Adaptive notch-filtration to effectively recover photoplethysmographic signals during physical activity. (February 2022)

• Record Type: Journal Article
• Authors: Zheng, Xiaoyu; Dwyer, Vincent M.; Barrett, Laura A.; Derakhshani, Mahsa; Hu, Sijung
• Is Part Of: Biomedical signal processing and control. Volume 72(2022)Part A
• Publication Date: 2022-02
• Subjects: Motion artefacts (MA) -- Multi-wavelength illumination optoelectronic patch sensor (mOEPS) -- Adaptive notch-filtration (ANF) -- Real-time signal processing -- Heart rate (HR) -- Respiration rate (RR); 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.2021.103303
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Low-complexity PPG signal recovery algorithm for real-time health monitoring. ANFA to remove in-band and out-of-band motion artefact at various physical activities. Accurate HR and RR values obtained from four stages of two exercise protocols with 24 subjects. Smaller absolute error in HR and RR values using ANFA, than using published methods. Abstract: Physical activity can severely influence the quality of photoplethysmographic (PPG) signals due to motion artefacts (MA). This study aims to extract heart rate (HR) and respiration rate (RR) values from raw PPG signals captured from a multi-wavelength illumination optoelectronic patch sensor (mOEPS) during physical activity of different intensities, and to do this in an effective manner. The proposed method, combined with a 3-axis accelerometer as a motion reference, was developed for the extraction of the desired PPG signals. The adaptive notch-filtration architecture (ANFA) comprises three parts: 1) the adaptive moving average filter, 2) the adaptive notch filter, and 3) extraction for physiological parameters. 24 healthy subjects completed four stages of exercise of increasing intensity. The recovered PPG signals for the calculation of HR and RR were comparable to the measurements from commercial devices, with an average absolute error for HR of < 1.0 beats/min for the IEEE-SPC dataset, and 1.3 beats/min for HR, and 2.8 breaths/min for RR, from the in–house dataset. The ANFA has been proofed to have a good robustness and low complexity to be suitable for application in real-time physiological monitoring.