novel denoising framework for ECG signals using a combination of Empirical Mode Decomposition (EMD) and adaptive filtering to effectively remove multiple noise types—baseline wander, power line interference, electrode motion artifact, and muscle artifact. The proposed parallel EMD-adaptive filter structure achieves the highest signal-to-noise ratio (SNR) improvement on the MIT-BIH arrhythmia database, demonstrating superior performance in complex, multi-noise scenarios.

ABSTRACT

Electrocardiogram (ECG) signal is an important physiological signal which contains cardiac information and is the basis to diagnosis cardiac related diseases. In this paper, several innovative and efficient methods based on adaptive filter and empirical mode decomposition (EMD) to denoise ECG signal contaminated by various kinds of noise, including baseline wander (BW), power line interference (PLI), electrode motion artifact (EM) and muscle artifact (MA), are proposed. We first present a novel method based on EMD and adaptive filter for the removal of BW and PLI in ECG signal. We then extend the method to the complex scenario where four most common noises, PLI, BW, EM and MA are present. The proposed Parallel EMD adaptive filter structure yields the best SNR improvement on the MIT-BIH arrhythmia database, corrupted by the four types of noises.

Motivation & Objective

• To address the challenge of ECG signal degradation due to multiple overlapping noise sources such as baseline wander, power line interference, electrode motion, and muscle artifacts.
• To develop an efficient, adaptive signal processing framework that enhances ECG quality for accurate cardiac diagnosis.
• To improve signal-to-noise ratio (SNR) in real-world ECG recordings contaminated by complex noise environments.
• To validate the proposed method on a standard ECG database (MIT-BIH arrhythmia) under realistic noise conditions.

Proposed method

• The method employs Empirical Mode Decomposition (EMD) to decompose the noisy ECG signal into intrinsic mode functions (IMFs), isolating noise components from the primary cardiac signal.
• An adaptive filter is applied to the IMFs containing noise, using a reference signal derived from the EMD process to track and suppress non-stationary noise components.
• A parallel structure is designed to simultaneously process multiple noise types—baseline wander, power line interference, electrode motion, and muscle artifacts—by assigning specific IMFs to targeted filtering.
• The adaptive filter uses a least mean square (LMS) algorithm to iteratively adjust filter coefficients based on the error between the estimated and actual noise components.
• The reconstructed signal is formed by summing the denoised IMFs, preserving the morphology of the original ECG while removing artifacts.
• The method is evaluated on the MIT-BIH arrhythmia database with synthetic corruption of all four noise types to simulate real clinical conditions.

Experimental results

Research questions

• RQ1 Can EMD effectively separate ECG signal components from multiple overlapping noise types such as baseline wander, power line interference, electrode motion, and muscle artifacts?
• RQ2 How does the integration of adaptive filtering with EMD improve SNR compared to conventional denoising methods in multi-noise environments?
• RQ3 What is the optimal configuration of the EMD-adaptive filter structure for simultaneous suppression of diverse ECG noise sources?
• RQ4 Does the proposed parallel filtering architecture outperform sequential or single-filter approaches in terms of SNR and signal fidelity?

Key findings

• The proposed parallel EMD-adaptive filter structure achieves the highest signal-to-noise ratio (SNR) improvement among tested methods on the MIT-BIH arrhythmia database.
• The method effectively suppresses all four major ECG noise types—baseline wander, power line interference, electrode motion artifact, and muscle artifact—simultaneously.
• EMD successfully isolates noise components into distinct IMFs, enabling targeted adaptive filtering with minimal distortion of the underlying ECG morphology.
• The adaptive filtering component significantly reduces residual noise, particularly in non-stationary and transient noise conditions.
• The method maintains high fidelity to the original ECG signal, preserving clinically relevant features such as QRS complexes and ST segments.
• The framework demonstrates robustness and adaptability in complex, real-world noise environments, outperforming traditional filtering techniques.