Assessment of cardiac function of a patient is very important for understanding a patient's physiological state. Remote measurements of the cardiac pulse can provide comfortable physiological assessment by minimizing the amount of wires and cables and allowing for near continuous measurements. It has been found that state-of-the-art algorithms based on independent component analysis (ICA) suffer from a sorting problem which hinders their performance. This effect is demonstrated in this work. The automated pulse detection techniques are applied to RGB color video recordings of the facial region of a person being monitored for cardiac function in a remote sensing environment. Automated face tracking is employed to locate the region of interest and address motion artefacts. This work proposed and evaluates a novel algorithm based on constrained source separation, aka, constrained independent source separation (cICA) to accurately estimate the pulse rate of a patient by solving the sorting problem observed in the ICA based approach. The constrained optimization problem incorporates prior information and additional requirements in the form of constraints. A reference signal with a single tone frequency corresponding to a possible heart rate is fed to the cICA algorithm. This forces the output signal to match the reference signal embodying prior knowledge about an underlying IC. It is also shown that with this algorithm a near photoplethysmography (PPG) signal corresponding to the variations in blood volume in the body can be extracted. An IRB approved study encompassing 45 subjects resulted in Bland-Altman analysis with an FDA-approved finger blood volume pulse (BVP) sensor demonstrating that the proposed algorithm provides significantly improved accuracy.
Library of Congress Subject Headings
Heart function tests--Remote sensing; Signal processing--Digital techniques; Source separation (Signal processing)
Department, Program, or Center
Microelectronic Engineering (KGCOE)
Kyal, Survi, "Constrained independent component analysis for non-obtrusive pulse rate measurements using a webcam" (2013). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus