Low-Cost rPPG Application for Real-Time Heart Rate Monitoring: Robustness Analysis of Signal Processing Techniques

A new application has been developed for remote PhotoPlethysmoGraphy (rPPG) tests in real-time using a low-cost laptop or smartphone cameras. This offers a non-contact and convenient solution for heart rate (HR) monitoring, which doctors can use to view all recorded rPPG signals. The application uses a new heart rate estimation method developed from facial videos. It employs skin pixel semantic segmentation and signal processing techniques, including a deep learning model to detect skin pixels from non-skin pixels. The spatial RGB means of the segmented skin pixels are used to extract the rPPG signal, which is then post-processed to remove outliers and calculate the heart rate. The proposed rPPG-based heart rate estimation method has achieved a mean absolute error of ±2.5 bpm, comparable to HR sensors. This low-cost and convenient solution for heart rate monitoring holds promise in medical and biometric applications, providing a new approach to health monitoring.