Abstract
Remote heartbeat rate estimation is computation of cardiovascular activities without any physical contact of the individual with the device. Monitoring of heart rate has become increasingly essential especially during the pandemic phase. This state-of-the-art model is expected to have a lasting impact on healthcare practices, thereby estimating the heart rate remotely which can help decrease the potential risk to exposing patients and even medical staff to infection. Thus, it will make monitoring and estimation of heart rate more accessible. This project is an effort to develop a remote heart rate estimation which would aid in keeping a track of heart rate more easily and effectively.
Conventional heart rate measuring device is an Electrocardiography (ECG), which require several electrodes to be attached to different points on the body which is not a practical solution for general daily use. Thus, this application solves that problem of having direct contact of the device with the subject. Moreover, it also provides the subject to monitor heart rate more frequently due to its feasibility.
This state-of-the-art project uses primarily reptile meta learner and transformer. It collects a short video of the subject’s face while performing different activities. These different scenario videos are analyzed and aid in predicting the heart rate when the subject is in different situations. Evaluation of the project has been done on Google Colab using GPU and CUDA using a large dataset of various kinds of subjects.