TY - GEN
T1 - Real-Time PPG-Based HRV Implementation Using Deep Learning and Simulink
AU - Esgalhado, Filipa
AU - Batista, Arnaldo
AU - Vassilenko, Valentina
AU - Ortigueira, Manuel
N1 - info:eu-repo/grantAgreement/FCT/OE/PD%2FBDE%2F150312%2F2019/PT#
PY - 2022/6
Y1 - 2022/6
N2 - The Heart Rate Variability (HRV) signal computation relies on fiducial points typically obtained from the electrocardiogram (ECG) or the photoplethysmogram (PPG). Generally, these fiducial points correspond to the peaks of the ECG or PPG. Consequently, the HRV quality depends on the fiducial points detection accuracy. In a previous work, this subject has been addressed using Long Short-Term Memory (LSTM) Deep Learning algorithms for PPG segmentation, from which peak detection can be achieved. In the herein presented work, a Simulink® implementation of the LSTM algorithm is obtained for real-time PPG peak detection. HRV and outlier removal blocks are also implemented. The obtained code can be used to be embedded in hardware systems for real-time PPG acquisition and HRV visualization. A Root Mean Square Error (RMSE) mean of 0.0439 ± 0.0175 s was obtained, and no significant differences (p-value < 0.05) were found between the ground truth and the real-time implementation.
AB - The Heart Rate Variability (HRV) signal computation relies on fiducial points typically obtained from the electrocardiogram (ECG) or the photoplethysmogram (PPG). Generally, these fiducial points correspond to the peaks of the ECG or PPG. Consequently, the HRV quality depends on the fiducial points detection accuracy. In a previous work, this subject has been addressed using Long Short-Term Memory (LSTM) Deep Learning algorithms for PPG segmentation, from which peak detection can be achieved. In the herein presented work, a Simulink® implementation of the LSTM algorithm is obtained for real-time PPG peak detection. HRV and outlier removal blocks are also implemented. The obtained code can be used to be embedded in hardware systems for real-time PPG acquisition and HRV visualization. A Root Mean Square Error (RMSE) mean of 0.0439 ± 0.0175 s was obtained, and no significant differences (p-value < 0.05) were found between the ground truth and the real-time implementation.
KW - HRV
KW - PPG
KW - Real-Time
KW - Simulink
UR - http://www.scopus.com/inward/record.url?scp=85134353826&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-07520-9_10
DO - 10.1007/978-3-031-07520-9_10
M3 - Conference contribution
AN - SCOPUS:85134353826
SN - 978-3-031-07519-3
T3 - IFIP Advances in Information and Communication Technology
SP - 103
EP - 111
BT - Technological Innovation for Digitalization and Virtualization
A2 - Camarinha-Matos, Luís M.
PB - Springer
CY - Cham
T2 - 13th Advanced Doctoral Conference on Computing, Electrical, and Industrial Systems, DoCEIS 2022
Y2 - 29 June 2022 through 1 July 2022
ER -