MSEMG: Surface Electromyography Denoising with a Mamba-based Efficient Network

TL;DR

This work tackles ECG contamination in surface electromyography by introducing MSEMG, a lightweight denoising framework that fuses the Mamba state space model with convolutional components. By leveraging a selective state-space mechanism and Half Normalized Filters, MSEMG captures both local and long-range temporal dependencies with low computational cost, enabling real-time-like denoising. Across NINAPro DB2 sEMG data and MIT-BIH NSRD ECG signals, MSEMG achieves higher $SNR_{imp}$ and lower RMSE than HP, TS, FCN, and SDEMG, while using far fewer parameters than the diffusion-based baseline. The method shows strong robustness across varying SNRs and holds promise for downstream tasks such as prosthesis control and gesture recognition, with future work extending data and applications.

Abstract

Surface electromyography (sEMG) recordings can be contaminated by electrocardiogram (ECG) signals when the monitored muscle is closed to the heart. Traditional signal processing-based approaches, such as high-pass filtering and template subtraction, have been used to remove ECG interference but are often limited in their effectiveness. Recently, neural network-based methods have shown greater promise for sEMG denoising, but they still struggle to balance both efficiency and effectiveness. In this study, we introduce MSEMG, a novel system that integrates the Mamba state space model with a convolutional neural network to serve as a lightweight sEMG denoising model. We evaluated MSEMG using sEMG data from the Non-Invasive Adaptive Prosthetics database and ECG signals from the MIT-BIH Normal Sinus Rhythm Database. The results show that MSEMG outperforms existing methods, generating higher-quality sEMG signals using fewer parameters.

Figures (4)

• Figure 1: The Selective State Space Model.
• Figure 2: The denoising framework and the model architecture of MSEMG.
• Figure 3: Comparison of denoising methods with SNR$_{imp}$ under various SNR conditions.
• Figure 4: Performance under the scenario simulating trunk sEMG with ECG contamination.