An Effective Photoplethysmography Denosing Method Based on Diffusion Probabilistic Model.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Photoplethysmography (PPG) is vital for health monitoring but susceptible to motion artifacts.
• Existing denoising methods struggle with non-Gaussian noise common in PPG signals.

Purpose of the Study:

• To develop a novel PPG denoising method using diffusion probabilistic models.
• To address the challenge of non-Gaussian noise in PPG signals caused by motion artifacts.

Main Methods:

• A diffusion probabilistic model was adapted to incorporate noisy PPG signals in both diffusion and reverse processes.
• The model was trained and evaluated on a dataset of PPG signals from 15 subjects performing various motion tasks.

Main Results:

• The proposed model significantly reduced the Peak-Rejection-Rate (PRR) from 0.24 to 0.03.
• Enhanced accuracy in heart rate (HR) and heart rate variability (HRV) estimations was observed.
• The model demonstrated robustness and generalization across different motion tasks and subjects.

Conclusions:

• The diffusion probabilistic model effectively denoises PPG signals, mitigating motion artifact impact.
• This approach improves the reliability of PPG for health-related information gathering.
• The method shows promise for enhanced wearable health monitoring devices.