IoT Technologies for HealthCare. 6th EAI International Conference, HealthyIoT 2019, Braga, Portugal, December 4–6, 2019, Proceedings

Research Article

A Real-Time Algorithm for PPG Signal Processing During Intense Physical Activity

  • @INPROCEEDINGS{10.1007/978-3-030-42029-1_2,
        author={Andrea Gentili and Alberto Belli and Lorenzo Palma and Salih Egi and Paola Pierleoni},
        title={A Real-Time Algorithm for PPG Signal Processing During Intense Physical Activity},
        proceedings={IoT Technologies for HealthCare. 6th EAI International Conference, HealthyIoT 2019, Braga, Portugal, December 4--6, 2019, Proceedings},
        proceedings_a={HEALTHYIOT},
        year={2020},
        month={6},
        keywords={PPG monitoring HR estimation Motion artifacts},
        doi={10.1007/978-3-030-42029-1_2}
    }
    
  • Andrea Gentili
    Alberto Belli
    Lorenzo Palma
    Salih Egi
    Paola Pierleoni
    Year: 2020
    A Real-Time Algorithm for PPG Signal Processing During Intense Physical Activity
    HEALTHYIOT
    Springer
    DOI: 10.1007/978-3-030-42029-1_2
Andrea Gentili1,*, Alberto Belli1, Lorenzo Palma1, Salih Egi2, Paola Pierleoni1
  • 1: Università Politecnica delle Marche
  • 2: Galatasaray University
*Contact email: a.gentili@pm.univpm.it

Abstract

Photopletismography (PPG) is a simple, low cost and noninvasive technique, implemented by pulse-oximeters to measures several clinical parameters, such as hearth rate, oxygen saturation (Spo), respiration and other clinical diseases. Although monitoring of these parameters at rest does not present particular problems, processing PPG signals during intensive physical activity is still a challenge, due to the presence of motion artifacts that affect its true estimation. In our work, a novel time-frequency based algorithm is presented to properly reconstruct PPG signal during intensive physical activity with respect to the ECG signal reference. Starting from raw PPG and acceleration signals, the proposed algorithm initially removes motion artifacts, providing an accurate heart rate estimation. Subsequently, it reconstructs PPG waveform based on both the heart rate information previously computed and the optimal selection of frequency-domain components representing PPG signal. Evaluating our proposed method on a dataset containing signals acquired during high speed running, we found for heart rate estimation an average absolute error of 1.20 BPM and very similar frequency dynamics between the ECG reference and PPG reconstructed HRV time series from a physiological point of view based on visual inspection.