Robust Estimation of Pulse Rate from a Wrist-type PPG During Intensive Exercise

Area of Science:

• Biomedical Engineering
• Signal Processing

Background:

• Photoplethysmography (PPG) signals from wrist-worn devices are susceptible to motion artifacts (MA) during exercise.
• MA corrupts PPG signals in both time and frequency domains, complicating accurate pulse rate estimation.

Purpose of the Study:

• To develop and validate a novel method for deriving pulse rate from wrist-type PPG signals during intense exercise.
• To improve the accuracy of pulse rate estimation in the presence of significant motion artifacts.

Main Methods:

• Employed Ensemble Empirical Mode Decomposition (EEMD) to decompose the PPG signal.
• Utilized power spectral analysis to extract the pulsatile component and estimate pulse rate.
• Validated the method on an open-access database with PPG and ECG-derived heart rate data from 12 subjects during running.

Main Results:

• Achieved high estimation accuracy with a mean absolute error (MAE) of 2.14 bpm.
• Demonstrated a strong correlation coefficient (0.98) between estimated and ground-truth pulse rates.
• Over 97.5% of estimates were within a 10% margin of the ground truth.

Conclusions:

• The proposed method effectively estimates pulse rate from wrist PPG during intense exercise, outperforming previous approaches without accelerometry.
• The technique shows significant potential for integration into next-generation wrist-worn wearable monitors for sports and clinical applications.