Authors - Mostafa Salah, Loay Hassan, Somaya Abdel-khier, Ammar M. Hassan, Osama A. Omer Abstract - Recently, there is an increasing interest in approximating the contact PPG measurements through a remote camera-based sensing. However, extracting biomedical signs such as cardiac activities from facial videos is more challenging than contact based measurements. In this paper, the spectral signature of the contact PPG (cPPG) is investigated against the corresponding remote-PPG (r-PPG). In spite of the noisy nature of the video extracted signal, the semi-periodic nature of the pulsating blood volume is properly employed in tracking the spectral signature by periodic filtering around fundamental frequency and the first two harmonics. That salient spectral signature is reflected in better estimating inter-beat interval (IBI) and heart rate variability (HRV). The essential tradeoff resides in estimating the high (featuring detail preserving) beat wave quality against the main signal approximation (timing preserving) quality. Fortunately, IBI and HRV are less sensitive to the quality of the estimated beat wave. They only need sharp timing information of the wave. Our claims are assessed using videos from public database with the corresponding ground truth (GT) PPG. The results reflect robust IBI estimation with high accuracy compared to the state of the art methods.
Open Zoom