Remote heart rate estimation from palm videos via spatial and channel module integration

Photoplethysmography (PPG) signals are traditionally acquired using contact sensors to monitor heart rate and other physiological parameters. However, prolonged use of contact sensors may cause patient discomfort, leading to an increased interest in non-contact physiological monitoring technologies for their comfort and convenience. Among these, visual-based methods have enabled the extraction of PPG signals from facial videos, offering a noninvasive approach to cardiovascular health monitoring. We extend the application of remote photoplethysmography (rPPG) by focusing on palm videos for heart rate estimation. A palm video database was established for rPPG signal analysis. In addition, a deep learning framework termed spatial and channel module integration was developed, which integrates spatial and channel information to extract more discriminative features and reduce estimation errors in heart rate prediction. The experimental results demonstrate that the proposed method achieves superior accuracy and stability in palm-based heart rate estimation, presenting a promising solution for non-contact physiological monitoring through palm video analysis.