A Non-contact heart rate estimation framework based on photoplethysmography amplitude variation elimination and data fusion

The evaluation and assessment of Heart Rate Variability (HRV) from contact-based methods have had quite a journey throughout the annals of medical diagnosis. However, non-contact measurement of heart rate (HR) is gaining lots of attention recently. In this paper, we introduce a framework to make a robust contact-less estimation of HR from facial recorded videos. Our framework exploits various color space representations of the video frames and different techniques to extract initial Photoplethysmography (PPG) signals. The acquired PPG signals are then enhanced using the novel PPG amplitude variation elimination technique (AVET) embedded in our framework which is based on their analytic signals. Various HR is estimated from the refined PPGs in order to be fed to the fusion algorithm which is intended to procure a more accurate estimation of the true HR. The evaluation on the publicly available UBFC- Phys dataset shows that the proposed framework has a superior performance compared to the baseline techniques such as ICA and Green methods.