Multi-Mode Particle Filtering Methods for Heart Rate Estimation From Wearable Photoplethysmography

Area of Science:

• Biomedical Engineering
• Signal Processing
• Wearable Technology

Background:

• Accurate heart rate (HR) estimation from photoplethysmography (PPG) is challenging during physical activity due to motion artifacts (MAs).
• Existing methods struggle to provide reliable HR readings when PPG signals are corrupted by MAs during exercise.

Purpose of the Study:

• To develop and evaluate novel multi-mode particle filtering (MPF) methods for accurate HR estimation from PPG signals.
• To address the limitations of current methods in handling MAs during intensive physical exercise.

Main Methods:

• Proposed four MPF algorithms with variations in particle weighting and HR determination.
• Compared MPF methods against single-mode particle filtering (SPF) and other state-of-the-art techniques.
• Evaluated performance on PPG recordings from two databases during intensive exercise.

Main Results:

• MPF methods achieved an average absolute HR error below two beats per minute, outperforming SPF and other methods.
• Demonstrated robustness even with severe MAs corrupting PPG signals across multiple windows.
• MPF methods require minimal processing time (6.4-6.5 ms within an 8s window).

Conclusions:

• MPF methods significantly improve HR estimation accuracy and reliability during strenuous exercise.
• The developed algorithms are suitable for real-time implementation in practical applications.
• MPF methods show potential for monitoring other time-varying physiological features.