An Evaluation of Non-Contact Photoplethysmography-Based Methods for Remote Respiratory Rate Estimation

Area of Science:

• Physiological signal monitoring
• Biomedical engineering
• Remote sensing

Background:

• Respiration rate (RR) is crucial for health and emotional state assessment.
• Traditional RR monitoring devices (e.g., respiration belts) have limitations in telemedicine and comfort.
• Photoplethysmography (PPG) offers a less invasive approach for RR estimation.

Purpose of the Study:

• To systematically evaluate and compare PPG-based RR estimation methods.
• To address the lack of standardized evaluation on public datasets.
• To determine the most reliable methods for estimating Respiratory Induced Variations (RIVs).

Main Methods:

• Evaluation of two PPG processing method families: morphological feature extraction and blind source separation (BSS).
• Experiments conducted on the public BP4D+ dataset.
• Comparison against a state-of-the-art Deep Learning (DL) approach for remote RR estimation.

Main Results:

• Morphological feature extraction methods demonstrated higher reliability for RIVs estimation compared to single-channel BSS.
• This finding held true for both contact and remote PPG measurements.
• Morphological methods outperformed a representative DL-based approach for remote RR estimation.

Conclusions:

• Morphological feature extraction is a more robust approach for estimating RR from PPG signals.
• The study provides a benchmark for evaluating future RR estimation techniques.
• Less invasive PPG-based methods show promise for widespread RR monitoring.