Multi-Feature Adaptive Variational Mode Decomposition for Wearable ECG Devices.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Wearable Technology

Background:

• Motion artifacts significantly degrade the quality of electrocardiogram (ECG) signals from wearable devices.
• Accurate ECG analysis is crucial for diagnosing cardiac conditions, but noise interference poses a major challenge in real-world applications.

Purpose of the Study:

• To develop and validate an adaptive motion artifact removal framework for wearable ECG monitoring.
• To enhance the accuracy and reliability of ECG signal processing in dynamic environments.

Main Methods:

• An improved Variational Mode Decomposition (VMD) algorithm with parameter self-adjustment and multi-feature fusion mode selection was developed.
• The proposed method was evaluated against traditional wavelet transform, Recursive Least Squares (RLS), and conventional VMD using the MIT-BIH Arrhythmia Database.

Main Results:

• The improved VMD algorithm significantly enhanced signal-to-noise ratio (SNR) by 5.17 dB and reduced Percentage Root Mean Squared Difference (PRD) to 49.13%.
• The method demonstrated high real-time processing capability (RTR = 22.5) and preserved clinically significant ECG features in pathological recordings.
• An arrhythmia classification task using a CWT-CNN classifier achieved 91.67% accuracy on denoised signals, a 2.67% improvement over raw signals.

Conclusions:

• The proposed adaptive VMD framework effectively removes motion artifacts from wearable ECG signals.
• The enhanced signal quality supports more accurate AI-based cardiac diagnosis and improves the clinical utility of wearable ECG monitoring devices.