Measuring pulse rate variability using long-range, non-contact imaging photoplethysmography.

Camera-based measurement of the blood volume pulse via non-contact, imaging photoplethysmography is a very popular approach for measuring pulse rate using a remote imaging sensor. Comparatively less attention has been paid to the usefulness of the method for measuring features of pulse rate variability, and even less focus has been put on the accuracy of any cardiac activity feature that can be achieved at long imager-to-subject distances. In this study, video was recorded from 19 participants, while at rest, at a distance of 25 meters from the imaging sensor. A digital camera was used to record video while cardiovascular measures of both electrical and optical physiological ground truth were recorded. Pulse rate data obtained from the imager using a common blind source separation and periodogram approach were compared to physiological ground truth signals. The quality of the recovered blood volume pulse morphology was sufficient to calculate time-domain measures of pulse rate using inter-pulse interval (IPI) time series. Following this, several features of pulse rate variability were calculated from the IPI time series and compared to those calculated from the corresponding physiological ground truth signals. Use of the time-domain data as compared to the periodogram approach to measure pulse rate reduced the error in the estimate from 1.6 to 0.2 beats per minute. Correlation analysis (r2) between the camera-based measures of pulse rate variability and ECG-derived heart rate variability ranged from 0.779 to 0.973; these results are of comparable outcome to those obtained at imager-to-subject distances of no more than 3 meters. This study demonstrates that pulse rates of less than one beat-per-minute error can be obtained when the recovered blood volume pulse morphology is of adequate quality to resolve systolic onsets for individual cardiac cycles. Further, this approach can yield data of very promising quality for estimating measures of pulse rate variability.