Robust Photoplethysmography Signal Denoising via Mamba Networks

TL;DR

This work tackles the challenge of denoising photoplethysmography (PPG) signals in wearable settings while preserving physiologically important information, particularly heart rate (HR). It introduces DPNet, a Mamba-based denoising backbone coupled with an auxiliary HR predictor (HRP) and a scale-invariant SI-SDR loss to maintain waveform fidelity and physiological consistency. Across clean BIDMC references and real-world WristPPG motion artifacts, DPNet outperforms conventional filtering and prior neural denoising models in both waveform reconstruction (lower $MSE$) and HR estimation accuracy (lower HR-MAE). The combination of long-range temporal modeling via BMamba, SI-SDR-driven waveform preservation, and HR-guided supervision enables more reliable denoising with practical wearables impact, bridging signal quality and clinical utility.

Abstract

Photoplethysmography (PPG) is widely used in wearable health monitoring, but its reliability is often degraded by noise and motion artifacts, limiting downstream applications such as heart rate (HR) estimation. This paper presents a deep learning framework for PPG denoising with an emphasis on preserving physiological information. In this framework, we propose DPNet, a Mamba-based denoising backbone designed for effective temporal modeling. To further enhance denoising performance, the framework also incorporates a scale-invariant signal-to-distortion ratio (SI-SDR) loss to promote waveform fidelity and an auxiliary HR predictor (HRP) that provides physiological consistency through HR-based supervision. Experiments on the BIDMC dataset show that our method achieves strong robustness against both synthetic noise and real-world motion artifacts, outperforming conventional filtering and existing neural models. Our method can effectively restore PPG signals while maintaining HR accuracy, highlighting the complementary roles of SI-SDR loss and HR-guided supervision. These results demonstrate the potential of our approach for practical deployment in wearable healthcare systems.

Figures (2)

• Figure 1: Architectures of (a) DPNet, (b) Mamba layer, and (c) HRP.
• Figure 2: Denoised PPG signals using different methods.