The photoplethysmogram (PPG) is a noninvasive cardiovascular signal related to the pulsatile volume of blood in tissue. The PPG is user-friendly and has the potential to be measured remotely in a contactless manner using a regular RGB camera. In this dissertation, we study the modeling and analytics of PPG signal to facilitate its applications in both robust and remote cardiovascular sensing.

In the first part of this dissertation, we study the remote photoplethysmography (rPPG) and present a robust and efficient rPPG system to extract pulse rate (PR) and pulse rate variability (PRV) from face videos. Compared with prior art, our proposed system can achieve accurate PR and PRV estimates even when the video contains significant subject motion and environmental illumination change.

In the second part of the dissertation, we present a novel frequency tracking algorithm called Adaptive Multi-Trace Carving (AMTC) to address the micro signal extraction problems. AMTC enables an accurate detection and estimation of one or more subtle frequency components in a very low signal-to-noise ratio condition.

In the third part of the dissertation, the relation between electrocardiogram (ECG) and PPG is studied and the waveform of ECG is inferred via the PPG signals. In order to address this cardiovascular inverse problem, a transform is proposed to map the discrete cosine transform coefficients of each PPG cycle to those of the corresponding ECG cycle. As the first work to address this biomedical inverse problem, this line of research enables a full utilization of the easy accessibility of PPG and the clinical authority of ECG for better preventive healthcare.