Evaluation of video magnification for nonintrusive heart rate measurement

Abstract

Measurements of physiological signals have been used in many areas, including medical, sports, and automotive applications. Cardiac signals such as electrocardiogram (ECG) and heart rate variability (HRV) are particularly important, but conventional measurement techniques require wired devices and static subjects. These limitations preclude application of these devices from various dynamic scenarios. Recently, however, developments in computer vision have shown that some physiological variables related to the heart can be measured in a nonintrusive way. "Video magnification" (VidMag) is one such technique, and it has been used to measure blood volume pulse (BVP) from face video data in laboratory settings. This paper discusses the readiness of VidMag for psychophysiological assessment more generally. The approach is to assess beat-by-beat comparison of VidMag with BVP signals extracted from medical devices. In addition, a novel skin detection algorithm, which does not use color cues, has been proposed in this paper as a preprocessing step to support VidMag. The paper also presents a systematic post-processing strategy using Savitzky-Golay filtering to improve the accuracy of the raw output from the video magnification algorithm.