PPGMotion: Model-based detection of motion artifacts in photoplethysmography signals. (May 2022)

• Record Type: Journal Article
• Authors: Maity, Akash Kumar; Veeraraghavan, Ashok; Sabharwal, Ashutosh
• Is Part Of: Biomedical signal processing and control. Volume 75(2022)
• Publication Date: 2022-05
• Subjects: Photoplethysmography -- Motion artifacts -- Morphology; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.103632
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A novel method, PPGMotion is proposed to detect motion contamination in PPG signals. The method uses PPG shape as well as quasi-periodic structure to detect motion. The results demonstrate significant improvement in detecting periodic motion. Abstract: Photoplethysmography (PPG) is used widely in health wearables to monitor biomarkers like heart rate. However, motion activities degrade the quality of the measured PPG signal, thereby reducing the accuracy of heart-rate estimation. Existing state-of-the-art methods for motion detection rely on the semi-periodic structure of PPG to detect the aperiodic motion artifacts, thereby failing in scenarios when motion contamination tends to be periodic. We propose a novel technique, PPGMotion, for detecting all types of motion artifacts in PPG signals with high accuracy, without the need for any reference motion signals. Our approach relies on the morphological structure of the artifact-free PPG signal. We compare our method against some classical methods on one synthetic and four real datasets – dataset (1) and (2) are obtained from finger pulse-oximeter under motion activities, dataset (3) and (4) are obtained from a wearable smartwatch. We show that for the synthetic dataset, the performance of PPGMotion is significantly better than existing work as the contaminated PPG tends to become periodic, with an increase in sensitivity of at least 20% over state-of-the-art methods. For real data, PPGMotion achieves similar performance for random motion artifact detection as the classical methods but performs significantly better when motion tends to be periodic, with at least 10% increase in sensitivity in detecting motion artifacts in datasets (2), (3), and (4).