SpO$_2$ Predictor-Guided Stage-Wise Time-Frequency Reconstruction of Low-Quality Dual-Wavelength PPG for Oxygen Saturation Estimation

2026-07-08

AI Summary

This study addresses the degradation in SpO₂ estimation accuracy caused by waveform distortion in low-quality dual-wavelength photoplethysmography (PPG) signals. To mitigate this issue, the authors propose a physiology-informed, stage-wise time–frequency joint reconstruction framework that uniquely integrates the SpO₂ prediction objective directly into the PPG signal reconstruction process. The approach combines time-domain waveform loss with short-time Fourier transform (STFT)-based frequency-domain loss and leverages a pre-trained SpO₂ predictor as a physiological constraint to guide reconstruction. A four-stage training strategy is employed to ensure preservation of SpO₂-relevant physiological information. Evaluated on both public and private datasets, the method achieves state-of-the-art performance with subject-level mean absolute errors (MAE) of 2.882% and 2.359%, respectively.

Abstract

Continuous oxygen saturation (SpO$_2$) estimation from wearable photoplethysmography (PPG) is important for long-term health monitoring, but low-quality red and infrared PPG segments can distort waveform morphology and degrade SpO$_2$ prediction accuracy. Existing PPG denoising and reconstruction methods usually optimize waveform fidelity or heart rate characteristics, while time-domain waveform loss on PPG signals alone insufficiently preserves frequency structure and SpO$_2$-relevant information. This paper proposes a SpO$_2$ predictor-guided stage-wise time-frequency reconstruction framework for low-quality dual-wavelength PPG signals. The proposed method first selects high-quality PPG segments to pretrain a SpO$_2$ predictor. A masked reconstruction model is then trained to recover randomly masked PPG regions using a joint reconstruction objective that combines time-domain waveform loss with frequency-domain loss computed from the short-time Fourier transform (STFT). To make the reconstruction task physiologically relevant, the pretrained SpO$_2$ predictor is incorporated as an additional constraint, encouraging the reconstructed PPG to preserve SpO$_2$ information rather than only minimizing waveform reconstruction error. The SpO$_2$ predictor and PPG reconstructor model are optimized through four training stages. Experiments on the public OpenOximetry Repository and a private wearable PPG dataset show that the proposed approach achieves the lowest subject-level MAE, with 2.882\% on the public dataset and 2.359\% on the private dataset.

Problem

• SpO2 estimation
• low-quality PPG
• dual-wavelength PPG
• time-frequency reconstruction
• wearable monitoring

Innovation

• SpO2-guided reconstruction
• time-frequency PPG denoising
• dual-wavelength PPG
• physiological constraint
• masked signal reconstruction