Contactless Heart Rate Estimation from Facial Videos Using Signal Processing and Machine Learning

Abstract

This research presents a contactless heart rate monitoring system that utilizes facial video analysis to estimate heart rate in real time. The system integrates computer vision, signal processing, and machine learning to extract physiological signals from facial regions, process them using Fourier Transform (FFT), and refine heart rate predictions using a trained Random Forest model. Unlike traditional ECGs or wearable devices, this approach eliminates the need for physical contact, making it a hygienic and non-intrusive alternative for healthcare, fitness tracking, and remote patient monitoring. Experimental validation against a pulse oximeter demonstrates reliable accuracy, with a Mean Absolute Error (MAE) of 2.15 BPM and a Root Mean Squared Error (RMSE) of 2.84 BPM, indicating minimal deviation from ground truth measurements. The system maintains stability under varying lighting conditions and minor user movements, further proving its practicality.