Recovering Physiological Signals From Facial Videos: Recent Advances and Applications in Intelligent Vehicles

Abstract

In recent years, non-contact physiological signal monitoring based on facial video has garnered significant attention due to its convenience and low cost. Unlike traditional methods for physiological signal monitoring, which necessitate complex equipment and stringent monitoring conditions, remote photoplethysmography (rPPG) technology relies solely on a camera to recover photoplethysmography (PPG) signals and analyze a broad spectrum of physiological metrics. This approach can be easily integrated into existing sensors in smart vehicles, enabling in-vehicle occupant status monitoring. In this paper, we conduct a comprehensive review of current research progress in detecting physiological signals using rPPG technology, specifically focusing on smart vehicles. This includes benchmark datasets, video preprocessing methods, unsupervised, supervised, and self-supervised signal restoration techniques, as well as post-processing methods applied to the signals. We also provide a performance summary of all these methods across various datasets. Additionally, we delve into the primary applications of rPPG technology in intelligent vehicles and highlight the current challenges. Finally, we conclude with a discussion on future research directions in this area to facilitate broader application of rPPG technology in the field of intelligent vehicles.