Authors:
Nkengue Marc Junior
;
Xianyi Zeng
;
Ludovic Koehl
;
Xuyuan Tao
;
François Dassonville
and
Nicolas Dumont
Affiliation:
Laboratoire Génie et Matériaux Textile (GEMTEX), Université de Lille, ENSAIT, F-59000, Lille, France
Keyword(s):
Signal Processing, Wearable and Mobile Devices, Artificial Intelligence, Health Monitoring Device, COVID-19.
Abstract:
We propose a new low-cost wearable system to guaranty patient mobility and robust monitoring of COVID-19 using physiological signals. Considering the correlation between two key signals (ECG and PPG), the proposed wearable system will integrate an Variational AutoEncoder (VAE) with self-attention block to reconstruct robust ECG, PPG Red and IR signals from a noisy ECG time series. The model performance is evaluated using the Mean Square Error (MSE), the root-mean-square error (RMSE), Mean Absolute Error (MAE) and the Signal-to-Noise Ratio (SNRoutput) for the signals. With a low MSE, RMSE and MAE, as well as good SNR, the model can generate robust and clean data from the noisy ECG waveform measured by the wearable system. we believe that the proposed wearable system can not only help to provide robust online COVID-19 symptoms monitoring but also for other applications.