Non - Contact Heart Rate Monitoring System using Deep Learning Techniques

Individual HR monitoring has diverse applications in fields such as healthcare, sports, and security, among others. Traditional HR measurement techniques can be useful, but they have limitations. This work introduces an LSTM network based on deep learning that can evaluate HR in real-time video streams. After face recognition and object tracking, a collection of face rectangles is generated, which are later analyzed for color differences. A signal corresponding to heart rate is represented by the average hue of a face-selected region of interest (ROI). Using signal processing, we can determine the heartbeat frequency from this information. In addition to being precise, this method is also highly reliable. With this method, we were able to simultaneously identify the faces of four individuals and obtain heart rate readings with an error rate of 3-5%. This method of combining object tracking and facial recognition reduces the required computing power and improves scalability.