Speaker: Nicolas Petrochilos, CReSTIC, Université de Reims Champagne-Ardenne (France)
Title: Cardio-Respiratory separation using non-uniform temporal alignment
Location: SiMul meeting room, Faculté des Sciences et Technologies, Henri Poincaré Building, 4th floor
Abstract:
Using a Doppler radar to monitor the heartbeat activity of a patient is not the most straightforward approach. But an important benefit is that it can be operated as a remote sensing operation in rare cases where it is necessary. In doing so, the acquired signal corresponds to the chest displacement, which is the combination of the respiratory signal and the cardiac one. Even assuming that the combination is linear, several challenges remain: The noise power is quite high and not white, the power of the respiratory signal is at least 20 dB higher than the heart, and array source separation is not effective. Fortunately, the heartbeat signal can be described as a repeating template which frequency or phase [1] is randomly slowly varying. Using this information and applying a non-uniform temporal alignment, can some separation be done?
[1] P. E. McSharry, G. D. Clifford, L. Tarassenko, and L. Smith, « A dynamical model for generating synthetic electrocardiogram signals}, » IEEE Trans. Biomedical Engineering, vol. 50(3), pp. 289-294, March 2003.