Unity in Diversity: Sampling Strategies in Wearable Photoplethysmography

Abstract

Wearable PPG sensors have become a popular way to unobtrusively provide information about the wearer's vital signs, allowing continuous monitoring of fitness and health. Although plenty of research has studied the parameters that such sensors can use to take their readings, we decided in this paper to focus on what measurement strategies can be observed in off the-shelf wearable products. In recent years, the variety of light wavelength used in current wrist-worn devices has settled on mostly using green light since it is robust against motion artifacts and provides a larger signal amplitude. All other parameters that one might consider in sampling PPG data are less settled. We noticed that some commercial devices apply traditional uniform sampling at a fixed-sampling frequency, but many other wearables follow different strategies as a tradeoff to achieve a good signal while using a minimum of energy. They tend to follow different variations of multichannel sampling, average sampling, and burst sampling strategies at various frequencies, yet only part of this diversity can be explained by the need for different qualities of readings. Even though PPG has been integrated into many commercially available wrist-worn wearable devices, their sampling strategies have thus yet not converged to a single solution.