Influential Factors in rPPG: Insights from a Diverse and Inclusive Empirical Study

Remote photoplethysmography (rPPG) allows to optically measure vital signs, such as heart rate, without physical contact. Signal quality impacts the reliability of derived measurements, but the effects of influential factors in rPPG are not well understood. This research specifically examines five factors that are hypothesized to be important: camera type, skin tone, age, gender, and body mass index (BMI). We investigated these factors using a purposely collected dataset from a comparatively large and diverse population (n=126). For each participant, two simultaneous video streams were recorded using different quality hardware to allow to study the effect of choice of camera type. Statistical analysis based on two quality metrics (signal-to-noise ratio and mean absolute error in heart rate measurements) shows that the choice of camera type is important. Generalized linear mixed models provide evidence of a compounded effect between low quality camera and young age with respect to both signal quality decrease and an increase in measurement error. Analysis of the models coefficients brings evidences that darker skin tone also appears to reduce signal quality, but the results are statistically inconclusive in what concerns heart rate measurements. We observe no significant effect of gender or BMI on rPPG in this study. We believe that comprehensive understanding of the influential factors in rPPG will lead to more reliable and inclusive technologies.