Abstract
Remote Photoplethysmography is the methodology for calculating the beats per minute of subjects using the contactless method. Predicting heart rate has become an important role in SARS-CoV-2 (Covid 19) pandemic. The existing approach for remote heart rate tracking has several challenges in unseen subjects with a large frame rate and sequence of videos. For many analytical goals, a combination of clinical and claims is used. The project aims to provide a heart failure prediction using a new deep learning algorithm with an interpretable predictive model for healthcare applications.Our model addresses the challenge by developing a meta-transformer model, a deep learning model for computing the heart rate of humans with contactless measurements. In our meta-transformer project, we proposed a contactless heart rate and blood volume pulse measurement by adapting a state-of-art performance model. Our method uses a sequence of videos and performs effectively by generating performance metrics better than other deep learning models. We implemented a method to improve the health prediction accuracy using TensorFlow and PyTorch. Our project will enable heart rate monitoring more accessible and convenient for daily life. We evaluated our model with two benchmark datasets namely UBFC and LGI-PPG datasets.