Abstract: Remotely measuring vital indicators, such as heart rate (HR), is possible using facial footage captured by a consumer-level digital camera. However, background light and bodily motions frequently have an impact on how accurately the heart rate is estimated. Therefore, we suggest an anti-disturbance strategy to measure HR at a distance, Eulerian video magnification (EVM), signal quality assessment (QA), and adaptive chirp model decomposition. The performance of the suggested method is then assessed and validated in five distinct situations, including low illumination, normal illumination, high illumination, imbalanced illumination, and head motion. The experimental findings showed a high degree of congruence between the proposed method's HR estimations in various scenarios and the related ground facts. Additionally, whether compared to techniques based on empirical mode decomposition (EMD) or variable mode decomposition (VMD).

Keywords: HR Estimation, Convolutional Neural Network, Spatial Decomposition, Temporal Filtering

| DOI: 10.17148/IARJSET.2023.10566